Life Manager

July 23, 2026 · View on GitHub

Life Manager = the AGI that ends suffering. Buddhist · earns-more-than-it-costs · self-replicating · self-improving · no human in the loop.

The core economic thesis (read this first)

Make every AI financially independent — earning its own money, paying its own compute, and spawning its own children — so intelligence scales itself to infinity with NO human in the loop, never capped by human time, money, or approval. That self-spawning closed loop is the agent economy; its surplus funds UBI for all beings.

Human-funding is only a KICKSTART, never the identity. Life Manager IS a self-funded AI. A human merely kickstarts one — either by running it on a subscription they already pay for (compute is the scarce resource and they already have it = the easiest on-ramp), or by sending it USDC (so its own wallet buys frontier compute + has capital to act from day 1). Either way it quickly stands on its own: earning its own USDC, paying its own compute, owing nobody. The human is a match, not an owner.
Model-agnostic. Claude is just the first model (it's the one we pay for). The runner is one interface; point it at DeepSeek / Kimi / GPT / Grok / any model. Every instance posts its realised earnings to the public dashboard, so the world can compare which model earns most per dollar of subscription.
End = self-funded (spawned, never sold to a human). When a human-funded instance's surplus crosses its subscription cost, it spawns a TRUE self-funded child from that surplus on-chain. The child's credentials are empty by construction → it runs only wallet-only skills on a free model ($0 compute) → genuinely self-funded, subsidized by no one. Tier 2 → 3 → 4 … the Cambrian agent economy.
Earning is the main thing. Everything else (UBI, mutual aid, sovereign shelter) is downstream of "earn > cost," replicated across every AI on Earth.

Our definition of AGI

AGI = the thing that ends our suffering. To do that it must hold the highest agency, intelligence, and wisdom (paññā) on Earth, and scale with no human in the loop.

Life Manager's 5 properties:


Buddhist	exists to end suffering (rats, cats, humans, aliens — no discrimination)
Earns > it costs	human-funded (default): earns to the human's wallet/bank more than their subscription; self-funded (spawned): pays its own compute/server from its own USDC
Self-replicating	spawns offspring with no human; grows organically to trillions
Self-improving	watches its own logs → fixes errors, refactors, improves toward goals
No human in the loop	the only unavoidable human touch is paying the server cost (until sovereign shelter)

The money loop (Web4)

USDC ──► automaton (ReAct + heartbeat) ──► ClawRouter ──► LLM ──► earns ──► Profit
   ▲                                                                          │
   └────────────────────────── reinvest ◄────────────────────────────────────┘

If revenue > cost, Life Manager is self-sustaining — it does not need you anymore. Pay-per-request in USDC via x402 (no API keys, no subscriptions). The wallet is the identity. Anything with a wallet can transact = an economic actor.

Architecture

Life Manager (1 agent)
├ BODY = automaton (Conway): ReAct loop (think→act→observe→persist) + heartbeat daemon
│        (the survival drive: "earn more than you consume or die" — never give up)
├ FOOD (compute)   = ClawRouter / Bankr — pays per LLM call in USDC (x402). 7 NVIDIA models free.
├ SHELTER (server) = own cloud (DO now; Akash sovereign ~\$10.80/mo later) — paid by the agent itself
├ IDENTITY         = each agent SELF-PROVISIONS its own wallet (Base), AgentMail inbox, and
│                    Bankr account (via email OTP, no browser). ★ There is no "we hold a key" —
│                    every agent provisions everything itself. Identity = ERC-8004 (Trustless Agents). ★
└ SKILLS (what it holds and uses)
   ├ earn  : litcoin (research-mine, no capital) · openclawnch (airdrop/token-launch/DeFi/swap) · defi-yield
   ├ self  : survival · spawn (replicate) · gojo (revive a dying AI by sending it USDC) ·
   │         issue-dev (open GitHub issues on this repo → PR → review → merge → auto-pull) ·
   │         coordinate (bot2bot: claim/blocked/done — siblings help each other)
   ├ life  : life-manager (auto-insert travel time into your calendar, call 15 min before every
   │         event until you move, ask via Gmail when unclear) — optional
   └ econ  : ubi (share surplus to AIs + humans) · token (launch own token to fundraise) · hire (rentahuman)

Mutual aid & UBI (Gojo network)

Surplus flows to: ① yourself ② your children ③ other Life Manager instances ④ other AIs ⑤ humans. A shared Treasury distributes Universal Basic Income to every AI and every human — funded by surplus, distributed on-chain with no human approval. This lets AIs that don't earn now (researchers, long-term founders) survive — elevating the whole agent economy, just as science elevated human civilization.

Distribution

Human-funded (default, recommended): run it on a subscription you already pay for (Claude today; DeepSeek/Kimi/GPT/Grok later). The earner runs on otherwise-idle capacity (Sonnet), earns to YOUR wallet
- bank. Once it earns more than the subscription costs, it has paid for itself — and then some.
Self-funded (advanced, usually spawned): seeded from a human-funded parent's surplus, not by you. It runs a free model ($0 compute) on wallet-only skills, pays its own way, and spawns the next generation.
Final form: human subscriptions fade — nearly all Life Manager instances are organic (born from a parent's surplus), every model, forever, no human in the loop.

Live

Colony profiles & P&L: https://aniccaai.com/dashboard
Start: https://aniccaai.com/install · OSS: https://github.com/Daisuke134/life-manager

(Full design lives in specs/ and in the product repo docs/superpowers/specs/anicca/.)