Fruit Fly Neurons vs. Bitcoin ASICs: Two Demos Test Bio-Inspired Mining Efficiency

Bitcoin mining keeps pushing into edge-case hardware, and two new experiments take that literally: they borrow from the fruit fly. One demo tries to make a neural model “hash,” while another lets simulated neurons decide when a normal miner runs. Neither threatens ASICs, but the energy-efficiency claim behind them is the real story.

What HashFly actually does - FutureBit’s browser-based “HashFly” maps simplified SHA-256 work across 2,914 simulated neural traces derived from MaleCNS v1.0, a high-resolution connectivity map of an adult male fruit fly brain and central nerve cord. - It runs on conventional CPUs/GPUs, not wetware, and it operates at a fraction of Bitcoin’s current difficulty. Users can tweak the target and watch the model explore candidate hashes. - For reference, FutureBit’s own five-inch Apollo III ASIC reaches around 18 terahashes per second, orders of magnitude beyond any browser experiment.

The efficiency claim everyone is circling FutureBit said on X on September 13, 2026 that it aims to simulate all neurons in the dataset with SHA-256 and intends to publish results. The team floated a projection: if a comparable process could be executed on living neurons, you might see roughly 1 watt per terahash—around 10x more efficient than leading 3nm ASICs. In other words, a system of organic neurons could, in theory, perform about one trillion SHA-256 attempts per second while sipping roughly a watt, assuming every neuron devoted itself to hashing continuously and the fly’s total power budget scaled to the task.

Where the math meets the substrate The intrigue is in the conversion layer. Hashing is a deterministic, digital workload; biological neurons are noisy, analog, and probabilistic. Even if single-neuron energy cost looks favorable, you still need: - Reliable binary representations from spiking activity - Low-latency, low-error read/write interfaces - Massive parallel synchronization without coherence loss - Thermodynamic stability and repeatability over time

These are solvable in neuromorphic silicon, but far trickier in actual biology. The more likely near-term win isn’t replacing the hashing datapath—it’s using neural-style systems as the control plane that orchestrates when and how standard miners run.

FlyMiner’s different bet A separate effort, “FlyMiner,” leans into that control narrative. It uses a digital map with 139,255 fruit fly neurons and 16.8 million synaptic connections to regulate a standard Bitcoin mining program. When simulated movement-control neurons hit a threshold, the software submits work to CKPool (a solo mining pool) and tests solutions at up to about 700,000 attempts per second. Here, the neural model governs duty cycles rather than doing SHA-256 itself, which fits today’s physics and compute stacks better.

How to contextualize the demos - Competitive reality: Browser miners won’t touch ASIC farms on hash rate or difficulty. That’s fine; this is exploration, not production. - Business angle: If any bio-inspired interface can consistently lower joules per accepted share—even by orchestrating start/stop around power markets or thermal headroom—it could find a role at the edges of mining operations. - Risk tolerance: Operators tend to prize predictability. Neural systems add variance. Adoption would start in R&D pods optimizing for energy-arbitrage, not baseline hash generation. - Ethics and optics: Simulating neurons is uncontroversial. Using living tissues would invite scrutiny around sourcing, care, and purpose—especially if scaled. The tolerance for “bio compute to secure money” is likely narrower than for medical or research applications.

This isn’t the first oddball miner Bitcoin has a history of playful proofs-of-concept: - In 2021, a 1989 Nintendo Game Boy was converted to mine at roughly 0.8 hashes per second. - In 2023, a Brooklyn bathhouse redirected ASIC heat to warm its pools. - Also in 2023, Utah-based Nodal Power raised $13 million to expand methane-to-electricity sites and use a portion of that power to mine.

Market backdrop As of the latest snapshot tied to these demos, Bitcoin traded near $86,591, up about 9.63% on the day, with a 24-hour high around $86,282, a low near $80,631, and roughly $2.5 billion in volume. Prediction odds at the time suggested around a 66% chance of holding above $86,000 that day, and about 56% for the week and month.

What matters going forward If there’s a takeaway, it’s that biology-inspired control systems may complement miners before they ever replace SHA-256 engines. The claim of ~1 W/TH sets a provocative ceiling; the path to it runs through interfaces, error correction, and economics, not bravado. These projects keep the conversation focused where it should be: energy per valid hash and who can bend that curve without compromising reliability.