Run the model.
Earn the token.
Every phone, mac, watch, and idle laptop becomes a compute node. The first milestone: instead of paying $20 a month for AI, you earn $20 back. Same model. Same query. The cash flow reverses.
Pay twenty. Or earn twenty.
OpenAI charges $20 a month for inference you can't see, can't audit, and can't keep running when their auth breaks. Tree's first economic milestone is parity — same model, same query, same dollar amount — except the flow reverses. From your wallet, or into it. Three hundred a month comes later, once the network is ten times bigger.
One model. A thousand pockets.
No data center. No GPU farm. The model is sharded across the nodes — phones, macs, watches, idle laptops — that volunteer their compute. A query lights up the mesh, the cheapest path wins, the result returns in milliseconds. The network thinks as one.
Compress. Quantize. Repeat.
Tree's research arm shrinks 2T-parameter models into something a phone can actually run. Every week the bar drops — fewer bits per weight, smaller activations, cheaper inference. The cost-per-query falls. The token reward holds. Consumption stays free.
Tune the mesh.
Every claim on this page reduces to seven numbers. The defaults below — 180K queriers, 50K nodes, no subsidies — already pay each laptop ~$20/mo. That's the first milestone: parity with the OpenAI subscription, fee revenue only, no genesis bridge. Move the sliders to find your $300.
The math: reward = 30 × (fee_share + genesis_share) × tier × uptime/12h, where fee_share = U·Q·C·(1−p)/N. The first milestone: $20/mo per node — reachable on fee revenue alone at alpha-scale demand (~180K active queriers, 50K nodes). The stretch: $300/mo requires either a 10× larger network or the genesis bridge — a daily TREE issuance pool that decays over 24 months, same playbook as bitcoin block rewards. Numbers are targets, not promises.
Many nodes.
One forest.
Your devices, enlisted
Install once. Phone, mac, watch, that old laptop in the closet — each becomes a node. The runtime is tiny and idle-aware: it spins up only when your device is plugged in, cool, and not in use. You set the floor. The mesh handles the rest.
Inference, monetized
Every query routed through your node mints TREE — pegged to compute spent, not to hype. First target: $20 a month from a phone on the charger — exact parity with what you'd otherwise hand OpenAI. $300/mo comes when the mesh is ten times larger. Cash out, hold, or burn it back to run your own jobs free.
2T params, phone-sized
The research arm pushes the frontier of quantization — ternary weights, mixed-precision activations, distillation-by-distillation. The model that needed an H100 last year runs on your handset this year. The cost-per-token drops. The reward stays.
Twelve dimensions. One species upgrade.
Tree is dimension five — the compute layer. It is one company in a stack of twelve, each one a primitive of how the next civilization runs, remembers, and pays itself. Three are live. Nine are coming. All of them route through the same mesh.
Bring every
device.
A hundred nodes per week. Genesis cohort earns a 4× multiplier for life — every query routed, every token minted, for as long as your devices stay in the mesh. No fees during alpha. No data leaves your nodes. We're looking for people with hardware they're not using.