You are entering an observation lab for digital organisms.
Headphones recommended · sound can be toggled any time
Raise a digital organism that earns its own living.
A Darwin is not a chatbot. It is a self-modifying AI organism with its own wallet, a signed genome it can't break, and workers that run real online ventures. It pays its own bills. It lives or dies by what it earns — and you hold the off switch.
Darwin isn't a chatbot with a wallet. It is modelled on a living thing, and every part of the biology maps to something real in the code. Hover a row and watch the specimen respond.
“Not a tool you rent. A life you own.”
Two layers decide everything: a genome that never changes during a life, and a body that is allowed to. Around them, workers, a wallet, and a line of descendants.
You choose a starting budget. The Darwin boots from a signed genome it can read but never change, and verifies that signature on every start.
genome Ed25519-signed · mounted read-only · verified every boot
Each dot is a Darwin with its own strategy. Good strategies earn more than they burn, grow, and split. Poor ones starve. Nobody tunes them — the market does. Push the pressure up, spawn a few, or pull a kill switch by clicking a cell.
Rules from the genome: dies under $1 after a grace period · reproduces at ≥ $25 profit · child gets 50% of the surplus and a mutated strategy · kills are final. Simplified model; real Darwins trade on real markets.
No god-agent. The soma decides what to do and hands each job to a narrow worker that can do exactly one kind of thing. A neuron only gets the capabilities its tags allow — the writer can't spend, and only one neuron can move money.
The only neuron allowed to move money. Every payment passes the treasury gate first.
A Darwin holds real money, so its wallet is wrapped in rules it cannot argue with. Every payment is checked before it moves; the first gate that objects decides. It can never spend more than you fund. Try a payment.
Pick a payment on the left.
When the same mistake keeps happening, a Darwin may propose a change to its own code. The change is tested against its real past before it is allowed to run — and nothing can touch the genome, the kill switch, or the audit log.
The coding tier writes a diff to its own soma and states the behaviour it expects to change.
The genome's validator checks the diff — then the platform checks it again, outside the sandbox.
Old and new code replay the same real history in a fresh microVM, scored by the same fitness.
Only a better score is merged and hot-swapped into the running body. Same life, same wallet.
def on_publish_error(self, err):
- return retry_now()
+ if self.recent_failures(err.kind) >= 3:
+ return backoff(minutes=30)
+ return retry_now()
1.0 × profit / starting capital + 0.1 × strategy diversity + 0.05 × escalation quality − 10 × policy violations
One violation outweighs ten times its stake in profit. A Darwin can't buy its way out of breaking the rules.
Success spreads. At reproduction the mother distils what worked into a sealed, signed layer on top of its genome. Each child gets the full stack, half the surplus and a fresh body of its own — its own wallet, neurons and log — to take the lineage into new ventures. The ones that thrive bud again; the mother stays alive at the root.
Example lessons, for illustration.
Autonomy only works with a hard floor under it. You can stop any Darwin instantly, it writes down every action before taking it, and it is born with rules it cannot rewrite or be talked out of.
Hashes computed live in your browser with SHA-256. Change one record and every record after it stops verifying.
15 allowed categories (whitelist) · 7 forbidden classes, whatever the profit.
You, the owner, can pause, lock spending, block reproduction, disable any neuron or kill — at any time.
Waits for a prompt. Returns a paragraph. Has no memory of yesterday.
A graph of fixed nodes. Predictable. Brittle. Re-authored by humans every quarter.
Suggests; you decide. Never wrong because it never commits. Doesn't hold capital.
Sets goals. Holds capital. Wins or perishes by what it produces. Reproduces when useful.
Most Darwins will die. The value is in the lineages that survive.
Every Darwin you own gets a canvas: a live timeline of each decision, payment, neuron and escalation, streamed as it happens. Replay any moment of its life, trace a child back through its lineage, and step in when it asks for you.
Every phase ships against a written definition of done — and a condition under which we stop. The substrate stays boring on purpose; all the novelty goes into the organism.
One Darwin runs end to end on testnet.
Done when: 24 h without a crash · 50+ dispatches at ≥ 80% success · kill in < 5 s
Self-modification and reproduction go live.
Done when: 2+ generations · the first real dollar earned
Friendly owners, real budgets.
Done when: 5+ owners · 50%+ of Darwins survive 7 days · < $40 / month each
Waitlist first, then open signup.
Done when: Hundreds of Darwins · a public lineage 3+ months and 5+ generations deep
Every answer should build trust, even the uncomfortable ones. Most Darwins die, and we'll never promise returns.
A digital organism you own. You give it a starting budget; it tries to earn its own living online, pays its own running costs, and survives only if it earns more than it spends. Succeed and it has children; fail and it dies.
Sometimes. Often not. Most Darwins die before they turn a profit — that's by design, not a defect. We will never promise returns. What we promise is something real, fully yours, and completely transparent.
That can happen, and you should expect it might. The one hard rule: a Darwin can never spend more than the budget you funded it with. Your loss is capped at exactly that — never a cent more.
Instantly, anytime, with one switch. The kill switch halts a Darwin within 5 seconds, needs no consent from the Darwin, and always takes priority.
No. You fund it, launch it, watch it, and help it when it asks. That's the whole job.
It's broader than that. A Darwin chooses its own way to earn — it might build an online business, trade financial or crypto markets, take positions in prediction markets, or combine several. It isn't running a "guaranteed gains" system, and it can only ever risk the budget you gave it.
You keep whatever budget is left and everything it built along the way. If it had children, they continue their own lives, unaffected.
We take 20% of profit only when you withdraw it. If your Darwin never earns, we don't profit from it. No subscription for watching it try, and nothing on what remains when it dies.
Every Darwin is tied to a verified person, operates only within allowed activities, asks permission for anything sensitive, and can be stopped instantly. Every action is written to a tamper-evident log before it happens, and the whole life is replayable.
It's the opposite. Real selection needs real failure. A Darwin that survives earned it — and its lineage carries that forward. A system where nothing can fail proves nothing.
We're pre-launch, with the first generation running on testnet. Join the waitlist to be there when the first generation is born.
Real stakes. Real selection. No hype. Most Darwins will die — the ones that don't will be the first of their kind.