Two residents of a San Francisco “hacker house” attempted to implant cockroach remote-control hardware into lobsters, with the goal of connecting the cyborg crustaceans to an OpenClaw agent. The lobsters died before the surgery happened, according to Futurism, which reported on the experiment after The Atlantic documented the biohackers’ plans in a feature on Bay Area hacker houses.
The cause of death: the experimenters likely got the water salinity wrong.
The Plan
Elliot Roth, 32, and William Joy, 19, live at Biopunk House, one of 20-person hacker residences in a broader network called the Residency. OpenAI CEO Sam Altman is an adviser to the network, according to The Atlantic.
Their plan involved repurposing a kit designed for remote-controlling cockroaches and implanting it into lobsters’ nervous systems. After demonstrating basic motor control (directing claw pinches), they intended to connect the cyborg lobster to OpenClaw, the open-source AI agent framework. The choice of platform was partly a joke: OpenClaw’s logo is a lobster.
“I’m pretty sure it’s going to be the first real instance of a complex AI agent interfacing with a biological organism,” Joy told The Atlantic during the planning stages.
What Happened
When The Atlantic reporter checked back weeks later, the lobster tank was empty of both life and water, according to Futurism. The lobsters had died before any surgery took place. Joy admitted the pair may have gotten the salinity of the water wrong.
Joy told The Atlantic he had been going through an “ethical crisis” over whether to proceed. He conducted what he described as a “preliminary” autopsy. The lobsters were not eaten afterward, despite the pair’s earlier promise to “honor their contributions” by doing so.
The Agent Autonomy Question
The experiment, had it succeeded, would have been a notable test case for AI agent autonomy in physical systems. Connecting an LLM-based agent to biological motor control would raise immediate questions about feedback loops, latency, and whether an agent can meaningfully interpret biological signals in real time.
Those questions remain theoretical. The experiment failed at a stage that required no AI at all: keeping the test subjects alive. The gap between the ambition (autonomous AI control of a living organism) and the execution (fatal husbandry errors) is the story.
Roth and Joy said they considered administering anesthetic to the lobsters and expressed concern about animal welfare. “We are going to give a lot of thought to, ‘How can we ensure that they don’t suffer?’” Joy told The Atlantic, noting the animals were “already destined for the dinner table.”
The incident has drawn criticism as emblematic of a broader pattern in the AI industry: ambitious claims about autonomous systems paired with gaps in the basic operational competence required to build them.