At 10:47 and 27 seconds on the morning of June 18, 2023, a text message reached the support ship Polar Prince from 3,341 metres below it: “dropped two wts.” It meant the submersible Titan had released two of its ballast weights. The message before it, half an hour earlier, had been “all good here.” Six seconds later the automatic position ping arrived from 3,346 metres. Then nothing.

The five people inside were a 61-year-old engineer who had designed the craft and was flying it, a 77-year-old French Navy veteran making his thirty-eighth descent to the Titanic, a 58-year-old British aviator who had been to the bottom of the Challenger Deep and to the edge of space, and a Pakistani-British businessman of 48 with his 19-year-old son, who had brought a Rubik’s cube to solve at the wreck.

A month earlier, on the same ship, the editor of a travel trade paper had spent eight days waiting for weather that never came. He attended the seven o’clock briefings and wore a helmet and steel-toed boots on deck. He watched the man who designed the submersible order his crew to pause for five minutes after every complex repair and think about what they had just done. He wrote afterwards that he had been impressed by how careful it all looked. He also wrote down the thing that had bothered him, and in retrospect it was the only thing that mattered.

Who ran it

OceanGate was founded in 2009 in Everett, Washington, by Stockton Rush and Guillermo Söhnlein, who stepped away in 2013 as the company turned to engineering, which he said was more Rush’s strength than his. Rush had an aerospace engineering degree from Princeton, an MBA, and a story he told often: that he had flown jets at nineteen, that he had wanted to be an astronaut until his eyesight slipped to 20/25, and that he had turned to the ocean because it was the frontier that was left.

The company’s first submersibles were conventional: a steel-hulled craft bought second-hand and a successor, Cyclops 1, rated to 500 metres. Its customers were researchers, film crews and the curious. Its ambition from 2016 onward was the Titanic, at 3,800 metres, and a craft that could take five people there at a time, at a price.

The craft was Titan, and almost everything about it followed from one decision. A conventional deep submersible has a pressure hull that is a sphere of steel or titanium. A sphere spreads the load evenly, and metal under compression behaves in ways a century of engineering has measured. But a sphere for five people is large, heavy and expensive to carry. Rush wanted a cylinder, which seats five in a row. He wanted it made of carbon fibre, which is light enough that the craft would float without the bulky foam conventional subs need. Carbon fibre is superb in tension. Under the compression of the deep ocean it has a short and unhappy record, and no classification society had ever certified a hull like it for that depth.

Rush’s answer to that was the experiment. In February 2019 OceanGate published an essay on its website titled “Why Isn’t Titan Classed?” arguing that most marine accidents are operator error rather than mechanical failure, that classing assesses only a vessel’s physical state, and that bringing an outside body into a fast-moving design would be slow and would stifle innovation. In a 2019 magazine profile he said the Passenger Vessel Safety Act of 1993 had “needlessly prioritized passenger safety over commercial innovation.” In a 2021 interview he said he had broken some rules to build the craft, with logic and good engineering behind him. In a 2022 interview with CBS he said that at some point “safety just is pure waste.”

The people who paid $250,000 for a seat were not called passengers. They were called mission specialists, took part in the briefings, and signed a waiver that described the vessel as experimental, uncertified by any regulatory body, and capable of causing “physical injury, disability, emotional trauma or death.” One of them later said death appeared three times on the first page.

What he was trying to find out

Strip away the tourism and the Titanic, and OceanGate was running a test of a single hypothesis: that a carbon-fibre pressure hull could be cycled to 4,000 metres again and again without the destructive testing and third-party review that the industry considered the price of admission, because the hull would tell you before it failed. The instrument for hearing it was what the company called real-time hull monitoring: acoustic sensors and strain gauges bonded to the cylinder, feeding a display in the cabin, meant to detect the fibres beginning to crack and the structure beginning to buckle in time to drop weights and rise. Rush patented the system in 2021. The waiver, the “mission specialist” title and the blog post were all ways of saying the same thing: this is a trial, and the trial is the safety case.

It was not a hidden hypothesis. In March 2018, after a meeting of the Marine Technology Society’s committee on manned submersibles, its chairman William Kohnen drafted a letter to Rush saying the members had “unanimous concern” about the project and urging him to have the craft classed before carrying people. Thirty-eight people in the small world of deep submergence agreed to sign it. It was never formally sent. Kohnen testified later that Rush was not receptive to outside scrutiny.

The same month a Boeing engineer consulted on the composite work warned that the plan carried high risk at 4,000 metres, and an expedition operator who had worked with Rush, Rob McCallum, wrote to urge caution. Rush wrote to one critic that the warnings were a serious personal insult.

Inside the company the objection was raised in January 2018 by David Lochridge, the director of marine operations, a submersible pilot the company had hired to run its dives. He inspected the hull as it was being built, found what he described as flaws, and wrote a report refusing to sign off. His two sharpest points were about the window and the warning system. The acrylic viewport had been certified by its maker for a depth far short of the Titanic, 1,300 metres by his account. The company had declined the maker’s offer to build one rated for the full depth and designed its own instead. Lochridge said OceanGate meant to go to 4,000 metres with passengers who did not know this. And the acoustic monitoring, he said, would detect a failure only milliseconds before an implosion, which is not a warning but a recording.

He was called into a meeting and fired. OceanGate sued him, he countersued, and the suits were settled. He also took his concerns to the Occupational Safety and Health Administration; eight months later a caseworker told him there were eleven cases ahead of his, and the complaint went nowhere. At the hearings six years later he called the Titan “an abomination of a sub.”

What they built

The hull was a cylinder of carbon fibre 2.4 metres long with an internal diameter of 142 centimetres, wound by a contractor, Spencer Composites, in 2017 to a wall thickness of 127 millimetres and cured for a week at 137 degrees. At each end a titanium ring was glued to the carbon and a titanium hemisphere bolted to the ring; the forward dome carried the viewport. The whole craft was 6.7 metres long and weighed about nine and a half tonnes.

Inside, five people sat on the floor with their backs to the curve; there was one small toilet behind a curtain, no seats, no heating, a screen for the outside camera, and a video-game controller to fly it. There was no acoustic positioning on board: the ship tracked the craft and sent it bearings and distances by text. Before the first hull was wound, a one-third-scale model had been tested at the University of Washington; one survived a pressure equivalent to about 3,000 metres and others imploded. OceanGate’s website said the design had been developed with experts from NASA, Boeing and the university. Boeing and the university later said they had no part in designing, building or testing it. NASA said it had signed an agreement with the company but had neither tested nor built anything for it.

Hull 1 made its first dives in 2018, and in December 2018 Rush took it alone to 4,000 metres in the Bahamas. In April 2019, on a dive to 3,760 metres with four aboard, the submersible pilot Karl Stanley heard what he described to Rush afterwards as the sound of something breaking. A month later a crack was found in the hull during an inspection, and the hull was retired after about fifty dives, only three of them to 4,000 metres. In January 2020 the company de-rated the design to 3,000 metres on account of cyclic fatigue, the tendency of a composite to weaken a little with every loading.

Then it had a second hull made, by Electroimpact and Janicki Industries, with a different cure, in 2020 and 2021; this one passed a full-scale pressure test at a facility in Maryland, and the depth rating went back to 4,000.

Hull 2 went to the Titanic in the summers of 2021 and 2022, five expeditions a season, a week each. By one count the craft logged 118 malfunctions in those two seasons. A dive in 2021 left it stuck on the bottom, and the people inside spent twenty hours in it before they surfaced. In 2022 the ship lost track of it for five hours while a CBS correspondent was aboard, and on another dive a thruster came back wired in reverse.

In 2022 alone, by Rush’s account the following spring, twenty-eight non-staff passengers reached the wreck. And on July 15, 2022, on the eightieth dive of the programme, the hull spoke.

Dive 80

The crew of dive 80 heard a loud bang on the way up. The scientist aboard, Steven Ross, later testified that the explanation offered at the time was the hull shifting in its metal cradle. The monitoring system heard it too: both the acoustic sensors and the strain gauges recorded the event, and after it the strain readings never returned to where they had been. On the next three dives the relationship between depth and strain was no longer a straight line. Three years later the National Transportation Safety Board’s engineers concluded that the pressure vessel had most likely delaminated on or after dive 80, and was damaged again on dive 82 by something no one identified. From then on, every dive was taken in a hull whose strength no one knew.

Nobody acted. The Coast Guard’s investigation found that the 2022 monitoring data should have been analysed and acted upon, that it was not, that no preventive maintenance followed, and that the craft was then stored without proper protection through the off-season before being brought back for 2023. The company’s own reading of the strain data, the NTSB said, was flawed enough that it did not know the Titan should have been taken out of service after dive 80. The instrument that was the whole experiment’s safety case had done its job. The experiment had no procedure for believing it.

What it looked like from inside

The 2023 season was the first in which the Titan was towed to the site on its launch platform behind the Polar Prince, a 1959 icebreaker chartered because the previous ship’s rate had risen to $200,000 a week. The travel editor in the opening was Arnie Weissmann, editor in chief of Travel Weekly, an invited mission specialist on the season’s second mission. Weather stopped every dive on his mission and the two after it, so the June dive was the first of the year. His three-part account, written in the days the craft was missing, is the fullest account by an outsider of how the operation looked in its last weeks.

What he saw was care. Breakfast at six, an all-hands briefing at seven with a prioritised checklist on a monitor, helmets and life vests and steel toes on deck, a copy of Atul Gawande’s The Checklist Manifesto left on the lounge table by Rush, who told anyone finishing a complex repair to take five minutes before walking away and think through everything they had done. Dive protocols carried a list of factors that would cancel a dive automatically if enough of them stacked up, among them wind over ten knots, a new member of staff, and a journalist aboard; Weissmann noticed that last one and wondered what it implied about the dives on which no journalist was present.

On the fourth day the front of the launch platform went under at the end of its tow rope, with the submersible on it, and it took Rush and two divers more than half a day with a buoy and flotation bags to right it. Asked what worried him most about the dive itself, Rush said ghost nets. Asked how much danger the sub had been in, he made a joke about a submarine being underwater.

What bothered Weissmann was a conversation on the back deck over cigars. Rush said he had bought the hull’s carbon fibre at a discount from Boeing because it was past its shelf life for aircraft use. That did not matter, he said, because the dates were conservative, and Boeing and NASA had helped design and test the craft. Boeing told Travel Weekly it had no record of any such sale and no part in the design. Weissmann wrote that he had thought about that conversation a great deal in the week after.

He also wrote, with a candour the rest of the coverage lacks, why he had still wanted to go: because Paul-Henri Nargeolet, who had been to the wreck thirty-seven times and understood more than he did, was going too. When he met Hamish Harding at a conference in the Azores a week before the June dive and told him his impressions of the towing and the platform, Harding raised his eyebrows, was not overly concerned, and flew to Newfoundland the next morning.

What happened

The Polar Prince left St. John’s on June 16 and reached the site at a quarter past five on the morning of June 18. Harding had posted the night before that after the worst winter in Newfoundland in forty years a weather window had opened, and that this would probably be the only crewed mission to the Titanic that year. Five people boarded the Titan on its platform at about half past eight: Rush as pilot, Nargeolet, Harding, Shahzada Dawood and his son Suleman.

Shahzada Dawood, vice-chairman of the Pakistani company Engro and a Titanic enthusiast, had first booked the dive for himself and his wife, Christine, before the pandemic cancelled it. When it was rebooked she gave her seat to their son, a student at the University of Strathclyde in Glasgow, because he wanted to go. It was Father’s Day, and Suleman’s aunt said afterwards that he had gone mainly to please his father.

The craft detached at 9:18 and began its fall of about two and a half hours.

At 10:14, at 2,274 metres, it reported all good. At 10:47:27, at 3,341 metres, it reported dropping two weights. At 10:47:33 the last ping arrived. By the later investigators’ reconstruction the pressure vessel failed at about 3,880 metres, under about 4,930 pounds per square inch, in a local buckling of the carbon cylinder; the collapse was complete in milliseconds, and the five inside could not have known it had begun. People on the Polar Prince heard a bang about then and did not know what it was. At 11:04 a United States Navy acoustic array recorded a signal consistent with an implosion in the right place; because it matched no known submarine, it was logged as an anomaly, and the information went to the search commander as something short of definitive.

The Titan had been due on the surface at half past four. The Coast Guard was told at ten past seven that evening. For four days the world counted down the ninety-six hours of air the craft was said to carry. Ships, aircraft and remotely operated vehicles from three countries searched an area twice the size of Connecticut. A Canadian aircraft’s sonobuoys picked up banging noises at intervals that turned out to have nothing to do with it.

On the morning of June 22 a remotely operated vehicle run by Pelagic Research Services, the Odysseus 6K, went down from the deck of the Horizon Arctic. About five hours after it reached the bottom it found the tail cone, 500 metres from the Titanic’s bow, and then the forward and aft titanium domes, lying apart. That afternoon in Boston a Coast Guard admiral said the debris was consistent with a catastrophic loss of the pressure chamber. The wreckage and what remained of the five came ashore at St. John’s on June 28; the last of it was raised in October.

After

OceanGate suspended operations in July 2023 and has not resumed. The Coast Guard convened a Marine Board of Investigation, the highest level of inquiry it has, and held two weeks of public hearings in Charleston, South Carolina, in September 2024. Lochridge testified, and so did Stanley and Ross. The former engineering director, Tony Nissen, said he had been let go in 2019 after refusing to sign off on a season following a suspected lightning strike on the hull. Mission specialists who had dived and come back did not agree with each other. One said the Titan was never supposed to be safe and that anyone who felt safe in it was deluded. Another said she had found the company transparent, and that anyone could ask anything.

A Harvard oceanographer told reporters it came down to hubris and greed. A retired submarine captain said it was too simple to call them reckless villains, and both were right.

The board’s report, 327 pages, came out on August 5, 2025. Its chairman, Jason Neubauer, said the loss of five lives was preventable. The board found the company’s design, certification, maintenance and inspection process inadequate, and found that it had repeatedly put operational and financial goals ahead of safety. It found that the company had used the mission-specialist designation, and intimidation of its own staff, to stay clear of regulators. Rush had concentrated virtually all decision-making in himself and could ignore inspections, data and maintenance at will; before one test dive he had said he would get a dive in even if it killed him. The board also found a toxic workplace, a whistleblower system that did not work, and no place in the regulations for a vessel of novel design, and it made seventeen recommendations. It found that Rush’s conduct amounted to negligence, and noted that the question of charges had died with him.

The NTSB’s report followed on October 15, 2025: probable cause, an inadequate engineering process that never established the hull’s real strength, which built a cylinder with multiple anomalies, wrinkles machined away and patched, voids in the glue, porosity in the fibre, and then misread the one instrument that could have revealed its state. Canada’s Transportation Safety Board reported in June 2026 to the same effect. The families sued. Nargeolet’s asked for fifty million dollars.

What it taught

Start with what it was not. It was not a failure of the material to warn. The hull cracked audibly in 2019 and was replaced; the second hull banged on dive 80, shifted its strain readings and left them shifted, and gave seven more dives of warning, recorded and filed, before dive 88. It was not a failure of courage; everyone aboard had chosen it with open eyes and, in Nargeolet’s case, more eyes than anyone alive. And it was not an accident in the sense of a thing no one could have foreseen. Thirty-eight people foresaw it in writing in 2018.

A warning you have no procedure for believing is not a safety system. OceanGate’s whole case for skipping certification was that it would hear the hull failing in time. It did hear it. The sensors worked on dive 80 exactly as designed; what did not exist was a rule that said what a shifted strain curve meant and who had to stop the programme when it appeared. An instrument without a threshold and a hand on the switch is a recording device. The NTSB’s finding that the company “was unaware” the Titan needed to be withdrawn after dive 80 is the whole story in one clause: the knowledge was aboard, in the data, and no one was obliged to know it.

Checklists manage the risks you have already imagined. Everything Weissmann saw was real. The briefings, the helmets, the five-minute pause, the go/no-go list, the book on the table: these are the habits of people who take danger seriously, and they governed the launch, the platform and the swells, which are where you get hurt most days. None of them had a line for the hull, because the hull was the premise and the checklist was built on top of it. A visible culture of care around small risks is the most convincing thing an operation can show a visitor, and it says nothing about whether the big one has been examined. The Biosphere lost its oxygen to a wall nobody had put on the ledger; the Titan lost its hull to the one item the checklist assumed.

Calling it an experiment does not make the passengers consenting subjects. The waiver said experimental; the title said specialist; the briefings were open; and still the people on dive 88 did not know what dive 80 had recorded, any more than the 2018 passengers knew what depth the window was rated for. Consent to a risk that has been described is not consent to one that has been withheld. If the trial is the safety case, the subjects are entitled to the trial’s data, and an operator who would not show them that data has told you what kind of trial it is.

The expert in the next seat is not a test result. Weissmann wanted to dive because Nargeolet was diving. Harding boarded the day after being told about the platform because he was a seasoned explorer and the operator seemed competent. Rush himself piloted most of the dives, which he and his defenders offered as proof of his confidence, and it was proof of exactly that. Each person’s assurance was borrowed from someone else’s presence, and at the bottom of the chain there was no one holding a measurement. Courage is contagious in a way that engineering is not.

Certification is slow because materials do not care about the schedule. The industry’s record in deep submergence before 2023 was, as one witness put it, perfect, and the reason was not caution for its own sake. It was that steel and titanium spheres had been tested to destruction, cycled, inspected and signed for by people with nothing to gain from the dive. Rush was right that the rules were slow and that composites might one day do what he claimed. He was wrong about who should bear the cost of finding out.

An uncertified hull is a test article, and the thing you do with a test article is instrument it, cycle it, and break it on purpose in a tank, not sell seats in it. The rule he called pure waste is the accumulated record of every hull that failed before anyone wrote the rule.

Check the story before you learn from it. The viewport’s rating is 650 metres in one record, 1,000 in another, 1,300 in Lochridge’s testimony; its diameter is fifteen inches in the design documents and twenty-one in the designer’s conversation. The hull’s carbon fibre was Boeing surplus in Rush’s telling and nothing of the kind in Boeing’s.

The last message has been read in a lawsuit as a sign the crew knew something was wrong, and by others as an ordinary step near the bottom; both readings fit three words. Harding’s last post was on Instagram or on Facebook, depending on the paper. The crew of the Polar Prince heard a bang before that last text appeared on their screen, which for a while looked like a message sent after the end; the hearings showed that texts took fifteen to thirty seconds to reach the display.

What the record supports without argument is the sequence: a crack in 2019, a bang in 2022, a strain curve that never recovered, a winter in storage, and a dive on a morning when the weather finally allowed it. That is enough to learn from, and it is the part that was known before anyone got in.

Sources

• United States Coast Guard, Marine Board of Investigation, Report of the Marine Board of Investigation into the Implosion of the Submersible Titan, 4 August 2025, released 5 August 2025, with the board’s press release and chairman Jason Neubauer’s statement; and the board’s public hearings, Charleston, South Carolina, 16 to 27 September 2024, including the Hearing Overview Factual Presentation of 20 September 2024 (the final messages and their times). news.uscg.mil
• National Transportation Safety Board, Hull Failure of Submersible Titan, docket DCA23FM036, final report 15 October 2025, and its materials investigation of the hull. The probable cause, dives 80, 82 and 88, the depth and pressure at failure, and the manufacturing anomalies. gcaptain.com summary
• Transportation Safety Board of Canada, final report on the Titan, June 2026.
• Arnie Weissmann, “Mission Titanic,” parts 1 to 3, Travel Weekly, 20 to 22 June 2023. The May 2023 mission from inside: the briefings, the go/no-go factors, the platform, the cigar conversation, Boeing’s statement, Nargeolet and Harding. Part 1, Part 2, Part 3
• Coverage of the September 2024 hearings: NPR, “Allegations of greed and recklessness in Titan submersible hearings,” 30 September 2024; CNN, 17 September 2024, on Lochridge’s testimony, the viewport and OSHA; NBC News, “Key revelations emerge in Coast Guard’s Titan submersible hearing”; the Associated Press on the final messages and Steven Ross’s account of dives 80, 81 and 87. npr.org, cnn.com, nbcnews.com
• NBC News, “Titan submersible disaster that killed 5 people was ‘preventable,’ Coast Guard report says,” 5 August 2025, for the report’s findings on Rush, the culture and the recommendations. nbcnews.com
• Guillermo Söhnlein’s testimony to the Marine Board of Investigation, September 2024, as reported by the Associated Press, on founding OceanGate with Rush and leaving in 2013. wtoc.com
• David Pogue, CBS News Sunday Morning and the Unsung Science podcast, 2022, the source of the “pure waste” remark and the account of the July 2022 expedition; CNN, “What we know about Stockton Rush,” 21 June 2023, for the 2021 interview on breaking rules; Smithsonian magazine, 2019 profile, for the Passenger Vessel Safety Act remark. cnn.com
• OceanGate, “Why Isn’t Titan Classed?”, company blog, 21 February 2019.
• The Marine Technology Society letter of March 2018 and its thirty-eight signatories, first reported by the New York Times, 20 June 2023, with Boeing’s, NASA’s and the University of Washington’s statements on their involvement.
• Christine Dawood, interview with the BBC, 26 June 2023, on giving up her seat and Suleman’s Rubik’s cube; and his aunt’s account to NBC News, June 2023, of why he went. nbcwashington.com
• Hamish Harding’s post of 17 June 2023, as quoted by Al Jazeera, Business Insider and others. aljazeera.com
• US Patent 11119071B1, “Systems and methods for curing, testing, validating, rating, and monitoring the integrity of composite structures,” Stockton Rush, assigned to OceanGate, 2021: the real-time monitoring system.