The coelacanth, and what it has to do with us
The coelacanth was thought extinct for sixty-five million years.
In December 1938, a trawler working off the mouth of the Chalumna River in South Africa hauled in a five-foot, hundred-and-twenty-seven-pound fish with limb-like fins and steel-blue scales the captain had never seen before. He set it aside for the curator of the East London Museum, a young woman named Marjorie Courtenay-Latimer, who took one look at the specimen on the dock and recognised that she was looking at something the fossil record said had been gone since the late Cretaceous. She sketched it, preserved what she could in the Christmas heat, and wrote to the ichthyologist J. L. B. Smith. He named the genus Latimeria in her honour. A second living species, Latimeria menadoensis, was caught off Sulawesi in 1997. Both still live, in deep water, mostly unchanged from their fossil ancestors.
The discovery rearranged the textbooks. Coelacanths sit close to the lineage that crawled out of the water and became tetrapods — us, eventually. Their fins are lobed, with internal bones arranged in a pattern that looks unmistakably like an early limb. They were not the missing link. They were a quieter thing: a working organism that the science of the day had written off, swimming the whole time at six hundred metres, where almost nobody was looking.
We named a company Candlefish, and we have written before about the candlefish itself — Thaleichthys pacificus, the Pacific smelt that you can dry and burn. We are writing now about a different fish in the same lineage of attention, because the coelacanth keeps showing up in the studio's visual language and we want to say plainly why.
The coelacanth is not a metaphor for the company. It is a metaphor for the work.
A whole register of craft has been declared dead in the last two years. Careful integrations. Calibration against a specific operator's actual workflow. Reading a codebase before changing it. Sitting with a client long enough to see the second-order effect. Writing software that survives a staffing change. The discourse has moved on. The conferences are about something else. The trade press has filed the obituary.
The work, however, kept running. It is mostly being done by small teams in unfashionable industries — cabinet shops, lumber yards, embroidery franchises, regional construction companies — and by the operators inside those businesses who still know what good looks like and have not been talked out of it. None of this work shows up in a leaderboard. None of it is on stage. It runs at depth, the way coelacanths run, and it has been running the whole time the surface has been arguing about whether it exists.
Our bet is that the next generation of useful AI gets built by the people who never stopped doing this kind of work. Not by retrofitting hype onto craft, and not by pretending the craft is new. By recognising that the work was always there, in deeper water, and that the rediscovery — when it comes — will look less like an invention and more like Marjorie Courtenay-Latimer at a dock in 1938, sketching a fish nobody else had patience to look at.
One thing to do with this. Next time you read a piece of software-industry writing that announces the end of something — the end of careful code review, the end of human-on-the-loop, the end of small consultancies, the end of the operator who knows the spreadsheet by heart — go and look in the deeper water. Find the team that is still doing the thing. Read what they ship. The fossil record is wrong more often than the discourse will tell you.
The animal that was not supposed to exist is the one with the most to teach about how anything survives at all.