Revisions

This page lists every change made to this site since it was first built, newest first. Each entry states what changed and what the change rests on. Where a revision closed a gap that the site had previously recorded as open, that is stated too, because the gap statement itself was a published finding.

14 August 2026 — twenty-first build, the defects move into the vessel's own history, and a fifth newspaper

Eight documents, corpus 338. Yesterday's build put the leaked defect records into the August reporting. That was the wrong place for half of them, and this build fixes it.

The records now sit where the defects happened

The dates in those documents are 2024, 2025 and early 2026. They belong in Part I of the chronology — the vessel's own institutional record — alongside the 2015 inspection that called her deck "a mass of junk and rust", the 2017 decision to rehabilitate rather than replace, and the March 2026 hull tender. They are now there, at the dates on which each defect was recorded, each marked [REPORTED AUGUST 2026] so that no reader mistakes when it happened for when it became known.

What that placement shows, which the August-only version did not: bilge pump valves jammed from January 2024 on one account; a generator out of service and a heat exchanger at 197–203°F at the end of November 2025; the radio logged as completely inoperable on 22 December; fuel lines "heavily corroded" on 21 January 2026 — and then, on 9 March 2026, the state putting a further G$124.5m into the hull by tender. Those last two entries are now six weeks apart on the same page. This archive draws no conclusion from that and neither should anyone else — a hull tender and an engine defect log are different documents about different things, and nothing establishes that anyone reading one had seen the other.

The same entries are now on the landing-page strip as a vessel-phase development, which is where a reader looking at the ship's history will find them.

A fifth outlet, and why it was missed

Kiskadee Watch published its own account of the records on 12 August, alongside three of the four this site had. It was missed because every attempt to fetch that outlet on 13 August timed out, and this archive wrote the failure down as settled — for the second time in two days, after the identical error about Village Voice News. Re-tested on 14 August, it fetched immediately.

What was sitting behind that failure: the fifth account of the defect records; the verbatim MARAD statement; the only account in this corpus of a crew member's family being warned about the vessel; the only passenger description of the voyage; and the GDF helicopter question. The rule from here is that no retrieval failure is carried forward without being re-tested at the next sweep.

A correction of a correction

On 13 August this archive shortened MARAD's quotation to "create public mischief", on the ground that "designed to" was the retrieval tool's connective prose. Kiskadee carries the sentence verbatim and it is longer than the archive allowed: "These reports are false and misleading and are designed to create public mischief." The removal was itself an over-correction. The fuller quotation is restored across the site. That is now the second over-correction caught in two days, and both were caught by re-reading a source rather than by reasoning about it.

What the crew's own families were saying, before any document leaked

Royston Frank, father of Raul Frank, 33, a T&HD sailor still unaccounted for, told Kiskadee Watch on 10 August — two days before the first document story — that his son had repeatedly warned the family about mechanical problems, that the vessel "had experienced engine problems on numerous occasions, including an incident earlier this year when its engine reportedly burnt out", that these were "well known among those who worked on the vessel", and that the 18 July voyage was to be his last before he sought a transfer. It is a bereaved father's recollection of private conversations and nothing can corroborate it. It is recorded because it reaches the question the documents reach, by a different route, and earlier.

Also new

The Commission of Inquiry's own progress report, via DPI on 13 August and Guyana Times on 14 August: arrangements for public hearings are being "finalised", and twenty-seven days after the sinking there is still no hearing date, no notice, no budget, no instrument number and no call for submissions. The Santa Rosa Village Council becomes the second Indigenous village council to demand the vessel be raised. Claude Benjamin, 64, of Sebai gives the first sustained passenger account of the 24-hour run in this corpus. And Kiskadee records that two of the GDF's three helicopters were in maintenance during the rescue, with a dated press query about their current status that went unanswered.

The verification pass, and what it caught

An adversarial pass over this build found twenty-four defects, all now fixed. Three are worth stating publicly because they are failures of method rather than of typing.

A scripted edit amputated four words from an unrelated entry and glued them onto a new one. The 2010 entry in Part I lost "on the North West route" and the new Part I note gained it, mid-sentence, after a full stop. Two broken sentences on the most-read page of the site, produced by a find-and-replace that matched a prefix. A second splice of the same kind put the 14 August clause on the method page after the sentence it was meant to go inside.

The chronology was left on four outlets while every other page moved to five. The 12 August entry's own headline said five newsrooms and named Kiskadee Watch; its body then said "four accounts, two stated acquisitions", attributed items to three outlets that four carry, and stated in the same paragraph that Kiskadee Watch could not be fetched. This site tells readers the record pages govern the analysis pages. The record page was the one that did not get updated.

And the absence claim overreached. Two pages said a search "across every Guyanese outlet in this corpus" found no government response. The state-owned Guyana Chronicle — 41 documents here, and the outlet most likely to carry a government statement — could not be fetched at all. The claim now carries that limit wherever it appears. The commit message for this build says "this time the archive can say it looked"; it could say that of most outlets and not of the one that mattered most.

Also corrected: the landing strip's new vessel entry was inserted out of chronological order, which is the fault this archive has now had to fix four times; nine Kaieteur-only defect entries in Part I were unflagged in breach of this site's own single-source rule; the ballast-tank valves were filed under a date the reporting does not give them; and the "designed to create public mischief" quotation was cited to three records, none of which carries it — the one that does, Kiskadee's, was not in the citation.

One passage was removed for prejudice. The 22 December entry recording the VHF set as inoperable had carried an instruction to "read against Part II", where the distress-call timing is disputed, followed by a disclaimer that no connection was being made. An archive cannot instruct a juxtaposition and then disown it, and the statement being imported was about the conduct of a man charged with murder and not yet tried. The cross-reference and the imported quotation are gone; the defect entry stands on its own.

What has not happened

Forty-eight hours after the first report of the defect records, no government statement addresses them. A search on 14 August across every Guyanese outlet in this corpus, the Department of Public Information and the Office of the President found nothing — and this time the archive can say it looked. The salvage tender closed on 14 August and no report of the proposals received has appeared. The Guyana Chronicle remains unfetchable behind HTTP 429 and a JavaScript wall; two of its 5 August reports are known by headline and are deliberately not in the corpus.

13 August 2026 — twentieth build, nine days missed, and a claim the state cannot now sustain

Twenty-seven documents, corpus 330. This is the largest single addition since the first week, and it exists because the archive stopped sweeping on 4 August and the story did not stop moving. Nine days of silence in a record whose whole method is daily capture is a failure, and the size of this entry is the measure of it.

What the record now holds that it did not

Internal maintenance records for the vessel, reported by four outlets — five, as it turned out the next day. Kaieteur News, News Source Guyana and Village Voice News published on 12 August; HGPTV on 13 August. Kaieteur and HGPTV each say they obtained the records; News Source says only that the documents were "seen by News Source"; Village Voice states no provenance. Their date ranges differ — Kaieteur August 2025 to March 2026, News Source 17 November 2025 to 9 March 2026, the others November 2025 to March 2026 — and so do their details, which is what overlapping copies of one file look like. Common to all four: a generator out of service, a heat exchanger running hot, an exhaust leaking into the engine room — though even there the details part company, Kaieteur dating the water-pump damage to 31 August 2025 where three others attach it to 30 November, and News Source alone declining to locate the heat exchanger. The 197–203°F figure comes from three of the four; the panel burnout at Port Kaituma on 28 November from three, and not from Kaieteur. Kaieteur alone reports a VHF communication system recorded on 22 December as completely inoperable, a damaged life raft cradle, fuel lines described on 21 January 2026 as "heavily corroded" and liable to "break at anytime while the vessel is in operation", a compiler named as Shore Chief Engineer Corwin Stephen, and the column for completed work blank throughout.

Set against Minister Edghill on 20 July: "There was no report of any kind that suggested the vessel was unseaworthy." That was always an absence-of-reports claim rather than a survey finding, and this site has said so since the first build. It is the one kind of claim a leaked report can break. As of this build no government statement addressing the records appears in anything this archive has retrieved — and on the same day it could not fetch the Guyana Chronicle or Kiskadee Watch at all, so that is a limit of retrieval and not a fact about the government. This archive holds none of the documents and cannot check a date, a signature or a word. A defect log is not a survey, and a blank column is not proof that no repair was made. Nothing here establishes a cause of the sinking, and Captain Price is described as a signatory to defect reports, which is a fact about paperwork and, if anything, the opposite of concealment. He and his co-accused are untried.

The salvage procurement conflict. A proposal attributed to a Dutch firm, dated 28 July, states the firm "has been invited by the Government of the Co-operative Republic of Guyana, through the Maritime Administration Department" — five days before the tender opened. On Friday 7 August MARAD, silent on the vessel since 24 July, said "no individual, company, or entity has been selected", that the process began 2 August, and that reports to the contrary are "FALSE and MISLEADING". Selection and invitation are different acts and only one has been denied. Village Voice News added a third date independently: MARAD's own Facebook page shows the notice issued 1 August.

The Commission of Inquiry, in its own words. On 9 August, ten days after being sworn in, chairman Justice Godfrey Smith: "We are still in preparatory stages, meetings…for now we are still organising ourselves." The day before, the President had answered demands for the minister's removal with "let the CoI work."

The end of the search, with figures. On 4 August the Office of the Prime Minister gave 73 dead, 76 survivors, about 30 missing, 5 unidentified; an inter-agency after-action review followed on 5 August and nothing of its content has been published. Also new: the Guyana Trades Union Congress's first intervention — Lincoln Lewis, "Accidents don't happen; they are caused"; an APA picket of about 200 outside the Office of the President on 12 August; the family of two missing brothers picketing the same building the same afternoon; and the Attorney General saying on 7 August that "more may be charged for all you know", which no one in government had said before.

Three corrections, and the first is the worst

1. A single-source flag that was wrong within a day. The first sweep of 13 August recorded the maintenance records as Single source to Kaieteur News. A second sweep the same day, deliberately aimed away from Kaieteur, found three other outlets holding them. The flag was not wrong because the evidence changed; it was wrong because the first sweep did not look. It has been withdrawn and the record now carries the correction on its face. This is the fourth time this archive has published a claim about what other outlets do not have without checking; the rule from here is that no single-source flag ships until a sweep has been run against it.

2. "No figures given at all," and a day out. The archive recorded, from Kaieteur's report of 5 August, that the search was called off without any figures being given. Wrong twice: the announcement was made on 4 August, and figures were given — News Room carried them that day, iNews Guyana and the regional wire the next. A single outlet's omission was written down as the state's silence, and a day's slippage travelled with it. Withdrawn.

Also withdrawn: an absence that this sweep itself destroyed. The chronology had recorded that nothing in the corpus reported the 3 August court hearing. A regional report of that date, retrieved on 13 August, carries it in full: the three appeared by Zoom before Magistrate Ravindra Mohabir, the prosecutor confirmed a fiat and made partial disclosure, and the matter was adjourned to 7 September. The stale entry had survived on three pages.

3. "Village Voice News: robots-disallowed… therefore wholly unrepresented." That note had stood on the method page for a fortnight. The outlet fetched without difficulty on 13 August and three of its reports are now in the corpus, including the one that supplied the MARAD Facebook date. One failed fetch had been left standing as a settled fact about a publisher, and a Guyanese newspaper covering this story heavily was missing from the record for two weeks because of it. Retrieval failures are re-tested from now on rather than carried forward.

What could not be captured

The Guyana Chronicle returned HTTP 429 on every attempt; one of its reports is held only through a syndicated republication and is marked a partial capture, and two others are known by headline and are deliberately not in the corpus. Kiskadee Watch timed out on robots.txt every time; at least six of its reports in this window exist and none could be taken. Both are logged on the method page.

Also worth recording

After 5 August this archive can find no regional or international coverage at all — the Gleaner, Nation News, CBC Barbados, Caribbean National Weekly, Caribbean Today, Loop, Reuters, the Associated Press and the BBC all return nothing dated 6–13 August. The story is now being carried in Guyana alone. That is an absence and is not cited as a finding anywhere on this site; it is recorded because it changes who is left asking.

4 August 2026 — nineteenth build, a second submission, the terms of reference, and an over-correction

Three documents, corpus 303. Two of this archive's longest-standing gaps closed, and it made and unmade a correction inside one build.

The gaps that closed

On 4 August the President published the Commission's terms of reference in the Extraordinary Gazette — seventeen days after the sinking, and the first time they have existed as a document rather than a paraphrase. Six heads: seaworthiness and safety; passenger and cargo operations; crew qualifications and conduct; search and rescue effectiveness; regulatory oversight by state entities; and human factors in emergency decision-making. A two-month deadline for final written findings — the first deadline of any kind in this record. Public hearings, a power to issue interim safety reports, and explicit authority to refer evidence of criminal negligence, official misconduct or statutory breaches directly to prosecutors. The inclusion of regulatory oversight answers part of Christopher Ram's objection of 28 July. Still absent: a Statutory Instrument number and a budget.

A Secretary was named the same day — attorney Javed Shadick — closing a gap Demerara Waves recorded as open on 30 July. And within hours David Patterson objected that he holds a function in the Guyana Civil Aviation Authority, which sits under the Ministry of Public Utilities and Aviation. That is the ministry MARAD answers to, on MARAD's own salvage notice of 1 August, and MARAD's regulatory oversight is one of the six heads. That last step is this archive's inference and is marked as such, not Patterson's. No government response appears, and it is the second set of conflict questions about this Commission to go unanswered in five days.

The second submission

Robert C. Millington sent a second analysis, 52 pages dated 3 August, published here in full. It widens from maritime law to sea and river, road, rail and air, and schedules 52 deficiencies, 31 marked Critical. Its sharpest instrument is the road comparison: Guyana has regulated driver fatigue since 1940 and no hours-of-duty or watchkeeping provision for a vessel has been identified; drivers face a prescribed alcohol limit, breath testing and disqualification, masters none; causing death by reckless driving carries ten years and there is no offence of causing death by reckless navigation; a Government road vehicle must hold a certificate of fitness and is exempted only from the fee, while a State vessel is exempt from the Certificate of Seaworthiness altogether.

He also identifies a reform available without Parliament: deleting the five words "vessels belonging to the State" from reg. 6(2)(a) of the Harbours Regulations 1939, which he calls "the single highest-value immediate reform available in the whole statute book". And a second route into survey that does reach State vessels — the Tax Act vessel licence, which cannot issue without a survey certificate dated within twelve months, and which contains no State exemption though the same Act exempts Government carts, mules and steam boilers. He states it "was open on 18 July 2026".

One caution deserves its own line. The 2022 Revised Edition of the Laws of Guyana was presented to the National Assembly on 19 June 2026 — twenty-nine days before the loss. Both his analyses rest on the 2012 edition, which is the one the Ministry of Legal Affairs still publishes.

The over-correction, which is the entry worth reading

The second submission sets out Cap. 49:03 s. 7(1), under which the master must declare in writing the number, names, sex and age of every person on board before clearance, with a year's imprisonment for a false declaration. This archive read that as contradicting the finding it had published from his first submission — that no counting duty exists in Guyanese law — and withdrew it, in capitals, on three pages.

That was wrong, and the withdrawal has itself been withdrawn. The same document, read whole, reconciles the two. Deficiency 18, marked Critical, repeats the earlier premise unchanged: the power to require particulars of all passengers "has existed since 1998 and has never been exercised". And his reform section still proposes creating a duty of passenger accounting — "a physical count reconciled against tickets before the brow is landed, and a signed and timed manifest lodged ashore before departure and preserved". A man does not recommend creating a duty he has just said exists.

The two provisions are about different things, and the distinction is the substance: s. 7(1) is a customs pre-clearance declaration; what SOLAS III/27 requires and s. 398(1)(b) never delivered is a headcount kept ashore. On his own account the form the declaration depends on has not been identified.

Every failure logged on this site so far has been an absence asserted without a complete search. This is its mirror image — a correction made without finishing the reading, which destroyed a sound finding instead of an unsound one. It is the more dangerous of the two, because it wears the costume of rigour. The rule now stated on the method page: one qualified person's reading is not corroboration of itself, and a second reading by the same person is not necessarily a correction of the first.

One conflict that is real, and recorded as unresolved

The same author now places the vessel inside Part X of the Guyana Shipping Act — "a mechanically propelled steel passenger ferry in commercial service falls within Part X" — which carries an annual survey duty and a requirement that life-saving appliances comply with the Safety Convention. His first analysis turned on a 24-metre band that the second document does not mention anywhere. He does not say he has changed his mind. Both readings are now recorded on the figures and analysis pages and neither is preferred.

Chronological order, rechecked across all three timelines

A recheck on 4 August found the fault logged on 30 July had recurred, and in a place nobody was looking.

The landing page's developments strip ended backwards — 4 August, then 1–2 August, then 31 July. Each new turn had been inserted before the previous tail entry rather than after it, so the three most recent entries on the site's most-read page ran in reverse. Corrected; the strip now ends 31 July, 1–2 August, 4 August.

Two smaller breaks in the chronology page's early parts, both from the same habit of inserting rather than appending: a note about conditions before departure sat below the 15:15 departure it preceded, and a question covering the 19:00–23:00 window sat below a 20:00 event inside that window. Both moved.

And the Part IV heading still read "19–30 July 2026" while the part itself ran to 4 August.

Checked and clean: Part I (10 dated entries), Part IV (49 dated entries, strictly ascending), the counterfactual page's nineteen-turn strip, and the overnight sequence in Parts II–III, where 01:36 and 03:15 correctly follow 23:01 across midnight. No entry was altered by any of the reordering, on either occasion.

Also fixed

Sixteen defects. The terms-of-reference closure had not propagated to findings 12 and 19, to the questions page, or to the landing page, which still told readers there were no terms, no secretary and no deadline. The site's revision stamp had not advanced. A commissioner was described as newly named when he was named on 26 July. An inference of this archive's was folded into Patterson's objection. "Six days" was five in four places. The severity split of the 52 deficiencies was given as "roughly half" when it is 31 of 52. And the 3 August Charity Magistrate's Court date passed with nothing in the corpus either way, which the chronology now records as a gap in retrieval rather than a gap in proceedings.

2 August 2026 — eighteenth build, the state moves to raise the ship

Five documents, corpus 300. One of them changes the biggest open question on this site.

The wreck

On 1 August MARAD, under the Ministry of Public Utilities and Aviation, issued an Invitation for Expressions of Interest for marine salvage contractors to recover the Barima, closing 14 August. Bidders must set out an approach protecting "forensic evidence for the ongoing Commission of Inquiry". The notice gives the wreck's position, a depth of 13–15 metres, a length of 38.6 metres and twin Caterpillar C18 engines, requires "no cure, no pay" terms, and reserves MARAD's right to cancel the process entirely.

The memorial plan is gone. The proposal that went to bereaved families as a draft declaration on 25 July was dropped seven days later — through a tender notice. No government statement in this corpus acknowledges the change of position or mentions the memorial proposal at all. Kaieteur News names the reversal and attributes it to "mounting pressure from stakeholders across Guyana"; that is a newspaper's account, not the state's, and it is recorded as one. Finding 3 is rewritten around it, and says plainly that the abandonment is an inference from a procurement document rather than anything the state has said.

The opposition has moved from whether to how fast. APNU's Ganesh Mahipaul argues the process pushes salvage to late August or September and calls for emergency procurement: "That timeline is simply unacceptable." His sharpest point is that the evidence the notice exists to protect is decaying while the process runs — "Salt water is unforgiving." No government response appears.

Also added: Kiskadee Watch's 30 July account of the former master Andrew Donald, requested by a reader. It carries his allegations at greater length than HGPTV's — twelve years at the T&HD, "a floating disaster waiting to occur", chronic overloading, bilge systems he says could not control flooding. It corroborates the HGPTV report; it is not a second witness, since both rest on Donald. Nothing in it is verified and this archive adopts none of it.

What the verification pass caught, and the one worth publishing

Eighteen defects. Four are worth stating because they are the kind this archive keeps making.

An absence falsified by the document added in the same sweep. Six pages carried "no other outlet in the corpus carries him" about Donald — while this build was adding the second outlet. One of those sentences sat thirty lines below a new entry saying the opposite, in the same file. Fixed in all six; this is the eighth or ninth time a correction or development has failed to travel between pages, and the argument for testing claims mechanically rather than remembering them is now overwhelming.

A novelty claim that the same page disproves. The build asserted that MARAD's notice was "the first document in this corpus in which the state itself describes the hull as evidence." It is not: the Prime Minister said on 28 July that the wreck "likely contains additional remains and investigative evidence" — and this site quotes him saying it, four paragraphs earlier in the same finding. What is actually new is that the state is now paying to recover it. Corrected.

A conflict this archive manufactured. The build flagged MARAD's 13–15 metres as contradicting the "three to nine metres" reported in the first week. It does not: that range was given for the Iron Punt area generally, "depending on the vessel's distance from the shoreline", and reconciles itself in its own sentence. The flag is withdrawn. The 38.6 m against 40.26 m divergence is real and stands.

A rejection that never happened. Two pages said the government had rejected the Opposition Leader's offer to fund a salvage. This archive's own record is that he received no reply at all — "No update, nothing at all." Corrected in both places.

Also fixed: Christopher Ram listed among those vindicated on the hull when the same finding records his demand as going to documents rather than the hull; a claim that the state gave no reason for the salvage when the notice states a purpose; "eight days" for a seven-day gap; the landing page's developments strip and lead cards, which still presented the memorial as the live position; a stale "thirteen days" on the landing page; the site's revision stamp, which had not been advanced; four stale figures on the method page; and a disclaimer on the Donald entry saying his account bears on nothing about the charged men, when it plainly bears on them — it is an argument that others controlled loading. It is now recorded as an untested allegation rather than disclaimed away.

Elapsed-time claims moved from fourteen days to fifteen, except on the counterfactual page, which covers a fortnight ending 1 August and whose counts were wrongly advanced past its own stated horizon.

1 August 2026 (late) — seventeenth build, ten documents, and a false silence withdrawn

A sweep of 1 August reporting added ten documents, taking the corpus to 295 and extending the record to today. Four are dated 1 August, two are 31 July reports the previous sweep missed, and four are state releases of 26 and 28 July that this archive had recorded as not existing.

The correction first, because it is the largest thing in this build

This site has said since 26 July that the Department of Public Information published seventeen releases dated 19–25 July, then four days of nothing, then one release on 30 July. That was false. DPI published on 26 July — the day the commissioners were named — and three times on 28 July, the day the charges were laid. The four days of silence did not happen. What happened is that this archive's sweeps of 26–29 July did not reach dpi.gov.gy, and it recorded a gap in its own searching as a gap in the record.

The claim is withdrawn on the method page, where the corpus now records twenty-two DPI releases dated 19–30 July and makes no claim to have found all of them.

It also took a finding with it. Finding 10, What stopped being said, was built on that silence: the state going quiet exactly as the questions turned legal. Most of that finding is gone. What survives is narrower and still real — the single figure showing how the manifest failure was distributed between the living and the dead was volunteered on day one, refused by the Prime Minister on day six, and has never been given since by anyone.

And the correction cuts the other way on the toll. The newly held DPI release of 28 July gives 73 recovered, 68 identified, five awaiting identification, 30 unaccounted for, 76 survivors — so the 73 that DPI itself abandoned without explanation on 30 July was DPI's own published figure two days earlier.

This is at least the eighth absence claim this archive has published and withdrawn. The pattern is always the same, and so is the rule: an absence claim is only as good as the search behind it.

What the day itself brought

A second investigative body, and it is not the Commission. President Ali has established a Safety and Compliance Audit Team — Dr Ivor English, a former Director-General of MARAD, with Col. Julius Skeete of the GDF and Capt. Joseph Lewis, Marine Officer and Surveyor of Ships — to review safety and compliance standards for marine vessels generally. Carried by both Kaieteur News and Guyana Times, so not single-source. It arrives, exactly as the Commission did, with no deadline, no budget, no terms of reference published as a document and no undertaking to publish, and nothing states how the two bodies relate.

The rescue timeline turns out to exist. The DPI release of 28 July carries the GDF Chief of Staff, Col. Kenlloyd Roberts, announcing an after-action review of the search and rescue operation covering communications between the control tower, the lighthouse, the Search and Rescue Coordination Centre and Coast Guard operations, from the distress signal to the end of the mission. Finding 15 is rewritten accordingly. The position moves from nothing here says such a record exists to the state has said it will compile one and has said nothing about who will see it — no deadline, no named recipient, no undertaking to publish, and no statement whether it goes to the Commission.

The President is reported to have committed to raising the ship. Single source Relatives of three missing Berbicians say that at a meeting on 29 July the President committed to attempting a salvage and that they were told a lift could take eight to ten weeks. No government release confirms either, and the Prime Minister on the same day restated that the government is still weighing salvage against a memorial. Both are recorded on finding 3; neither is preferred. The recovery phase itself has ended: diving has ceased and the French Guiana, Trinidad and Tobago and Brazilian teams have left the country, with about thirty people never recovered.

The first published answer to the self-incrimination question, and not from the government. A Kaieteur News column argues the Commission and the prosecution can properly run together, and claims the Commissions of Inquiry Act already contains provisions protecting witnesses from self-incrimination. If that is right it is the answer finding 23 has been recording as absent, and it has been in the statute all along. This archive holds no text of that Act and cannot check it — the same gap that leaves finding 26 unresolved.

And the opposition leader put the charging decision in the sharpest terms yet, calling the charges "ridiculous, vindictive, and intended to silence the workers" and demanding the state say which ministries owned the cargo aboard. Those are an opposition leader's assertions about a live prosecution and this archive adopts none of them. Three men are charged and untried, and nothing on this site bears on their guilt or innocence.

What the verification pass caught in this build

Sixteen defects, of which the first was the failure to propagate the DPI correction to finding 10 — the second time in two days a correction has been made on one page and left standing on another, and the strongest argument yet for a register of withdrawn claims that can be tested mechanically against every page. Also fixed: an assertion put in the opposition leader's mouth that his source does not support; "hours apart" invented for two same-day releases that carry no times; two events of 31 July filed under 1 August without noting the difference; a tool-condensed capture described as matching a release "almost word for word"; a claim that the manifests annotate every entry, when they cover 192 of 295; a new publisher string that would have split Kaieteur News in the outlet filter; and five stale counts.

Also

The counterfactual page's timeline heading now reads The whole road, start to finish under an AI Advisor, and the landing page links to it beside the chronology — labelled, because it is not the record.

1 August 2026 — sixteenth build, one new page

No documents were added in this build. (A later sweep the same day added ten and took the corpus to 295; see the entry above.)

The road not taken

A new page asks one question: if the Government had an AI advisor from the night the Barima went down, whose only job was keeping the public's trust, what would it have said to do? It runs day by day through the first fortnight and month by month to an ending in 2027, with a summary strip in the same visual form as the landing page's timeline.

It is invention, which is what this archive exists not to publish, so it is walled off rather than woven in. It sits on the utility row of the masthead with the apparatus, not on the record row. It carries a standing banner. The first fortnight is set against the real record and cited to nineteen documents; everything after 1 August is made up outright, and the strip marks that with a break line, hollow markers and a dashed spine so a reader who scrolls past without reading a caption can still see where the record stops. It never says what the inquiry would have found about why the ship sank, because nobody can know that. And it says nothing whatever about Kevin Price, Rondell Roberts or Delon Granderson, who are charged and not tried.

The page also states where it goes beyond what anyone actually proposed at the time. There are six such places and they are listed, the largest being the six-month claim deadline: nobody in Guyana raised it until a retired ship pilot did, unasked, on 29 July, and the page puts it on the 23rd. That is both the most useful thing on the page and the biggest liberty it takes.

What the verification pass caught, and one thing worth publishing on its own

The page was written three times — a two-column comparison that was accurate and unreadable, a twelve-thousand-word chronology still pitched at people who read policy for a living, and this one, under five thousand words in plain language. Each version was checked adversarially and each check found real damage. The pattern in the last round is worth recording because it is the specific hazard of writing simply: shortening a sentence quietly strengthens it. A hedge about "whatever hours the log shows" became an assertion that such hours existed. "Bought four days" became "lasted four days", which claims a finding was withdrawn that never was. An unanswered demand became one that "was refused", inventing an act of government. Millington, a retired ship pilot, was promoted to ship captain in five places, and Patterson, the opposition's shadow minister, became "a former minister of its own ministry" — which was the entire force of the sentence.

One error is worth more than the rest together, because this archive had already made it, corrected it, and then made it again. A draft claimed the state concealed the 397-licensed-against-250-life-jackets discrepancy for ten days until it was extracted by a retired ship pilot. That is false. Both figures were volunteered by the Minister on 19 July, are carried by five outlets that day, and one is in the state's own release; what Millington did on the 29th was set them against each other. Finding 24 on the analysis page opens with a correction saying exactly that, published on 30 July after an earlier version made the same mistake — and the new page reinstated the withdrawn claim, in stronger form, a day later. That is the second time a correction has failed to travel between pages here, and it is the clearest argument yet for a register of withdrawn claims that can be tested against every page mechanically rather than remembered.

The honest version is smaller and worse for the state: the figures were never hidden. They were published on day one and left unreconciled for a fortnight, beside a ministerial assurance that they met the required standard, and nobody in government has been asked to reconcile them yet.

Also in this build

Elapsed-time claims moved from thirteen days to fourteen — the fourth consecutive build chasing rotting day-counts by hand.

31 July 2026 (late) — fifteenth build, one hour of broadcast and a new kind of source

One document was added, and it is the first of its kind in this corpus: an hour-long podcast interview, captured from the auto-generated caption track of the published video. It takes the corpus to 285. It is worth a build of its own for two reasons — what the guest says, and what capturing it forced this archive to decide about fidelity.

The source, and what is wrong with it

Beyond the MV Barima ferry wreck, OGGN Oil Talk, streamed live 30 July on the Kaieteur Radio Guyana channel. Host Janette Bulkan, who already appears in this archive as the author of the 25 July letter setting out the unanswered rescue-timeline questions. Guest David Patterson, AFC shadow minister for public works and a former Minister of Public Works — the first person to speak at length in this corpus who has held the ministry these vessels sit under. He did not run the Transport and Harbours Department itself, and this site records elsewhere that MARAD and the T&HD answer to different Ministers, so the distinction matters.

The capture is imperfect and the file says so. Of 1:04:15, this archive holds 0:47 to 59:55 — 3,548 seconds of 3,855, or 92 per cent; the closing four minutes would not load. (The record file first said 93 per cent. That was a rounding done by eye rather than by arithmetic, and it is corrected.) More importantly it is a machine caption track, not a publisher's transcript. It mangles proper nouns throughout: the host's own name renders wrongly, so does the vessel's. The record therefore carries a new field, capture_fidelity: auto-caption, and every quotation drawn from it is marked on this site as substantially accurate rather than verbatim. A caption track is a usable source. It is not a transcript, and this archive does not treat it as one. The method page now describes the capture type in full.

The strongest thing in it goes to the document this site has been waiting on since day one

Finding 2 has recorded from the beginning that MARAD made a categorical "not overloaded" finding on 22 July and never produced the paperwork behind it — and it has carried, honestly, the caveat that this analysis cannot establish that certificates of this kind are normally published in Guyana, so the absence might mean nothing.

Patterson does not produce the certificate. What he does is narrow the question, and he does it as someone who held the portfolio: he treats annual certification as routine and asks "how can you say you don't have the registration certificate of a document… but you are certifying it every year." He likens it to annual vehicle fitness certification, says speed boats and cargo boats renew at MARAD each year, and concedes a survey "may not check every single item." If he is right, the question stops being whether such a document should exist and becomes where this vessel's is. Finding 2 is updated, and the document that would settle it now includes MARAD's register of annual certifications. One former minister's recollection, uncorroborated, from a caption track — recorded at that weight and no more.

A statutory collision this archive cannot resolve, and should not pretend to

Patterson cites Cap. 49:07 section 18 as having given inspectors of shipping sweeping powers before 18 July: to board, inspect machinery, examine certificates, prevent overloaded vessels from sailing, and detain unfit ones.

But this archive already holds a reading of the same Act. Millington's submission reads section 3 as disapplying Part I to any ship belonging to or in the service of the State — and the Barima belonged to the State. If section 18 sits inside Part I, then the powers Patterson says went unused were not available at all. If it sits outside, they were. This archive cannot tell which, because it does not hold the text of the Act. That gap has been listed on the method page as "no statutory text" since 30 July; it now has a concrete consequence attached to it, which is a better argument for closing it than any abstract one. Finding 26 records both readings and prefers neither. The resolution costs one PDF from the Official Gazette.

The self-incrimination problem may be larger than three men

Finding 23 recorded that the three accused may be unable to give evidence to the Commission, and that four remedies for it had been drafted and none chosen. Patterson argues the effect runs through the operator's whole management chain: the marine superintendent "is not going to incriminate himself", nor the chief engineer, "nor would the general manager of transport and harbours" — leaving the Commission with "circumstantial" evidence and "he say, she say." If that is right it damages the one remedy that looked robust: severance, reconstructing the casualty from documents and institutional evidence, assumes the institutional evidence remains obtainable. He adds a second prediction — that masters may simply refuse to sail. Both are predictions, neither has happened, and the Commission has not sat. Finding 23 is extended and says so.

Two figures, recorded and not adopted

Bulkan states the missing as "probably between 30 and 50. We still do not know how many persons are unaccounted for." Almost every other source here gives about 30, and 30 is the figure that makes the arithmetic close against 179 aboard — the exception being News Americas on 19 July, which gave 49 while the count was still moving. Fifty does not close against 179 at all, and this archive holds no source for a higher complement. It is recorded on the figures page as a divergence and not adopted; the figure of record remains 30. What it marks is that a well-informed critic does not regard the residual as settled.

Patterson separately offers the only explanation anyone has given for any of the count movements: that survivors picked up by one agency were not counted by another, so the published number jumped. It is uncorroborated, dateless, and does not touch the 133-against-179 gap. Recorded because nothing else in thirteen days has attempted the question.

A possible explanation for one of this archive's own standing gaps

The method page has recorded, since early on, that no Auditor General finding specific to the T&HD could be located despite targeted searching. Patterson says the department has not turned a profit since "2007 or 2009" and has not been officially audited since "most likely I think 2005, 2006." If that is right, the reason the search found nothing is that there is nothing — which would convert a gap in this archive into a finding about the operator. It is one man's recollection, hedged twice in his own words. The method page now records it as a lead and keeps the Auditor General's reports on the retrieval list.

Also recorded

Bulkan's arithmetic of the rescue gap, from the Prime Minister's own account — the first rescue "took place at 27 minutes past 2 on Sunday, that's more than 3 hours later" — which is the sharpest published statement of the rescue-timeline question in this corpus and is not a new fact but the existing facts subtracted from one another. Finding 15 carries it, with the note that it is her arithmetic and not a state statement of its own response times. Patterson's account of the pre-departure travel plan and the "lighthouse" as maritime nerve centre goes there too, uncorroborated. His relay of the captain's account of missing equipment is not adopted: it is the untested assertion of a man awaiting trial, at second hand. The positions page now carries his account at length, and a new entry for Bulkan as interviewer, including her relay of MP Amanza Walton-Desir's argument that a marine safety investigation under SOLAS cannot be substituted by a Commission of Inquiry. The developments strip gains a twenty-second turn.

What the verification pass caught, and one correction it forced elsewhere

Two adversarial passes were run over this build before it was published, both instructed to refute rather than confirm. Between them they found twenty-one defects in material written in the last hour, and the count is worth stating plainly because it is the argument for running them at all.

The instructive ones. Three headings dated the broadcast to 31 July — the day this archive retrieved it — when it was streamed on 30 July, and each heading was contradicted by its own next clause. Retrieval date is not utterance date and this site knows the difference; it slipped three times in one sitting. A blanket "none of it is corroborated by any document here" was written twice, and was false both times: Capt. Adams's 2015 letter touches the certification point and the captain's own court statements of 29 July carry the equipment allegations Patterson relays. A headline said Patterson "says the certificate exists." He does not; he asks a rhetorical question implying one should. The changelog paragraph above said so correctly while two other pages overstated it. The coverage figure was 93 per cent by eye and is 92 by arithmetic. And a paragraph explaining a count movement described a movement that had not happened.

One of the passes also found a defect that predates this build and is worth its own line, because it is the second time this archive has published an absence it had not checked. The method page has said since 26 July that the residual — the people neither recovered nor accounted for — "has never appeared in a government document." That is wrong. DPI's own release of 30 July gives "30 could not be accounted for," and the Prime Minister gave 30 unaccounted for on 24 July. The true statement is narrower and still damning: no government document has ever added the recovered to the unaccounted for and said what the sum means. The line is corrected on the method page. The rule this archive published on 30 July after the first such error applies again without amendment: an absence claim is only as good as the search behind it.

31 July 2026 — fourteenth build, day thirteen

Eighteen documents added, all dated 31 July, taking the corpus to 284. Nothing on the site's standing gaps closed. The day's movement is almost entirely in what people are now doing about the gaps rather than in the gaps themselves.

The defence moves to stop the two proceedings running together

This archive published finding 23 on 30 July: that the three people who know most about the voyage were charged two days before the inquiry was sworn, that this raises a self-incrimination problem nobody in government had addressed, and that a document setting out four ways to manage it had been sitting on this website since 29 July.

On 31 July it stopped being theoretical. Attorney Dexter Todd, who acts for all three accused and is also an APNU Member of Parliament, announced he is preparing a court challenge to the Director of Public Prosecutions' position that the prosecution and the Commission may proceed simultaneously. His framing of the clash: "What will stand down? Will the COI stand down or will the criminal court stand down?" Single source He made his clients' participation in the Commission conditional on three things — whether the criminal proceedings are stayed, whether adequate safeguards exist, and whether the terms of reference are released for review.

Finding 23 is updated accordingly. The point it now makes is narrow and, this archive thinks, fair: the state's failure to choose a mechanism in advance has produced exactly the outcome those mechanisms existed to prevent — the inquiry's two most important witnesses negotiating in public, through counsel, the terms on which they will speak to it. Whether Todd's legal argument is sound is for a Guyanese court, and this archive takes no view.

Seaworthiness is put at the regulator's door, and the regulator has now been silent for a week

Todd: "None of the parties charged are to determine seaworthiness of that vessel. Seaworthiness of the vessel carries a particular procedure, and that has to be determined by, of course, MARAD." Finding 26 recorded MARAD's silence since 24 July; it is now seven days, and the central question in the prosecution has been addressed publicly to a body that has not spoken. Everything relayed on the accused's behalf about the vessel's equipment remains untested allegation by a man awaiting trial, and this archive adopts none of it.

The toll divergence did not resolve. It went quiet on one side

Every Guyanese outlet that gave a number on 31 July used 73. The two that had carried 72 on 30 July — the Department of Public Information and Demerara Waves — published nothing at all, so neither corrected nor defended it. Guyana Times, the outlet closest to the administration, now carries 73 while the government's own information service still carries 72, and nobody has acknowledged the gap. Meanwhile the regional wires have started giving "more than 70" or no figure at all, and two letter-writers have gone the other way and treated the combined dead-and-missing total as a confirmed death count of 100. Finding 18 records all of it. No new bodies were reported recovered.

The wreck, and a second document put in front of bereaved families

Single source Kiskadee Watch reports that the government "circulated documents requesting relatives consent to designating the vessel a memorial site." If accurate that moves the memorial from a proposal discussed with families to paperwork placed before them, and it is the second such unpublished document this archive has had to record. The Opposition Leader says he has had no reply of any kind to his salvage offer — "No update, nothing at all" — and is escalating to a formal written request. Against the memorial, the Assakata Village Council in Region One formally resolved that the vessel be raised so a complete investigation can be made: the first recorded resolution of an Indigenous village council on the question, from the region that lost the most people. Finding 3 is extended.

Money: a framework team, and the word "compensation" avoided again

A team has been assembled to draft the national framework for the HEART initiative, and the assistance is again described as not compensation. No members named, no sum, no timetable. Thirteen days on there is still no published rate, no eligibility rule and no total disbursed. Finding 22 is updated; the limitation warning on it stands unchanged.

Also on the record

New detail on why the Port Kaituma wharf failed — an opposition MP names the designer and attributes the failure to pre-construction design and site investigation, describing piles that never reached firm ground, a former sawmill site with buried debris, undersized steel and tie rods that "snapped under the pressure." Single source, and an MP's account rather than an engineering report, but it bears on the chain the Prime Minister has already conceded. The resignation campaign reached the minister's gate, with a joint opposition vigil outside Juan Edghill's home. Kaieteur News records that President Ali has not faced the local media since returning after the disaster. And Single source Kiskadee Watch reports putting conflict-of-interest questions about two commissioners to three government bodies and receiving no response from any of them.

Day counts advanced, and a note on why that keeps needing doing

Every elapsed-time claim moved from twelve days to thirteen. Claims anchored to an event — the Commission was sworn in twelve days after the sinking — were deliberately left alone, because those are facts about a date rather than about now. This is the third build in a row in which rotting day-counts have had to be chased by hand, and it is the clearest argument in this project for building a register of time-relative claims that can be re-tested mechanically rather than remembered.

What could not be retrieved

Recorded because a reader should know the sweep's reach. Demerara Waves, News Room Guyana, News Source Guyana, DPI, HGPTV and Guyana Graphic published nothing dated 31 July that could be retrieved; several of those sites terminate at 30 July. The Guyana Chronicle's 31 July archive returns a JavaScript stub with no content, so this archive cannot say whether the state paper published that day. One aggregated item was dropped because only a headline could be reached. And a trap worth naming: three iNews Guyana articles appeared in an early listing as 31 July and are in fact dated 30 July on their own article pages — a snippet-level sweep would have filed all three under the wrong day.

One quotation was dropped in verification. A sentence in the Kaieteur report of Todd's challenge rendered once with a negation and once without it across two separate fetches, so it is not quoted anywhere on this site; only the wording that rendered identically on both passes is used.

30 July 2026 (late) — thirteenth build, the analysis doubles and the story gets a spine

Two things. The analysis page was expanded from nineteen findings to twenty-eight, after a verification pass that killed one candidate outright and reshaped three others. And the site now carries a short account of the story itself, for a reader who has just arrived and does not know it.

Notable developments: the story in twenty turns

Until now a reader landing here had two options — the chronology, which records every dated entry the corpus supports and runs to several thousand words, or nothing. That is a poor choice to offer someone who has heard about a ferry and wants to know what happened.

There is now a notable developments rail on the landing page and at the head of the chronology: twenty entries from the yard that built her in 1939 to the swearing-in of the Commission on 30 July. The test for inclusion is narrow — each has to be a moment after which something was different. The eight pre-disaster turns are collapsed behind a disclosure on the landing page, because a reader who has just arrived wants the fortnight rather than the ninety years, and open by default on the chronology. Markers are coloured by phase, and the colours carry no information the text does not, so the list reads correctly in monochrome and to a screen reader.

It is deliberately not a summary of the analysis. It records what happened, in the order it happened, and leaves the argument to the page that makes it.

Nine new findings, and one that did not survive checking

Every candidate was put to an adversarial pass before it was written: verify the factual base, then try to destroy it, then build the strongest innocent explanation available, then say which the record actually supports. Two of them came back materially different from how they went in, and one came back dead.

What was added.

Finding 20 — a law officer of the state publicly attributed alcohol to a named accused man. Every official account of the toxicology until 29 July says cannabis: the minister on 19 July, the Vice President on 22 July, DPI's own release the same day. On 29 July the Attorney General said "accounts indicate the captain was under the influence of alcohol and narcotics." No positive alcohol result appears in any test, statement, report or document in this corpus. An alcohol panel was run — that is on the record — and its result has never been published either way, which is the point. The finding is written strictly as a fact about what a minister said while the matter was before a magistrate, and it says so.

Finding 21 — the carrier decides whether its own cargo is dangerous, and on this voyage the carrier's parent ministry was the shipper. The Ministry of Public Works published its own consigned cargo on 29 July: one ATV, six drums of gasoline, four drums of diesel. Section 17(3) of the Transport and Harbours Act provides that "Any question as to whether goods are dangerous goods shall be determined by the Department" — the carrier — with no external standard, no review, and no requirement that the determination be recorded. As first published this finding said something else and something false: that nothing in this record mentions dangerous goods at all. It was withdrawn and rewritten within hours; the reason is set out further down this entry, and it is the most important thing in it.

Finding 22 — money is reaching bereaved families at figures nobody can explain, and a claim window may be closing behind them. Ranked critical for one reason: unlike everything else on the page it is prospective. It is also the one finding on this site with a practical instruction attached, which is repeated here: anyone who may have a claim should take Guyanese legal advice now rather than wait for the Commission.

Finding 23 — the three people who know most about the voyage were charged two days before the inquiry was sworn, and nobody in government has addressed what that does to the evidence. Written as four documented facts and one unanswered question, and expressly not as the inference in its own headline.

Finding 24 — licensed for 397 people, 250 life jackets aboard. Both numbers came from the same minister at the same press conference on 19 July, one of them in the state's own written release, and he described it as meeting the required standard. [WITHDRAWN THE SAME DAY — see below.] As first published this finding said that in twelve days nobody in this corpus had put them side by side. The submitted analysis does exactly that, and reports an opposition figure making the same point. Rewritten within hours; the claim is now that the government has never reconciled them.

Finding 25 — the state has had the loading on video since 20 July. Officials sat up overnight watching it and published what it says about people: 179 boarded. They have never said what it says about cargo — the thing three men were charged over eight days later. In between, the regulator cleared the vessel of overloading on the strength of, in his own words, "persons that we talk to."

Finding 26 — MARAD has said nothing since 24 July, and the master's dated allegation about its own inspection is sitting unanswered. Carries a genuine second gap: the state's first account of how the casualty was located rests on an electronic locator transmitter signal, and no document anywhere says what device produced it.

Finding 27 — nine weeks before charging three men over how a ferry was loaded, the same ministry advertised roughly a thousand people standing on a vehicle deck at Fort Island. Recorded as a pattern in how one body applied the same question to itself and to others, and expressly not as a finding that any rule was broken.

Finding 28 — on one reading of the Shipping Act, the official who cleared the vessel was, in law, her owner. Single-sourced to the submitted analysis, and marked as such.

Finding 3 was extended rather than duplicated. The disposal of the wreck — the one physical object in a live 72-count murder prosecution — has been placed in the hands of bereaved families, and no document in this corpus records the Director of Public Prosecutions, the police, the court or the Commission being consulted, asked or heard on it. Against that, the families most closely consulted appear to want it raised, which would mean the mechanism disposes of nothing.

And one candidate was killed. A finding was drafted on the Prime Minister publicly doubting the intent the prosecution must prove — that he "did not believe they set out to kill occupants." Verification found the quotation exists only as an undated back-reference, twice, in a single outlet, with no record anywhere of the occasion on which he said it — and that this is the same outlet whose framing sentence about the Director of Public Prosecutions this archive already had to withdraw on this exact story. It was dropped. Half of it survived and was folded into finding 4: the Director of Public Prosecutions has stated a basis for 72 murder counts without addressing intent or malice, and three outlets independently record the omission.

What the checking removed from the findings that survived

Two elements were cut from finding 21 as wrong. MARAD is not under the same minister as the Transport & Harbours Department — the corpus is clear on that, including a self-correction by the author of the submitted analysis, so the conflict recorded is between owner and shipper and not between operator and regulator. And 29 July was not the first the public heard of state cargo aboard: the Prime Minister volunteered the substance of it on 22 July. Both corrections make the finding weaker and both are published.

From finding 20, the drift over whether the second man who tested positive was the first engineer or the chief mate was cut, because it is not the Attorney General's drift — it begins in DPI's own release of 22 July and is already recorded at finding 7.

From finding 23, the causal claim was cut entirely. One defence lawyer's sub judice proposition is probably bad law and comes from counsel of record for all three accused; the other's suggestion about intention is expressly disclaimed by the outlet that carried it; and thirteen of the seventeen crew died, so the inquiry's evidentiary loss is overwhelmingly attributable to the sinking rather than to the charge sheet. What remains is the unanswered question.

The plain-language summary grew to twelve points

Three were added, all of them from the new findings and all chosen because they can be understood without knowing anything about ships: the life jackets against the licensed capacity, the video the state has never described, and the money going out against a deadline that may be running. The landing page previews all twelve.

One more correction found while doing this

DPI's release of 19 July states the vessel "departed Parika at approximately 3:15 p.m." Every other account, including the state's own later ones, has her leaving Georgetown. It is almost certainly a drafting error written in the first hours, and it is recorded at finding 17 because it sits in the state's first primary document and because the port of departure is not trivial here — the regulator's load-line reasoning turns on the density of the water she was floating in.

Then the verification pass found that two of the new findings were wrong, and why

An adversarial check was run over the rebuilt site before it went out. It killed the central claim of two of the nine findings just written, and the reason is worth publishing in full because it is a fault in method rather than a slip of the pen.

This archive searched its own markdown extracts and never searched the 253-page PDF it publishes on its own site. The corpus is 266 records; 265 of them are extracts, and one is a primary document held in full. The search that produced finding 21's claim — that nothing in this record mentions dangerous goods — ran over the extracts. The PDF contains eleven mentions of dangerous goods, six of the IMDG Code, and an entire numbered deficiency on the subject. The same search produced finding 24's claim that nobody had set 250 life jackets against a licensed capacity of 397. The same document does exactly that, on page 43, and reports an opposition figure making the same point publicly.

Both claims are withdrawn, and both findings have been rewritten around what the record actually contains. In each case the true version is stronger than the false one.

Finding 21 is now about section 17(3) of the Transport and Harbours Act, which provides that "Any question as to whether goods are dangerous goods shall be determined by the Department" — the carrier. No external standard, no review, no requirement that the determination even be recorded, and a carrier with a commercial interest in the answer. On 29 July the parent ministry of that carrier disclosed that it had itself consigned six drums of gasoline and four of diesel on the fatal voyage; the submitted analysis notes, of this casualty, that fuel in containers is capable of falling within IMDG Class 3. Nobody has connected the statutory point to the cargo list, and nobody has asked who answered the section 17(3) question, or whether anyone did.

Finding 24 is now that the government has never reconciled the two figures — not that nobody has. One person has: an unsolicited analyst writing to this archive ten days later, who adds the part that is still open, which is that the aggregate capacity of the vessel's eight liferafts is not established in the public record at all.

Both rewrites open by stating the error, because a correction buried at the foot of a page is not a correction.

Six smaller defects the same pass found

Finding 27's headline arithmetic was wrong. Fort Island was 25 May and the charges were 28 July: nine weeks, not seven. The Konawaruk 1899 was commissioned five months before Fort Island, not seven — the "seven months" came from a source describing the gap to the sinking, and was re-anchored to the wrong event when it was lifted.

Finding 25 and plain-language point 11 were a day out. The wharf footage was seized on 20 July and the charges were laid on 28 July: eight days, not nine. (An earlier version of this entry said "findings 11 and 25" — finding 11 is the medico-legal one and carries no such interval. Corrected.)

The developments rail stated a spend that may never have happened. It read "A further G$124.5m goes into her hull." The corpus holds only an engineer's estimate in bidding documents, and the Opposition Leader alleges the contract was never awarded or executed — which is why this archive's own finding 14 downgraded that tender back in an earlier build. The rail now says a repair was put out to tender and that whether it was ever done is not established.

The rail stated three contested things flatly. The build year, whether the Lady Northcote was withdrawn, and why the Ma Lisha never took the route are all marked contested on the chronology, and the rail — whose own preamble points readers to that chronology "with every conflict marked" — had dropped the marks. Restored.

The rail named three untried accused with no caveat near it. On the landing page the standing caveat follows a few hundred words later; at the top of the chronology it was thirty thousand characters downstream. The caveat is now inline in the entry itself.

Finding 20's heading called the Attorney General "a minister" and rendered his words as "drunk." He is a law officer, not a minister, and the record says "under the influence of alcohol and narcotics" — a harder word than the source supports, in a heading about a man who has not been tried. Both fixed.

And the claim that was killed was still live on the chronology

The finding drafted on the Prime Minister doubting the murder charge's intent element was dropped from the analysis page. The chronology still carried it, dated, under its original heading — a claim withdrawn on one page and asserted on another, which is the exact failure this site audited itself for earlier the same day. The chronology entry has been rewritten to record what it actually is: an undated back-reference, twice, in one outlet, with no occasion recorded anywhere, from the same outlet whose framing sentence about the Director of Public Prosecutions this archive already had to withdraw on this same story.

What the same pass confirmed

A second adversarial pass, run after these corrections, found four more — all of them the same failure, and all now fixed: the withdrawn life-jacket claim was still live as a plain-language point and on the landing page; the standing post-mortem claim is falsified by the same PDF, which lists "post-mortem findings" among the material a Commission should place under chain of custody; finding 23's claim that only a broadcaster had named the self-incrimination problem was falsified by the same document, which calls it "the single most urgent practical problem facing the Commission" and drafts four remedies; and finding 3 undercounted who has argued the wreck is evidence. The lesson is the same one and it is recorded rather than absorbed: an absence claim is only as good as the search behind it, and this archive ran three searches before it searched everything it holds. Beyond those, every factual assertion in findings 20 to 28 and in all twenty rail entries was checked against the corpus and holds, with quotations located: the four toxicology attributions and the absence of any positive alcohol result; the cargo list and the Prime Minister's earlier disclosure; the $300,000 and $350,000 figures and the limitation arithmetic; the 397 and the 250 and MARAD's 394; the seizure of the footage and the "persons that we talk to" quotation; MARAD's silence after 24 July, tested by sweeping every 26–30 July record for any MARAD utterance; the Fort Island figures and the "No vehicles will be permitted aboard the vessels" instruction; and the section 3(3) quotation, correctly marked single-source. The re-search of the PDF also broke this archive's longest-standing absence claim, the one about post-mortems. It searched for "coroner" and "inquest", found both as statutory concepts, and failed to search for the word in the claim's own headline. A later pass found it: the same document lists "post-mortem findings" among the physical and forensic material a Commission should place under documented chain of custody. Finding 11 and plain-language point 7 are rewritten. The claim that survives is narrower and, if anything, worse — not that the words are absent, but that no cause of death has ever been stated for a single one of these dead, and that the only document treating post-mortems as existing at all is an outside analyst telling a Commission what to secure.

Counts

Twenty-eight findings, twelve plain-language points, four self-corrections plus two more declared inside findings 21 and 24, and a citation block on every one of them. The corpus is unchanged at 266 documents: nothing was added to support these findings. Every one of them was already in the record and had not been read — including, in two cases, in a document this site has been publishing in full for a day and had never searched.

30 July 2026 (evening) — twelfth build, the Commission is sworn in and the whole site is audited

Two things happened here. The day's reporting was swept and filed, and every page was read against the corpus looking for statements that had quietly gone stale. Both are recorded below, the corrections first, because a correction matters more than an addition.

The day's news: the Commission of Inquiry was sworn in, with the press shut out

Twelve days after the sinking, on Thursday 30 July, the five commissioners took the oath of office at State House, Georgetown, before Chief Magistrate Judy Latchman, with President Ali present. Four were sworn in person; the chairman, Justice Godfrey Phillip Smith, is not yet in Guyana and was sworn virtually. Five sources in this corpus carry it, so it is not a single-source fact: DPI, Demerara Waves, News Room, News Source Guyana and Kiskadee Watch. iNews Guyana published a sixth account the same day; the sweep could not retrieve its text, so it is not held here and not counted.

The media were not invited. Demerara Waves and News Source Guyana both state so; News Source adds that the ceremony was streamed on the President's own social media accounts instead. That is now finding 19 on the analysis page, and it is on the landing page, because the independence of this Commission is the most contested question in this record and the one event at which that independence was on public display was held on a channel the state controls end to end.

The statutory basis is now official. DPI's release states, in the government's own words for the first time, that the Commission was established under section 2(1) of the Commissions of Inquiry Act, Chapter 19:03. This archive had recorded that citation only as Kiskadee Watch's reporting of 27 July and as the AFC's proposal. It is now confirmed, and the consequence is written into the positions page and the chronology: Christopher Ram's argument that Part XVIII of the Shipping Act should have been used survives as a criticism of the choice, and no longer stands as an open question about which statute was in fact used.

Everything downstream of the statute is still missing, and the site now says so in one list rather than four scattered sentences: no gazetted instrument or Statutory Instrument number, no terms of reference published as a document rather than paraphrased, no named secretary — Demerara Waves states that flatly — no reporting deadline, no budget figure or appropriation, no rules of procedure, and no route for a member of the public to make a submission. Kiskadee reports the President saying the commissioners will determine their own procedures, which is a statement that the rules did not exist on the day they were sworn.

The government's own release lowered the death toll from 73 to 72, and nobody announced it

From 24 to 29 July every official figure in this corpus gave 73 bodies recovered. On 30 July DPI's release and Demerara Waves gave 72. On the same day News Room, Kaieteur News in three separate reports, Kiskadee Watch and iNews Guyana still gave 73. No source announces a revision and none explains one.

This archive is keeping 73 and recording 72 as an unexplained divergence rather than as a correction, for a stated reason: 73 + 30 unaccounted + 76 survivors makes the 179 the state says were aboard, and 72 does not — yet every source giving 72 also gives 30 unaccounted. The sum stopped reconciling. The likeliest mechanism is contamination from the 72 counts on the charge sheet, and that is this archive's inference, marked as such. It is now finding 18, and it is the ninth point on the plain-language summary.

Twenty-five documents added, taking the corpus to 266

Twenty-one are original Guyanese reporting on the sinking, which moves the independent-reporting count for the first time since 29 July — from 107 to 128. The other four are one DPI release and three opinion pieces. Among the material new to the record: the charge particulars, published for the first time by Kaieteur News, which name the position 9.3 nautical miles on a bearing of 040° off Devonshire Castle and confirm that the 72 counts correspond to 72 named victims; the Ministry of Public Works' itemised list of its own cargo aboard — one Honda side-by-side ATV, six drums of gasoline, four drums of diesel — which is a statement about its own consignment and not a manifest; the Attorney General personally defending the charges and characterising the accused, recorded as allegation and as a fact about what he said; a belief attributed to the Prime Minister by one outlet, that he did not think the accused set out to kill anyone — later qualified, on 30 July, when verification could locate no occasion on which he said it (see the thirteenth-build entry above); the first published figure for a funeral payment actually received, $350,000, from a named relative; and two assessments of the state's care of the bereaved published the same day that do not agree — PAHO's praise, offered incidentally at an unrelated signing, against a clinical psychotherapist's account of an improvised response run out of an open public venue.

A gap this archive had published as open is now closed, and the note is withdrawn

The chronology recorded that the Amerindian Peoples Association's picket of the Office of the President on 29 July appeared in no press report in the corpus, and that only two social-media posts held it. That is no longer true. Kaieteur News, News Room, Kiskadee Watch and Demerara Waves all carry it. The note is withdrawn and the entry rewritten from the press reports, which supply crowd sizes, placard texts, named speakers and a third minister's name in the resignation demands.

Corrections found by auditing the site against its own corpus

Seven statements were still published that the archive had already established were wrong or had been overtaken. Each is a failure to propagate a correction, not a new error, which is its own kind of problem: a correction logged on one page and left standing on another is not a correction.

A withdrawn claim of absence was still on the figures page. It read: "No BBC byline surfaced. Reuters did contribute…so treat the wire coverage as present but not directly captured." Both are in the corpus and have been since 27 July — BBC News by Vanessa Buschschlüter and Tom Bennett, 19 July, and Reuters by Kemol King, 20 July. The method page had recorded that correction on 27 July; the figures page had not been updated to match. Removed.

The same withdrawn post-mortem wording was on the landing page. The analysis page had already narrowed it, on 29 July, to the Barima dead specifically, because the words do appear in the corpus in two columns about an unrelated drowning. The landing page still carried the unnarrowed version — "No post-mortem is mentioned anywhere in the corpus. Not once." Replaced with the narrowed wording.

The landing page stated the overloading inference more flatly than the finding it summarised. It said the vessel "was about 70 tonnes over"; finding 1 says the comparison is inference and that whether she was overloaded is not settled in either direction. The summary line now says which part is documented — the category error, uncorrected — and which part is this archive's inference.

The method page still described the Commission as it stood on 26 July: "No commissioners named, no chair." Five commissioners and a chair were named on 26 July and sworn on 30 July. Rewritten to state precisely what has closed and what has not.

PAHO was listed as absent from the corpus on both the figures page and the method page. It is not, as of 30 July. Both entries now record its statement and what it is worth — incidental, public-health, not the product of any assessment mission.

A cross-reference pointed at the wrong question. The standing note on the questions page said "Question five also draws on international instruments." It is question two. Corrected.

An investigation was attributed to the wrong outlet, twice. The 27 July investigation into whether delay on the Port Kaituma stelling kept the Barima on the route is Kiskadee Watch's, not Kaieteur News's. Corrected on the questions page and on this one.

Staleness that was not error, but had started to mislead

Eight day-counts had rotted. "Eleven days", "ten days" and "the eight days" were accurate when written and are not now. All are twelve days, and where a count refers to when a source published rather than to now — Andrew Donald's account, published eleven days after the sinking — it is left alone, because that one is a fact about the document.

The chronology's fourth part was not in chronological order. An earlier build appended two blocks of entries newest-first, so 29 July material sat above 28, 27 and 26 July in a page whose entire purpose is order. Part IV now runs 19 July to 30 July, ascending. No entry was altered by the reordering.

The complement chart contradicted its own prose. It labelled the first bar "Rescued 76" while the page around it said 77 were taken from the water and 76 survived. The bar is now "Survived", and the chart carries the 72/73 divergence in its description.

The inquiry chart stopped at day ten. It now runs to day twelve and records the swearing-in, the statute named for the first time, and the press exclusion.

The corpus enumeration did not sum to the corpus. The provenance table on the method page totalled 241 while the archive held 266. Every row was recounted from the files rather than adjusted to fit, and the independent-reporting figure was recomputed the same way.

One publisher had drifted back into two names. HGPTV was being recorded as both "HGPTV" and "HGPTV Nightly News" across sixteen files — the same regression the 27 July dedup was supposed to have ended. One name now.

What was checked and found correct

This is worth stating, because an audit that only reports what it fixed tells you nothing about its own reach. Every {{SOURCES:}} citation on the analysis and questions pages resolves to a document in the corpus — the build fails rather than emits a dead citation, so this is enforced rather than asserted. The per-outlet counts on the method page were recomputed from the files. Every standing claim of absence was re-tested against the corpus. An adversarial pass over the built site then found four that a first sweep had missed, all of them corrections that had been made on one page and left standing on another: the note that no newspaper reported the 29 July picket (withdrawn from the chronology, the positions page, the method page, the document browser and the source record it sat on), the claim that DPI published nothing after 25 July (it published on 30 July), the claim that no response to the announced membership of the Commission appears in the corpus (TIGI, the APA and the parliamentary opposition all responded on 30 July), and the claim that the particulars of the 72 counts appear nowhere (Kaieteur News published them on 30 July). Those are fixed above. The remainder still hold except PAHO's: no certificate, no gazetted instrument, no Auditor General finding, no trade union statement, no IMO statement, no MARAD advisory, no official statement of the number of counts, no statutory text. The 3 August Charity Magistrate's Court date is unchanged, and there is no bail application, no amended charge and no additional accused in any document dated 29 or 30 July. And the figures the archive carries across pages — 73 recovered, 68 identified and five not, 76 survivors from 77 taken from the water, 179 aboard against 133 manifested, the 23:01 distress call against Millington's 22:43 — are consistent everywhere they appear.

What could not be verified, and is therefore absent

A Guyana Chronicle report of 30 July, "Charged MV Barima suspects receiving psychosocial support – PM", could not be retrieved: the site returns nothing behind a JavaScript interstitial, and a syndicated mirror yielded only the lead paragraph. Its claim — that the three accused "remain registered as survivors of the maritime disaster despite the criminal proceedings against them" — appears nowhere else found and is new. It is recorded here as unretrieved rather than characterised from a fragment. No international wire copy dated 29 or 30 July could be fetched; the AP dispatch traced is dated 28 July and already held. Six Kaieteur opinion items of 30 July were not fetched and nothing is claimed about them.

30 July 2026 (morning) — eleventh build, a submitted analysis and the first public question

A master mariner submitted 253 pages, and it changes the analysis

Robert C. Millington sent this archive a statutory analysis of Guyana's maritime law and asked that it be incorporated. It now is, in full, and it is the first document here that is neither journalism, nor a state release, nor a social media post. A new tier — expert submission — is defined on the method page, labelled wherever it appears, and admitted on four conditions: a named author, stated qualifications, reasoning and citations supplied so a reader can check it, and publication in full so the checking is possible.

He is a retired United States Navy naval officer, a former licensed US Coast Guard Merchant Marine Deck Officer and First Class Ship Pilot, both Unlimited Tonnage, and has thirty-nine years in the maritime sector. He states he is not admitted to practise law anywhere, that nothing in the document is legal advice, that it makes no finding of fact, and that every person charged or named is presumed innocent.

It answers three questions this archive had recorded as open, and it answers them from a direction the archive could not reach.

Why nobody could produce a capacity figure. The Barima was 40.26 metres. The only Guyanese instrument containing stability criteria, stability information for operators, a record of passengers, and muster and drill requirements applies to vessels under 24 metres. The survey instrument covering vessels above 24 metres excludes passenger ships in five words. She fell outside both, and Part XV of the Shipping Act — headed "Passenger Ships" — is two sections, one an unused power to make regulations and the other creating offences of drunkenness and fare evasion. This archive had recorded three official capacity figures as officials contradicting each other. On this reading there may have been no properly constituted figure to contradict.

Why the manifest gap breached nothing. SOLAS requires all persons on a passenger ship to be counted before departure. The Guyanese provision that would do it is an enabling power under which no regulations were ever made. So the gap between 133 manifested and 179 aboard is not a breach of any counting duty, because no counting duty exists. The figures page has been corrected where it implied otherwise.

Why the load line was cited. He explains what a load line measures and what it does not — not metacentric height, not the righting-lever curve, not free-surface effect, not the downflooding angle, not passenger crowding — and then locates the fault in the statute rather than in the regulator: "A regulator with a load line and nothing else will use the load line."

And why the Commission sits under a general statute. Both maritime casualty regimes were unusable. The 1998 one gives the Minister a power and not a duty, with no threshold, no trigger, no time limit and no duty to publish. The 1883 one is disapplied by its own section 3 to any ship belonging to or in the service of the State, and the Barima belonged to the State. The general Commissions of Inquiry Act was the only route left — which he says was the correct choice and is also the clearest possible proof of the deficiency.

Two findings are harder than anything this archive had. Cap. 49:01 s. 429(3) allows a Board of investigation to cancel or suspend a certificate or censure its holder, and nothing else; no finding is available against a company, a Department, a Ministry, the regulator or any uncertificated person, because the Act creates no offence a Department can commit. And the Superintendent of Loading — the office the Minister publicly identified as responsible for inspection and loading — does not exist in law at all. It is not created by, named in, or given any duty by any of the four Acts. One of the three men charged is described as the Goods or Loading Superintendent. His observation, recorded as his and not adopted here, is that recourse to common-law murder follows from the Shipping Act's penalty structure — "a legislative failure, not a prosecutorial choice."

The penalties, on his reading, are inverted. Submerging a load line: $100,000. Sending a ship to sea so unseaworthy that life is likely to be endangered: $25,000 and six months, with a defence available that doing so was "reasonable and justifiable."

The 1891 river law is better than the 1998 sea law. Annual inspection, branded waterlines, the certificated maximum number of persons painted on both bows, one lifebelt per certificated person, a count entered in a book before departure, a prohibition on taking anyone aboard after the count, and — from 1918 — a self-executing rule of one passenger per five square feet of deck, excluding deck cargo, so the permitted number falls automatically as cargo is loaded. It applies on the Barima River and the Kaituma Creek. It does not apply on the sea passage between them and Georgetown.

The regulator said much of this seven years ago. He records that on 27 March 2019 MARAD's Director General, Claudette Rogers, stated publicly that Guyana had ratified the main conventions but the regulations still needed enacting; that the IMO Member State Audit of Guyana was completed in 2018; that Professor Duke Pollard, a retired Judge of the Caribbean Court of Justice, had been retained to update Guyana's maritime legislation and had completed the assignment with recommendations aimed at eliminating conflicts among agencies over responsibility; and that Guyana had not ratified the STCW 95 amendments while Guyanese surveyors issued certificates on that basis, which she called a significant breach of international maritime law. Four documents would largely answer the Commission's institutional question and all four exist: the audit report, Guyana's corrective action plan, the progress reports, and the Pollard report.

A deadline that may already be running

The most urgent thing in the submission is not analysis. He reads Cap. 49:04 s. 23 as requiring any action against the Transport and Harbours Department to be commenced within six months of the act or omission — on or about 18 January 2027 — on written notice naming the claimant's legal practitioner served a month before, so by about 18 December 2026. He gives six reasons it falls hardest on the people least able to avoid it: it runs from the act, not from knowledge or death or identification of a body; the Commission will probably not have reported inside six months, so a family that waits may find the period gone; claimants are disproportionately Region One residents with no road to Georgetown and the vessel that connected them is on the seabed; the notice must name a practitioner, so a family must find and instruct one first; five recovered victims were still unidentified and at least thirty unaccounted for, and a family with no body and no death certificate cannot readily obtain a grant of representation; and ex gratia assistance is not compensation and does not stop the clock. He adds that there is no insurer to claim against, Guyana having enacted the limitation of liability in Part XVI without enacting the liability.

This archive cannot confirm that construction of the section and does not give legal advice. It is on the figures page and the positions page because a bereaved family reading this site should learn that a deadline may exist. Anyone with a possible claim should take Guyanese legal advice now rather than wait for the Commission to report.

And it weakens part of this archive's own analysis, which is recorded rather than quietly absorbed

The analysis page argued that at every point where the state's account could have been checked against a document, the document did not appear. On this submission's reading, several of those documents were never required to exist. That is not a smaller problem, but it is a different one — not a state concealing what it knows, but a state with no instrument to know it. The analysis page now says so, in a section placed before its own findings rather than after them.

It does not touch the two findings that concern conduct rather than statute: the sequencing of the wreck decision, and the narrowing of disclosure between 19 and 25 July.

One conflict is recorded and not resolved. The submission gives the distress signal as 22:43, against the 23:01 this archive carries from the government's own information service and the BBC. He states the signal reached an aviation authority at 22:43 and took twenty minutes, through two intermediaries, to reach the maritime rescue authority. If that is right both times are correct and it explains why the foreign outlets named air traffic control while the Guyanese ones named the Timehri tower. The archive has not adopted it; 23:01 remains the figure used here, and the conflict is on the figures page and the chronology.

The first question from the public, and it found something

A member of the public asked why a ferry could serve Fort Island for the Independence flag-raising when the government says the newer vessels could not serve Port Kaituma for want of a wharf. It is now question 6, answered on the same terms as the archive's own five, and it is marked as a public submission.

The premise is wrong. Fort Island has a stelling — rehabilitated for G$82M under a contract awarded in August 2021, completed in seven months and commissioned in May 2022, with low and high ramps and a carriageway. The MV Malali resumed scheduled calls there in August 2022. So no ferry berthed where there was no wharf, and the answer says so first.

Two further distinctions are real and are given their full weight. The Fort Island operation carried no vehicles at all — the government's own release states "No vehicles will be permitted aboard the vessels," and attendees were shuttled by bus to Parika. And Port Kaituma's works failed geotechnically: an inadequate tie-rod specification, structural deficiencies on third-party review, fissures during pile-cap casting, revetment failure, and in Edghill's own words "significant movement of the soil".

And the question still lands, because it does not actually depend on Fort Island. The wharf explanation says why the Ma Lisha could not berth. It has never explained how the Barima had been berthing at Port Kaituma for four decades. Whatever facility she used on 18 July existed on 18 July. The real question is therefore not why the new vessel could not come, but why the only vessel using a route that demonstrably could be served was one built in 1939 — with a vehicle-capable ferry, the Konawaruk 1899, commissioned seven months before she sank and sent to Fort Island in May.

A search of Guyanese outlets found no published item connecting the two. Kiskadee Watch's own 27 July investigation into the Port Kaituma stelling delay does not mention Fort Island. Nobody is recorded asking the minister why an 87-year-old vessel could serve Port Kaituma but a 2023 vessel could not. The submitter identified a comparison the press has not made.

Seven documents were added to the corpus to support the answer, taking it to 240: the government's own transport release, two accounts of the ceremony, the 2022 stelling commissioning, the MV Malali's return, the Konawaruk 1899 commissioning, and a trade specification for the Ma Lisha giving its 3.0 m draught and its forward vehicle ramp.

The analysis page says what produced it

"What doesn't add up" is now labelled, unmissably and in four places, as a frontier AI analysis by Claude Opus 5 reasoning across the whole corpus, with no human investigator behind it. That belongs in the open: a reader weighing an inference should know what made it, and the point of naming it is that the inference can then be checked rather than trusted. Every one of the seventeen findings, all seven plain-language summary points and all four self-corrections now carry a citation block naming the documents behind them — twenty-nine in total.

The verification pass on that page found nineteen errors of its own and fixed them. Four were false claims of absence, including the assertion that no coroner or post-mortem is mentioned anywhere: those words do appear, in two Guyana Chronicle columns about an unrelated drowning, so the finding is now narrowed to the Barima dead specifically and states what the corpus does record. A claim that nobody asked whether the Kalliopi N could have sailed uncommissioned was false — Christopher Ram asked it on 28 July. A claim that every casualty figure is attributed to the Prime Minister was false. The statute count for the Commission was wrong. A sentence attributed to Minister Edghill was the newspaper's own summarising wording. The 72-count summary contradicted the finding it summarised. Several dates were a day out because the Friday 24 July briefing was reported on 25 July.

28–29 July 2026 — tenth build, the archive starts answering questions

This entry covers two days and two different kinds of work. The larger change is that the site now does something it had deliberately not done before: on one page it answers questions directly, in its own voice, instead of only reporting what other people said. On 29 July a further twenty-four documents were added and the archive corrected itself twice, both recorded further down. (An earlier version of this entry opened by saying no documents were added — that described the 28 July work only, and was left standing when the 29 July additions were folded into the same entry. Corrected on 30 July.)

A new page: Questions of the record

Until now every page here was a record page. Each statement carried a named publisher and a date, and where accounts conflicted the archive set them side by side without adjudicating. That restraint is what makes the record usable, and it is unchanged on those pages.

But restraint has a cost. A corpus of 209 documents contains findings that no single document states, and absences that are only visible once you know what a complete record would have contained. Leaving those unstated is not neutrality; it withholds the part of the work a reader cannot easily do themselves.

The questions page states them. Five questions, in plain language, answered by AI reading across every document the archive holds: who is accountable; how a disaster like this should be handled by international standards; what conditions the ferry was operating under and what could have been done differently; whether the boat was overloaded; and how many people actually died.

The page is open. Anyone can send a question in and it will be answered there on the same terms. A question the archive cannot answer will be published as such, because what the record cannot answer is itself a finding. That invitation still carries an honest caveat: no contact address has been published yet, so the route is not usable until one is. Publishing the invitation without the address would be worse than publishing neither.

How the answers are built, and the four rules they follow

Each answer opens with the short version as soon as you click the question. The reasoning sits behind a second disclosure, so a reader who wants the answer gets it in a paragraph and a reader who wants the working can have all of it.

Four rules govern the page, and they are published on it.

Where an answer and a record page conflict, the record page is authoritative. The analysis is downstream of the documents and can be wrong about them; the documents cannot be wrong about themselves.

Every answer names its evidence. Each one now ends with a list of the specific archive documents it rests on, linking to this site's page for each — publisher, date, and a link to the original. Where an answer infers, it says it is inferring. Where the record cannot settle a question, that is the answer given rather than the most plausible-sounding one.

The page does not speculate about the three men charged on 28 July, in either direction. It does not suggest they are guilty, and it does not suggest they are carrying blame for someone else's decisions. Both would be inference about people facing an active criminal charge, and an archive should manufacture neither.

One answer reaches outside the corpus and says so. The question about international standards is answered against the IMO's Casualty Investigation Code, its 2022 Model Regulations on Domestic Ferry Safety, its guidelines for investigators and the joint IMO/ILO guidelines on the fair treatment of seafarers. Each is linked. Every other answer on the site reasons only over documents the archive holds.

What the international-standards answer found, which is not what was expected

The obvious version of that answer would have been that Guyana is breaching its international obligations. It is not, and the page says so plainly.

SOLAS, the Load Lines Convention and — through SOLAS — the binding force of the Casualty Investigation Code all attach to ships on international voyages. The MV Barima ran between two Guyanese ports. It sat outside all of them, was never required to hold a SOLAS passenger safety certificate, and was not evading one. The IMO adopted model regulations for domestic ferries in 2022 precisely because of that gap, and made them voluntary.

The argument that survives is narrower and stands up better: the Casualty Investigation Code is the recognised international standard, Guyana already applies it to its ships on international voyages, and there is no principled reason to apply a lower standard of independence, evidence preservation and publication to a domestic route. Measured against that sequence — the first day, securing the evidence, standing up an independent investigation, the families, publishing — four things stand out, and they are set out with a stage-by-stage comparison table on the page.

One stage is deliberately not measured against anything. The international instruments deal with seafarers, not with bereaved families. Rather than dress a Guyanese civil-society proposal as an international norm, that stage names the Guyana Human Rights Association as its source and says there is no maritime standard to measure against.

Charts, and one that was thrown away

Four charts carry the numbers: the load against each stated limit, the complement split into rescued, recovered and unaccounted, the formal action taken by organisation, and the days from the sinking to each step toward an inquiry. Each value is labelled in text, each chart has a table view, and none of them carries meaning by colour alone. The palette was checked for colour-vision separation and contrast against both the light and dark backgrounds rather than judged by eye.

The first version of the accountability chart was discarded. It drew "3 charged", "at least 4 detained" and "1 team suspended" as bars of increasing length, so the longest bar carried the smallest number. Bar length has to encode the value or the chart is lying. It now shows one segment per person charged, and nought is drawn as a stub at the origin rather than as a full-width rule that would read as a bar.

The navigation, and the landing page

Adding a page took the navigation to eight items. They are now grouped, with a thin rule between groups: the record, voices, analysis, and the archive behind it. Nothing was renamed and no link changed.

Eight items also broke the header at two widths — the theme button dropped onto a third row on a laptop, and a sticky header ate a third of a phone screen. Below 820px the navigation is now a single scrollable row with the current page scrolled into view.

The landing page now carries a panel listing the five questions, each linking to its answer. It is generated from the questions page itself, so the two cannot drift apart.

The section was written twice, and the first version was replaced

The first version of this page was fifteen dense analytical questions written for a reader who already knew the case. It was researched, checked, corrected and published — and then replaced, at the site owner's direction, with five questions in plain language for a reader who does not. That is recorded here rather than quietly dropped, because the earlier version shaped the corrections below and much of its research survives inside the five.

What the checking found, in both versions

Every draft was put through the same adversarial process the record pages get: each number and quotation verified against the corpus, each claim of absence tested by actively hunting for a counterexample, and the whole page read again by a hostile critic from the government's side and from the opposition's. Every pass found real problems. They are recorded here rather than quietly fixed, because a page that reasons in public has to show its own corrections.

A claim of absence failed, and is withdrawn. An early draft said the safety measures announced in late July contained no new rule, only enforcement of existing ones. That is wrong. The Prime Minister announced expanded CCTV coverage and improved monitoring of wharves and loading areas — new capability, not enforcement. This is the third absence claim this archive has had to withdraw, and the reason is always the same: an absence is only as good as the search behind it.

Attributions drifted, twice, and were fixed twice. A draft put quotation marks around words a source gives as a reporter's indirect speech, and attributed to MP Terrence Campbell a framing that belongs to Opposition Leader Azruddin Mohamed. The rewrite that fixed it then did the same thing to the same speaker in the next paragraph. Campbell's own words — "Accountability must go from top to bottom" — are what appear now. A quotation from the Kaieteur News "Peeping Tom" column had also been extended past the part the source gives as verbatim, in a file this archive flags as a condensed capture, where its own standing rule forbids exactly that.

Contested facts had been quietly hardened. Drafts asserted who located the wreck, that the sister ship was withdrawn in 2019, and that a design fault removed the replacement vessel from the Port Kaituma route. All three are contested or unresolved on the record pages. They now carry their contests.

A single source had become several. A draft said "several survivors describe cargo stacked to the deckhead." Only one survivor in the corpus describes cargo, and at the rear of the vessel. The figures page already recorded her as the sole source, and the draft had contradicted it.

The most-cited document was read the way this archive says not to read it. The March 2026 tender's bill of quantities requires that deteriorated hull sections be renewed where necessary. That is a specification anticipating deterioration, not a survey establishing it — a distinction the figures page states explicitly. It is now described as what it is: an account of a document this archive does not hold.

Language about the accused was rewritten. Several sentences could be read as bearing on the guilt or innocence of the three men charged, in both directions — a hypothesis about who sent the distress call, an assertion that 46 people boarded unrecorded, a framing that presented the charges as falling within a category a minister had identified nine days earlier, and a chart that ranked the accused at the bottom of a hierarchy of responsibility. All are gone. The presumption of innocence now appears in the chart itself rather than only in a tooltip.

Balance. The hostile read found that every concession made to the government was immediately turned against it, while no opposition claim received the same treatment. The answers now carry the counter-arguments the corpus actually contains, including the "Peeping Tom" argument on ministerial responsibility, the government's own division-of-responsibility position, the argument that age is not unseaworthiness, and the government's stated case for the commission it appointed.

The site was redesigned to look like what it is

The presentation was sharpened toward an institutional register: a masthead with a heavy rule beneath it, uppercase letterspaced metadata, visible hairlines, square corners, and colour reserved for data rather than decoration. Nothing about the record changed; the point was that a public record should look like a public record rather than like a product page.

The navigation is now two tiers. The record itself — overview, chronology, figures, positions, questions — sits on the primary row. The apparatus behind it — documents, method and gaps, revisions — sits on a quieter utility row beneath, with the theme control. Everything stays one click away and nothing hides behind a menu, which matters on a reference site. The theme control itself is now a small sun-and-moon icon rather than a labelled button; its accessible name carries the meaning for anyone using a screen reader.

Two layout faults were fixed on the way. The document browser's filter bar was overflowing a phone screen and forcing the whole page to scroll sideways; it now folds into a two-column grid. And anchored links were landing underneath the sticky masthead.

Each answer now carries a reasoned view, fenced off from the evidence

The questions page previously stopped where the documents stopped. Each answer now ends with a block headed the reasoned view, which argues past what any single document states: what follows from the evidence, which competing readings are available, which is better supported, and what would change the conclusion.

That is a real change in what this site does, so the block is fenced. It has its own rule and label, it carries a standing note that it is inference rather than record, and every one of the five ends by naming what evidence would overturn it. The record pages remain authoritative over all of it.

These blocks were then attacked harder than anything else on the site, because inference is where an archive is most likely to go wrong, and the attack found a great deal.

Two claims of absence were false. The reasoning said the archive contained no document asking whether the vessel should still be carrying passengers. It contains several: a master mariner in 2015, residents of Port Kaituma in 2010, and the Amerindian Peoples Association on behalf of passengers over a period of years. The corrected version says what is actually true — the question was asked from outside and no body inside the operator, regulator or ministry is recorded answering it. Separately, a claim that "no board decision explains" the route assignment was asserted as a finding while the same paragraph conceded the archive holds no board minutes.

A logical test was stated backwards. The but-for test used to weigh preventive measures was written as "a measure counts if the disaster does not happen without it," which is the negation of what was meant and would have made the whole section self-defeating.

A false dichotomy was claimed as exhaustive. The reasoning said only two explanations could account for three untraceable capacity figures. At least three more fit, including a gross-tonnage figure misread as a cargo limit — a risk this archive's own figures page already flags.

Parsimony was used to license a conclusion it cannot license. The reasoning held that because legal caution sufficiently explains the absence of a published death toll, other explanations "are not needed." That is not what parsimony does: a sufficient cause is not thereby the only operative cause. And the explanation was not sufficient anyway — legal caution explains not certifying deaths; it does not explain declining to publish an estimate, which states routinely do and which this archive does on the same page.

One passage exculpated the operator, which breaches the page's own limit in the opposite direction from the one usually feared. A sentence describing three institutions as "each discharging its own function competently enough" and the failure as "not a bad decision" was both contradicted by the government's own 2025 finding of "longstanding issues of negligence, mismanagement, and accountability" at the department, and a view on the merits of operational conduct now before a court. The page's stated limit is that it takes no view on the guilt or innocence of the three accused. A sentence favouring them breaks that limit as squarely as one against them.

The reasoning leaned one way through under-supported steps, and that has been corrected. All five blocks landed on the state's capital and regulatory layer. Some of that is the evidence — the archive genuinely records no formal action of any kind above the operator. But in three of the five the landing was reached through a step that did not hold. Three counter-arguments that were available in the corpus and had not been made are now made: MARAD's stability argument, which is the only technical causal argument in the corpus and points at the vessel; the Associated Press finding that unmanifested cash ticketing was a practice of decades' standing at state terminals; and the government's own 2025 admission about the department. A claim that Guyana's administration must already have the capability to investigate to the international standard has been withdrawn, because the archive contains no record of it ever doing so.

A submission route that works, and the contact-address problem is solved

The invitation to send a question was published before there was any way to accept one. There is now an ask a question page carrying a form, which delivers by email through Web3Forms. It takes three kinds of submission: a question for the questions page, a correction to something on this site, and a document the archive is missing.

No email address is published anywhere on the site, which keeps it out of reach of address harvesters. The form is the contact route, and the note on the method page saying no address existed has come down.

Name and email on the form are optional, and neither is published unless the sender asks for it. The form works with JavaScript switched off — it falls back to the provider's own confirmation page, which the page says out loud rather than leaving as a surprise.

The short answers now carry the reasoning too

Each question opened with the short answer and hid the reasoned view behind "keep reading," which put the archive's actual conclusion two clicks from the reader. Every question now shows a one-paragraph distillation of its reasoned view alongside the short answer, marked with the same section sign as the full block so the two read as the same voice.

The rules were dialled back

The heavy rule earns its place under the masthead, where it separates the identity of the site from its contents. Repeated down the page — under every section heading, above the footer, around every panel — it stopped signalling anything. Everything below the masthead is now a hairline or a mid weight, and the labels carry the hierarchy instead.

The tooling is now stated on the site

The research, retrieval, cross-checking and reasoning behind this archive were carried out with Claude Opus 5, Anthropic's frontier model. That is now said on the method page, in the footer of every page, and on the questions page itself, rather than being left for a reader to infer. The model is a tool; the sources remain the authority, and every claim on the record pages is traceable to a named published document.

Shared links now carry a card

Every page has Open Graph and Twitter card metadata and its own 1200×630 image, so a link posted to a social platform, a group chat or a messaging app shows the page's title, one line of what it contains, and the archive's masthead — rather than a bare URL or whatever image a crawler happened to find first.

The cards are rendered at build time in the same browser the site is tested in, so the type on a card is the type on the page. Each carries the page title, a line of subtext written for the crop, and a footing line giving the document count and the compile date. The 209 source pages share one card and each carries its own canonical URL, title and description.

The corpus grew to 233, and the archive corrected itself twice

Twenty-four documents were added covering 28 and 29 July, each verified by fetching the published article rather than trusting a search result. Four of them are social media posts, labelled as such.

One of those four records something no newspaper here reported. On 29 July the Amerindian Peoples Association picketed the Office of the President, demanding accountability for the disaster and the removal of Ministers Edghill and Indar. Two Guyanese news accounts posted it to Facebook; a search of Guyanese outlets the same day found no press report of it, and none appears in this corpus. A street protest at the seat of government over the country's worst maritime disaster, and the archive can find no newspaper account. The absence is recorded on the positions page alongside the event.

Also added: Minister Edghill's first response to the resignation demands, which have run since 21 July. He does not engage them — "I have repeatedly said I will speak about this matter at the appropriate time", and on the protests, "My life is in God's hands and that's not today".

The archive published something it could not support, and has withdrawn it. An earlier reading of Demerara Waves' 28 July report had the Director of Public Prosecutions acknowledging that the vessel's navigational and safety systems were defective, and treated that as a state admission. It is not one. The line is the reporting outlet's characterisation in its own lead; there is no quotation of the DPP saying it, iNews reproduces the same release at near-full length without it, and HGPTV omits it too. It is now recorded as a single-source characterisation and explicitly not as an admission. This was the archive's error, not the outlet's.

The three accused are now named, as published across a dozen outlets: Captain Kevin Price, 40; Chief Mate Rondell Dwayne Roberts, 42; Goods Superintendent Delon Granderson, 33. Every page that names them states that they have not been tried and that this archive takes no view on guilt or innocence. Roberts's address is contested between two renderings and his forename appears two ways; both are recorded rather than resolved.

The charge count is 72, and it does not reconcile. Seventy-two counts of murder were laid jointly against the three, not seventy-two each. The figure is corroborated by five outlets but derives from one place only — defence counsel's account of prosecution disclosure, that 72 bodies were recovered and the accused were charged for all 72 deaths. Every official figure from 24 to 29 July gave 73 bodies recovered; the government's own release of 30 July then gave 72. (The charge particulars were published on 30 July and are recorded in the twelfth-build entry above.) No official release states any count at all. Both figures are recorded and the gap is left open, because closing it would mean choosing.

Two figures were corrected and one discrepancy closed. Of the 73 recovered, the record now reads 68 identified and 5 unidentified; it previously carried 69 identified, which was the 24 July figure, and no source explains the movement. The long-standing 76-versus-77 survivor difference is resolved: 77 were rescued and 76 survived, one having died after rescue.

New positions on the record: the DPP's defence of the charges and her statement that the inquiry and the prosecution can proceed simultaneously; Nigel Hughes and Darren Wade on the charges potentially silencing inquiry witnesses; Dexter Todd calling them holding charges; Christopher Ram arguing the inquiry should have been constituted under Part XVIII of the Shipping Act, now a third candidate statute with nothing yet gazetted; Opposition Leader Azruddin Mohamed's public offer to raise the wreck, and the government moving within a day from "likely to remain there as is" to acknowledging multiple salvage proposals and deferring to families.

A former captain of the vessel is on the record, and the archive has had to argue against itself. Andrew Donald, published by HGPTV and originating in a Member of Parliament's Facebook post, asserts the vessel was single-hulled contrary to SOLAS, lacked hull tanks and an adequate bilge system, and that wharf staff and managers controlled loading and sailing times while responsibility fell on the captain. It is single-source, HGPTV states it is not independently verified, and Donald names no individuals. It is recorded with all of that attached — and the reasoned view on Question 1 now says plainly that if his account is right it cuts against that view's own separation of operational from capital decisions. The reasoned views are the archive's inference, and inference that will not concede when the evidence moves is not worth publishing.

Social media sources are now labelled, and the rule is stated

Facebook was used as a lead source for this build. The rule is on the method page: institutions and public figures speaking on the record only, no private individuals, no comment threads. The first days of social media around a mass-casualty event carry the names of people reported dead who were alive, and an archive that fixed that in place would be doing harm in the name of completeness.

Where a post pointed at something real it was chased to published journalism and the journalism is what carries the weight. Two posts are held as documents in their own right: the President's statement on engagements with survivors and families, and the Member of Parliament's post carrying Donald's statement, which is the origin of an account four other outlets never ran. Both are marked with a red badge on the card, a red badge on the page and a standing warning above the text. The image attached to the second carries Facebook's own "AI-generated content" label and is recorded as such so that nobody reproduces it as a photograph.

A new page: what doesn't add up

It now leads the landing page. The analysis is the strongest thing on this site, and burying it beneath the questions panel understated it. It takes the top slot, previews all seven of its plain-language points in full, and the questions panel follows underneath. Each preview line leads with the claim rather than the setup, because a reader scanning a homepage gets one line to decide whether any of this matters.

One overreach caught in the same pass. The first draft of the short version said flatly that the boat "was about 70 tonnes overweight". Finding 1 further down the same page says the opposite — that whether she was overloaded is inference and not settled in either direction. A summary that contradicts the finding it summarises is worse than no summary. The lead now says what is documented, which is that a volume measure was used publicly as a licensed tonnage and nobody has corrected it, and points to the finding for why the overloading inference might be wrong.

It sits under questions rather than in the top navigation, which stays at five items, and it has a panel of its own on the landing page directly beneath the questions panel. It is linked from the top of the questions page and from every page footer.

It opens with a plain-language summary. Seven points, no jargon and no hedging, before any of the careful version. The archive's usual register is deliberately cold, and on this page that risks the analysis being legible only to people who already follow the story. The short version says what it comes to: that the state's proof the vessel was not overloaded appears to use a measure of interior volume rather than carrying weight; that one sheet of paper would settle most of it and has not been produced; that families were asked to accept the wreck as a permanent memorial two days before the assessment that justifies leaving it; that the minister located the fault at ground level on day one and the charges landed there nine days later; that disclosure was most forthcoming in the first forty-eight hours; that there are seventy-two charges against seventy-three bodies; and that no post-mortem is mentioned anywhere in the corpus. The detail, the counter-arguments and the sourcing follow underneath, unchanged.

Until now this site did two things — it recorded what documents say, and on one page it reasoned across them. It did not systematically ask what in the record fails to reconcile. It does now, on a new page, what doesn't add up.

Seventeen findings, ranked by how much each matters. For every one: the anomaly with the documents and dates that establish it, the innocent explanation given at its genuine strongest, the alternative stated flatly, an honest answer on which the record actually supports — including "neither, yet" where that is true — and the document that would settle it. The page accuses nobody of a crime. Three men are charged and untried, and nothing on it bears on their guilt or innocence.

The largest finding is a unit. The vessel's capacity was given publicly as 284 tonnes, then 126, then 120. This archive has recorded those for a week as three irreconcilable figures. They are not three guesses at one number — they are different quantities. The vessel is recorded at 284 gross tonnage and 198 deadweight. Gross tonnage measures enclosed volume, not carrying capacity. The manifest recorded 268 tonnes. On the most likely reading, the minister's day-one assurance that the vessel was "operating below its licensed limits" compared a cargo weight against a volume measure, and the real carrying figure was roughly 198 tonnes — meaning the manifested load exceeded it by about seventy. Opposition Leader Mohamed advanced the same arithmetic on 29 July, citing 198 deadweight, without reference to the source this archive holds. Two unconnected records give 198 and this archive had never joined them.

The page opens by correcting itself four times, because an analysis page that will not audit its own record has not earned the reader's attention. The 284 was previously flagged as possible contamination from another vessel's specification; it is a category error instead. Two "discrepancies" this archive asserted were its own arithmetic failures — 162 passengers plus 17 crew is 179, and the two search-area figures reconcile to within a rounding of each other. And the Mayday contradiction was overstated.

On that last one. This archive recorded the Prime Minister's "We have no such record" as a flat contradiction of the 23:01 distress call. Read exactly, he was denying a compound proposition — that the captain issued a Mayday and told passengers to don life jackets — which is a statement about the captain's conduct, not about whether a signal reached the state. The state's own Department of Public Information recorded the 23:01 call on 19 July. The figures page, the chronology and the positions page have all been corrected. What survives is harder and is kept at full weight: no government statement anywhere in this corpus identifies who received the call, and the rescue coordination log, publicly requested on 25 July, has never been produced.

One claim has been withdrawn for want of a source. An earlier note in this session reported that police barricaded the Charity waterfront and admitted only state media. A targeted search of every document in the corpus finds no such document. It is not recorded anywhere on the site and will not be until it is sourced.

Four of the seventeen cohere, and the page says what they amount to without asserting more than it can carry: at every point where the state's account could have been made checkable against a document, the document has not appeared — and conclusions have repeatedly been asserted that only that document could support. That is what an administration looks like when its early public statements outran its records. It is also what a cover-up looks like. The record as it stands cannot distinguish between those two, and the page says so rather than choosing.

The single document that would most change the picture is named: the vessel's Certificate of Survey and Load Line Certificate, with the registry entry giving gross tonnage and deadweight. One page decides whether 284 was a unit error, whether 268 tonnes was inside the limit or seventy tonnes outside it, whether the 394-passenger rating exists on paper and when it rose from 150 — and, because deficiencies are noted at the back of a certificate, whether anyone had recorded defects on that hull before 18 July. It is the cheapest document on the list, and the regulator has already said publicly that it relied on it.

Record-page corrections found along the way

Two dates were wrong. The complement figure of 179 was dated to 21 July. It was published on 20 July and restated by the Prime Minister on 21 July. And the tightening of ferry safety procedures was dated to 25 July, which is when the Jamaica Observer carried it; the DPI release and the Prime Minister's interview are both dated 24 July.

An attribution was over-precise. The 179 was described as attributed to the government's National Communications Network. The NCN is the named source for the 69 rescued and 83 missing; the 179 is attributed to "the government" generally and, in wire copy, to the Prime Minister.

A claim of absence on the figures page was false. That page stated that no official had characterised the payments made to families as "not compensation." The Prime Minister did, on 24 July: "That is not deemed compensation. That is the government's responsibility in times like this to assist families." Corrected.

A units discrepancy is now stated rather than silently resolved. The figures page rendered the cargo tonnages as tons and the positions page as tonnes. Most of the corpus says tonnes. The archive now uses the dominant rendering and says on the figures page that the sources disagree, rather than converting quietly between units in a comparison that turns on the size of the numbers.

Two claims went further than their sources. The chronology asserted that the Kalliopi N was still at the Kingston Goods Wharf on 18 July; what the corpus supports is that it had not entered service. And it filed the fisherman's location of the hull in a way that read as contradicting the state's own 13:31 timestamp on 20 July, when the two accounts are sequential. Both are corrected.

28 July 2026 — ninth build, charges laid

The largest single day of developments since the sinking. Three documents added; the corpus stands at 209.

Three employees of the operator are charged with murder

The archive had recorded, since first publication, that no charges had been filed. That is no longer true. On 28 July the captain, chief mate and goods superintendent of the MV Barima — all employees of the Transport & Harbours Department — were jointly charged with murder contrary to common law, on legal advice from the Director of Public Prosecutions. The indictable charge was read at the Georgetown Magistrates' Court by Chief Magistrate Judy Latchman, who remanded all three. The matter moves to the Charity Magistrate's Court on 3 August.

All three are accused persons who have not been tried, and nothing on this site should be read as a finding against them.

One fact about the charges is recorded because it is a fact about the record, not a judgement: every person charged is an employee of the operator. No charge against any official of the regulator, any board member, or any minister appears in the corpus.

Parliament sat, and did not resolve the inquiry question

The National Assembly met on Monday 27 July. It passed a G$54.9bn supplementary budget and several bills while opposition MPs from WIN, APNU and FGM protested inside the chamber, some displaying photographs of victims, and AFC and VPAC members protested outside. The Speaker briefly suspended the sitting. The Attorney General and the Prime Minister addressed the House on the tragedy. The demand has widened: the opposition now calls for the resignation of Minister Deodat Indar as well as Minister Edghill, the first reported inclusion of Indar.

Single source APNU's Dr Terrence Campbell states that Transport & Harbours employees told the opposition they had been instructed not to speak to the press or to opposition members. That allegation is recorded as his, and is not corroborated elsewhere in the corpus.

The government signals the wreck may stay where it is

On 28 July the Prime Minister stated the vessel is "likely to remain there as is", that no one has been contacted to raise it, that attempts to move it onto its keel had not succeeded, and that "it's obvious that bodies are trapped in the wreck". A draft "Declaration of Agreement" circulated to families on 25 July asks them to accept the wreck as an undisturbed "sacred memorial" with a monument bearing the victims' names.

Relatives object, and their objection is recorded alongside the government's position: that the proposal may wrap up salvage prematurely, and may obscure physical evidence before the Commission of Inquiry can inspect the hull. Kaieteur News sets the position against President Ali's assurance of 20 July that the effort would continue "until every possible avenue has been exhausted". This archive takes no position on which course is right.

Still open

The establishing instrument remains ungazetted. No secretary or reporting deadline has been named. The toll stands where it has since 24 July: 73 recovered, 76 rescued, at least 30 unaccounted for.

27 July 2026 — eighth build, fresh reporting swept

A sweep of the Guyanese outlets for anything published since the last build. Five documents added; the corpus stands at 206.

The status has not moved

The toll is unchanged at 73 recovered, 76 rescued and at least 30 unaccounted for — the same figures since 24 July. No charges appear. The Commission's establishing instruments are still not gazetted: on 27 July the President's own words remain that they "are currently being formalised". Kaieteur News reports that the Speaker had not yet ruled on APNU's Standing Order 12 motion, so Parliament has not resolved anything either.

A third source on the commissioners

Kaieteur News of 27 July names all five independently of the Guyana Chronicle and Kiskadee Watch. It adds that the chair, Godfrey Smith, previously served on the Commission of Inquiry into Guyana's disputed 2 March 2020 general elections.

Hinterland leaders, in their own names

Two Indigenous and community leaders have called for Minister Edghill's resignation personally, distinct from the Amerindian Peoples Association's institutional call: Kako Toshao Romario Hastings and Chris Debideen, Chairman of the Aruau Community Development Council, who states he lost two relatives. Both are now on the positions page. Hastings's statement about silence among Toshaos — that "every time accountability is traded for access, our people ultimately pay the greater price" — is recorded as his own characterisation.

The argument against resignation is now on the record

The corpus had accumulated many published demands for the Minister's resignation and no sustained argument against. Kaieteur News's "Peeping Tom" column of 27 July supplies one, and it is recorded on the positions page at the same length as the demands: that the Transport & Harbours Department's own board insulates the Minister from operational responsibility, that political accountability should be tested by whether Cabinet support and non-partisan public trust have actually been lost rather than by the volume of demands, that the cause is not yet established, and that the 2017 Camp Street prison fire — eighteen dead, a government commission finding a "well executed plan" enabled by overcrowding — produced no ministerial resignation, raising a consistency question. It carries the newspaper's disclaimer that the views are the author's.

The stelling, documented

Kiskadee Watch has published a chronology of the Port Kaituma stelling — the works the Prime Minister cited as the reason the MV Ma Lisha could not serve the route. It names the parties for the first time in this corpus: International Import and Supplies, awarded the G$1.4bn construction contract by public tender on 30 December 2022, and CEMCO as designer and supervisor. It records the project at 77 per cent complete per the Minister in January 2024 and about 90 per cent per the contractor in February 2024, when fissures halted work; the contractor's statement that the cause was design deficiency rather than construction; and the Minister's August 2024 confirmation that the design consultant had acknowledged flaws. Approaching four years after the contract, the project is unfinished. No finding of fault against any named company appears; the statements are recorded as made.

One contested figure resolved

The French deployment was recorded as 12 divers plus two doctors and a mechanic per BBC News against 15 divers per Ignite News. Kiskadee Watch gives a 15-member team from the French Armed Forces in French Guiana, so the two accounts describe the same deployment. The figures page now says so.

And one operational finding

The French dive team told relatives directly that it is too dangerous to enter the wreck — ropes and hammocks inside, buried in mud, no visibility, strong tidal currents — and that lifting the vessel would be "very, very difficult". That sits against the Prime Minister's statements of 24–25 July that boats were standing by for a possible righting operation. Both are on the chronology.

27 July 2026 — seventh build, consistency audit

A check across every page for stale counts, stale dates, and findings that had been added to the archive but never reached the pages a reader actually reads.

Counts

The document total had moved to 201 with Ignite News, but several figures had not followed: the provenance table still totalled 200, the state share read 62 of 200, the independent-reporting figure read 84 of 200, and the chronology and figures pages were still described as checked against 200 and 152 documents respectively. All corrected. Original reporting on the sinking is 125, not 124, and independent original reporting is 85 of 201.

Dates

Every page footer read "Compiled Sunday 26 July 2026" although the archive has been revised repeatedly on 27 July, including the addition of a 27 July document. Footers now read compiled 26 July, last revised 27 July, and the chronology, figures and method pages carry the same. The overview's summary of where the parties stand was dated "as of 26 July"; it is now as of 27 July.

Findings that had not reached the record

This was the substantive part. Material added over the previous three builds was sitting in this revisions log and on the positions page, but had never been written into the chronology or the figures page — so a reader consulting the record itself still got the older picture.

Now added to the chronology: the pre-departure accounts from Kiskadee Watch, set against Minister Edghill's statements; and the identification of air traffic control as receiving the distress call at about 23:00, reported independently by Deutsche Welle, CBC News and Ignite News, alongside BBC News giving the time as 23:01. Four outlets place the call and three name the recipient, against the Prime Minister's "We have no such record."

Now added to the figures page: the pre-departure machinery accounts under proximate cause, and five further discrepancies — the first search-area expansion given as 1,040 sq km by Reuters and CBC News against 1,070 by the Guardian and Ignite News; the French deployment as 12 divers, two doctors and a mechanic per BBC News against 15 divers per Ignite News; the location as off Iron Punt in most of the corpus against seven miles off Dartmouth per Ignite News; the two renderings of the lifesaving equipment; and who received the distress call.

27 July 2026 — sixth build, wording made consistent and Ignite News added

"Sources" and "documents" were being used interchangeably

The previous build changed the headline count to documents but left the word "sources" in the places a reader actually meets it: the browse page heading, the filter counter, the breadcrumb on all 201 source pages, and the navigation label. All now say documents. The word "source" is kept only where it means the origin of a particular claim, which is a different and correct use.

The chronology page was still described as checked against 152 sources, and the method page still said the corpus "skews heavily" to Kaieteur News at 32 — the same claim, now inverted, that was corrected on the overview in the fourth build. Both fixed.

Ignite News added

Ignite News is the Guyanese outlet CBC News credits as the origin of survivor Leon Murray's account. Its own 18–21 July chronology is now in the corpus. The specific Murray interview was not found on the outlet's site; CBC's account of it is here, the original is not, and that remains recorded as a gap.

It gives a third independent statement that air traffic control received the distress call at about 23:00, after Deutsche Welle and CBC News. Three outlets now name the recipient, against the Prime Minister's 25 July "We have no such record."

It also settles one figure conflict and opens two. The search area after the first expansion is given as 1,070 square kilometres, matching the Guardian, against 1,040 in Reuters and CBC. The French deployment is given as 15 divers, where BBC News gives 12 divers, two doctors and a mechanic. And the capsizing is placed seven miles offshore Dartmouth on the Essequibo Coast, where most of the corpus places it off Iron Punt near the Pomeroon mouth. All three are recorded, none adjudicated.

27 July 2026 — fifth build, outlet names deduplicated

The archive had been recording 51 distinct publisher names for what were in fact 42 publishers. HGPTV appeared five different ways; Demerara Waves, iNews Guyana, Democracy Guyana, the Department of Public Information and Wikipedia twice each. Twenty records were relabelled to a single canonical name per publisher.

This matters beyond tidiness. The fragmentation made the outlet filter list the same broadcaster five times, and it inflated the apparent breadth of the archive in the same direction as the source count did — making the record look drawn from more publishers than it is.

Merging the Department of Public Information's two spellings also moved its total from 21 to 22, so the state's share of the archive is 62 of 200 rather than 61. The overview panel and the provenance breakdown are corrected, and those figures are now computed at build time rather than written by hand.

Four sets of similar names were deliberately left separate because they are different publishers: BBC News, Canada's CBC News and the Caribbean Broadcasting Corporation in Barbados; the Jamaica Gleaner and Observer; the Guyana Chronicle and the Guyana Graphic; and iNews, News Room, News Source and Things Guyana. Associated Press entries remain separated by carrier, since which outlet carried a wire report is part of the record. CBC News is now labelled CBC News (Canada).

27 July 2026 — fourth build, prompted by a reader

"200 sources" overstated what this holds

A reader objected that calling these documents "sources" is misleading. The objection is correct and the framing has changed.

Each of the 200 is a distinct published document, so the count was never false. But a general reader hears "200 sources" as 200 independent accounts, and the composition does not support that reading: 18 are syndicated copies of other outlets' reporting, 18 are government communications, 14 are opinion or letters, 24 are pre-disaster context rather than coverage of the sinking, and four are duplicates or republications. Counting only original reporting on the sinking by outlets that are neither state-owned nor carrying another outlet's wire, the figure is 84.

The site now says documents rather than sources, and a full provenance breakdown is published on the method and gaps page so the number carries its own qualification. The underlying collection policy is unchanged: syndicated variants stay, because which outlet carried which figure on which day is itself part of the record.

A claim on the overview had inverted

The overview carried a panel headed "Why this corpus leans on one outlet", stating that Kaieteur News accounted for 32 of the sources. That was true when written. It is no longer the shape of the archive: with 40 Guyana Chronicle documents added on 26 July, the largest single share is now the state-owned newspaper, not the independent one. The panel is now headed "Who published this record" and states the position accurately — 61 of 200 documents, about three in ten, come from the state or a state-owned outlet, with Kaieteur at 32 the largest independent share.

Stale counts corrected

Six places on the overview and in the footer still read 192 after the corpus reached 200. Counts are now computed from the record set at build time rather than written by hand, so they cannot drift again.

Outlet counts were softer than they looked

The archive holds 51 distinct publisher strings but only 41 actual publishers; HGPTV appeared under four spellings, Demerara Waves and iNews under two each. Recorded on the method and gaps page.

27 July 2026 — third build

Chronicle source dates repaired

All 40 Guyana Chronicle sources added in the previous build carried no date. They had been written into the archive with a date field where the build reads published, so each resolved to a null value. On the sources page they appeared under "Undated", sorted to the end, and — because a null date compares as earlier than any real date — the period filter placed all 40 under "Before the sinking", the one period they could not belong to. Their dates now read correctly, 22 to 26 July, and all 40 fall under "22 July onward".

BBC and Reuters: a claim of absence withdrawn

The site recorded that no BBC byline and no directly-bylined Reuters article had been found, flagged as a limit of search reach rather than proof of absence. That caution was warranted: both exist. BBC News published on 19 July, updated 22 July, bylined Vanessa Buschschlüter and Tom Bennett; Reuters published on 20 July, bylined Kemol King. Both are now in the corpus, and the earlier statement has been withdrawn on the method and gaps page.

Both were found through the reference list of the English Wikipedia article on the vessel, which had not been consulted before and is now itself a source. The failure was one of method, not of availability.

Four further international sources added

The Guardian (on-the-ground reporting from Essequibo by Neil Marks and Natricia Duncan), Deutsche Welle, CBC News, and the Wikipedia article.

What these add to the record

Two of them bear on the Mayday contradiction. The site records that a distress call at 23:01 underpins every published chronology, and that the Prime Minister said on 25 July, "We have no such record." BBC News gives the time as 23:01 local; Deutsche Welle and CBC News independently state that air traffic control received the call at around 23:00. Three sources now place the call, and two name the body that received it. The contradiction with the 25 July statement is unresolved and is recorded as such.

Reuters names the private company VEHSI, working with Exxon Mobil and local fishermen, as having located the wreck — the only source in the corpus naming it.

A published inventory of lifesaving equipment now exists: 250 life jackets, six inflatable life rafts and two rigid life rafts, attributed to the Prime Minister by Deutsche Welle and CBC News and matching the Wikipedia entry. The Guardian renders the same equipment as "two lifeboats", attributed to Minister Edghill.

Two new figure conflicts are recorded rather than resolved. The search area after the first expansion is given as 1,040 sq km by Reuters and CBC News and as 1,070 sq km by the Guardian. The Wikipedia lede gives "at least 162 passengers and 17 crew", which does not match the 179 total used elsewhere in that same article.

Kiskadee Watch added, and what it changes

Kiskadee Watch, an independent Guyanese outlet credited by the BBC as the origin of Elena Moonsammy's account, is now in the corpus with two reports. Source count rises from 192 to 200.

The commissioners are corroborated. Its 27 July report names all five members independently of the Guyana Chronicle. The single-source qualification carried on the positions page has been removed.

The statutory basis is now on the record. The same report states the commission was appointed under Section 2(1) of the Commissions of Inquiry Act. The site previously recorded that no statutory basis had been cited by the President; that gap is closed. It also means the AFC's proposal to use that Act and the instrument actually used are consistent.

Survivor accounts on the machinery are now recorded alongside the Minister's. Minister Edghill stated on 19–20 July that "There was no engine problem or any mechanical problem" and on 22 July that there were "no reported engine troubles from the crew when the boat came in and when the boat left." Kiskadee Watch of 20 July carries a survivor stating the engine stopped twice before departure and that no pump was available when water entered, and a passenger stating the vessel was stuck on a sandbank at the wharf for almost an hour and then sailed on one engine with a mechanic having worked on it. A third account, already in the corpus via the Associated Press, places an engine failure about four hours out. This archive does not adjudicate between them. Both are on the positions page.

Commissioner backgrounds are recorded, including matters previously reported about three of them — each with its outcome stated, none resulting in a finding of wrongdoing — and an open question the outlet raises about a possible conflict of interest, which it states was unanswered at press time.

Still outstanding

Ignite News, credited by CBC as the origin of Leon Murray's account, has not been retrieved.

26 July 2026 — second build

The Commission of Inquiry was named

The site previously recorded that no commissioners had been named, no chair appointed and no terms of reference published. That is no longer the case, and those statements have been revised wherever they appeared: on the overview, on the positions page, and in the method and gaps page.

President Ali announced the five-member commission on Sunday 26 July. It is chaired by Justice Godfrey Phillip Smith of Belize, with Capt. Hamada Fouda of Jamaica, Nyree Dawn Alfonso of Trinidad and Tobago, Dr Andrzej Jasionowski of Poland, and Rear Admiral (Ret'd) Hayden Pritchard of Trinidad and Tobago. Terms of reference were announced in substance at the same time.

Single source at the time of this build, the commissioners' names and the terms of reference rest on the Guyana Chronicle's report of the announcement. No second outlet had published the names when this page was written. The badge will be removed once a second independent source is on the record.

Three related gaps remain open and are still recorded as open: the legal instruments were described in the announcement as "currently being finalised," so nothing is gazetted; no statutory basis was cited; and no secretary or reporting deadline has been named.

The Guyana Chronicle gap was closed

The first build recorded the state-owned Guyana Chronicle as the largest single gap in the corpus — the outlet was absent entirely, because every retrieval attempt hit a JavaScript redirect wall. Forty Chronicle articles have now been retrieved through a browser session and added, of which thirty-eight carry full text. Two are flagged as incomplete captures and are marked as such on their source pages.

This changes the shape of the corpus. The first build noted that the state's entire published output on the disaster ran through the Department of Public Information and ministers' social media accounts. With the Chronicle present, that is no longer accurate, and the observation has been revised.

The government responded on the IMO question

The first build recorded the demand for an International Maritime Organization investigation — from Transparency Institute Guyana, Rescue Guyana and the joint opposition parties — as unanswered by government. A Chronicle report of 25 July carries the Prime Minister saying: "If contacting the International Maritime Organization will help in this inquiry, we will do it." The positions page has been updated and the claim of no government response withdrawn.

Other additions

The Shipping Association of Guyana's statement welcoming the inquiry has been added to the positions page. Chronicle coverage of the search operation, the psychosocial support programme, the interfaith services, and the possibility of a full salvage as a last resort has been added to the chronology.

Source count rises from 152 to 192.

26 July 2026 — first build

Initial publication. 152 documents, compiled eight days after the sinking. Method, retrieval failures and the eighteen corrections made during an adversarial verification pass are set out on the method and gaps page.