News/Business/Coinbase Engineer’s Simulated Fly Brain Trades…
Business

Coinbase Engineer’s Simulated Fly Brain Trades $100 in Crypto, Down About $1

A Coinbase engineer wired a simulated fly-brain connectome into live trading; the $100 BTC, ETH, SOL and USDC stake is down roughly $1.

Coinbase Engineer’s Simulated Fly Brain Trades $100 in Crypto, Down About $1

A software engineer at Coinbase has connected a computer simulation of a fruit fly’s brain to the exchange’s own AI agent tools, letting the artificial neurons decide when to buy, sell, or hold Bitcoin, Ether, Solana, and USDC. After roughly a day of live trading, the portfolio that started with $100 was down close to $1, a loss of about one percent.

The current split, according to a post from the engineer, Alex Wormuth, shows $59 in USDC, about $5 in BTC, $17 in ETH, and $17 in SOL. That totals roughly $98, consistent with the reported single-digit-dollar drawdown.

What Two Outlets Actually Confirm

Both Protos and BeInCrypto reported on the project this week. Protos published the underlying detail: Wormuth’s own tweet describing the setup, the starting balance of $100, the asset mix, and the mechanism, dopamine-style neurons that fire when a trade is profitable, feeding signals that trigger buy or sell orders through Coinbase.

BeInCrypto’s summary line adds one further claim: that the engineer himself cautions the experiment “proves nothing.” That framing did not appear in the retrievable portion of BeInCrypto’s report, so it is worth treating as attributed to that outlet rather than independently verified. The core mechanics of the bot, the starting capital, and the near-flat performance are the parts that hold up across sources.

The Neuron Map Behind the Bot

The simulation is only possible because of a connectome, a full wiring diagram of a brain, released last week by Google and the Howard Hughes Medical Institute’s Janelia Research Campus. Researchers mapped an adult male fruit fly’s brain and ventral nerve cord down to 166,000 neurons and 125 million synaptic connections, a project that took 20 years to complete and is described as the largest brain map by neuron count produced so far.

The stated purpose of that research is medical, helping scientists study how brains process complex behavior and eventually inform work on conditions such as Alzheimer’s, dementia, and schizophrenia. A crypto trading bot was not the intended application. It is a side effect of the data being made public.

What Remains Single-Sourced

Protos also reported that other developers have taken the same fly connectome and pointed it at unrelated tasks, including one account that claims to have taught the simulated brain to parallel park a virtual Mini Cooper. That detail did not appear in the second source and should be read as a single-outlet claim rather than a confirmed fact.

None of this changes anything about BTC, ETH, or SOL price action. It is a novelty built on top of a legitimate, dated research release, running on a live exchange account with a small, disclosed stake. The number that matters here is the one both sources agree on: a $100 starting balance, four assets, and a loss of about $1 so far.

Sources

Related articles