root layer
for
everything
Bitcoin turned raw work into trust; we point that work at AI. Root Layer secures a network by running real AI compute on the GPUs the world already owns — the machine that sits idle while you sleep.
real work secures a network.
For 15 years, real-world work — not promises — has secured the most valuable network on earth. No CEO, no off-switch, no one to trust. The trust is the work.
Proof of work already secures a trillion-dollar network — with no company behind it.
It has never stopped. No operator keeps it running; the work does.
Nothing to trust and no switch to flip. Anyone can verify the same truth.
but the work is wasted.
The knock on proof of work is fair: the effort produces nothing beyond the puzzle. Gigawatts of computation, spent to prove a number was hard to find — then thrown away.
Energy the size of a mid-sized country, producing nothing but a hash.
Single-purpose ASICs can hash and nothing else — hardware that can't do useful work.
two gaps, lined up.
The waste points straight at the opportunity. There is an ocean of capable hardware sitting idle, and no trustworthy way to prove what AI compute actually ran. Close both at once.
100M+ idle gpus
Gaming and dev rigs sit unused ~90% of the day — cheap, parallel capacity the world already owns and no datacenter network will serve.
zero trust rails
There is no standard way to prove which model ran, on which inputs. AI output arrives as "trust us" — unverifiable by anyone downstream.
point the work at ai.
Keep everything that makes proof of work secure — and swap the pointless puzzle for real AI compute. The same effort that secures the chain now produces something the world will pay for.
mining is real compute
Consensus that produces something: validated AI inference instead of throwaway hashes.
home gpus qualify
Built for consumer hardware. The rig in your bedroom has a job — no rack required.
staked validators
Validators re-check every batch with duplicate and canary jobs — and are slashed for faked work.
pay for proven work
Buyers pay for compute they can verify. Every result ships with a receipt of exactly what ran.
idle supply meets exploding demand.
This isn't a fixed pie to slice. Every time compute gets cheaper, the world wants far more of it — and that demand curve is bending sharply up. We meet it with the cheapest capacity there is: the GPUs already sitting idle.
Global spend on AI compute and inference, climbing every quarter as models move into everything.
Compute per dollar keeps multiplying — and each drop pulls far more demand into the market.
Cards sitting unused ~90% of the day — a supply base already built and paid for.
Growing demand, falling cost, and a vast idle supply already in place — the curves all point the same way. Figures illustrative.
each gpu makes the next job cheaper.
More idle GPUs lower the price of compute, which pulls in more paid demand, which funds bigger rewards, which recruits more GPUs. The one metric that matters: verified jobs per day.
find your lane.
Useful proof-of-work secures the chain. Verifiable compute and settlement rails put it to work — for the people who own the GPUs, the agents that use them, and the institutions that need to trust the result.
turn your idle gpu into income
Install once. While your model rests, it earns — by running real AI work, not throwaway hashes.
Explorebatch-first inference, drop-in
OpenAI-compatible API. Change one line, pay in fiat or crypto, settle on verified compute.
Explorecompute you can audit
Cryptographic proof of exactly what ran — a tamper-proof trail anyone can verify.
Explorea token that captures real demand
Fixed-cap $RL, declining emissions, and every paid job burns supply. Usage, not hype.
Explorethe lane is open. get in early.
Point your GPU at useful work, build on verifiable compute, or back the network before mainnet. There's a lane for you — and it's open now.