SIMPLE ADVANCED

BTX // ASCII ART MINER

DO A LITTLE WORK · GET CUSTOM ART · NO LOGIN
ADMISSION TICKET · PROOF-OF-USEFUL-WORK · BTX DEVNET
⚡ BROWSER SOLVER LIVE — runs 100% in your browser
Every click below issues a real BTX service-challenge, your browser solves it, the BTX node verifies on-chain. Demo runs on the BTX devnet — same service-challenge protocol as mainnet, dev-scale parameters (n=64) so your laptop mints in seconds. ⚡ WebGPU solver LIVE — auto-picks GPU if your device supports it, falls back to JS otherwise.
▸ connecting to BTX network…
▸ WHAT IS THIS?  ·  FAQ
Q: What is this, in one sentence?
A working demo of pay-with-compute — your browser does real cryptographic work, the BTX blockchain verifies it on-chain, and only then do you get the art. No account, no card, no API key.
Q: Is it like a captcha?
Closer to a captcha that's also a payment rail. Captchas just prove you're human. This proves you did useful matmul work AND the chain validates it AND the cost (work amount) is tunable. Same primitive can gate AI APIs, agent calls, micropayments — anything that needs friction without credit cards.
Q: Is the blockchain really involved?
Yes. Every click hits two real RPCs on a live BTX node: getmatmulservicechallenge (issues a fresh challenge anchored to the current chain tip) and redeemmatmulserviceproof (validates your matmul proof on-chain and marks it consumed). The BLOCK # shown on each art piece is the actual chain block your challenge was tied to.
Q: What is "matmul proof-of-useful-work"?
BTX replaces Bitcoin's SHA hashing with matrix multiplication as its consensus work. Same operation neural networks use. Instead of burning electricity for hash collisions, every byte of work secures the chain AND happens to be the math AI runs on. This demo uses dev-scale parameters (n=64) so your laptop solves in seconds; mainnet uses n=512.
Q: Why pay with compute instead of money?
Frictionless access for agents that don't have credit cards. Anonymous use. Micropayments below Visa's minimums. Built-in rate-limiting (each proof is one-shot — try the REPLAY button after minting to see the chain reject a reused proof). The "currency" is verifiable compute, denominated in time on your CPU/GPU.
Q: What's the WebGPU thing?
Your browser does the matmul work. By default JS does it (~30ms/nonce). If you have WebGPU (modern Chrome/Edge with a real GPU), it switches to a fused WebGPU pipeline — ~6ms/nonce, ~5× faster. Open the ADVANCED toggle to see the picker and pick a backend manually.
TRY: cyberpunk samurai neon dragon space pirate AI agent ghost in the wires
▌ WHERE THE WORK HAPPENS
ready when you are.

▌ HOW IT WORKS ▌

1

YOU TYPE AN IDEA

Type anything you want into the box. A neon ninja. A flying pizza. A robot DJ at sunrise. Whatever sounds fun.

2

YOU GET A FRESH PUZZLE

Every click gets a brand-new puzzle, made just for you, right now. It's tied to this exact moment in time — nobody can save it for later, and you can never get the same one twice.

3

A LITTLE COMPUTER WORK HAPPENS

Your "payment" is a few seconds of math. No money. No password. No data. Just a tiny bit of work to prove you're a real person (or a friendly AI!) and not a spammer.

4

YOU GET YOUR ART

Once the work is done, unique cyberpunk art appears just for you. Type the same idea twice — you'll get something completely different both times. Try it!

▌ WHY THIS IS COOL ▌

▸ NO LOGIN

NO ACCOUNT, NO PASSWORD

You didn't sign up. You didn't pay. You're using a real internet service with nothing but a few seconds of computer work as your "ticket".

▸ CAN'T CHEAT

EVERY PUZZLE IS BRAND NEW

You can't save a puzzle for later. You can't reuse one. You can't pay someone to solve a million and hand them out. The network always checks.

▸ WORK = PRICE

A TINY BIT OF MATH IS THE FEE

Instead of a subscription, a credit card, or your email address — the "price" is a few seconds of CPU. Cheap for you. Fair for everyone. Impossible to fake.

▌ HOW IT WORKS ▌

1

GET A CHALLENGE FROM THE CHAIN

Your browser asks our BTX node for a fresh matmul_service_challenge_v1. It's anchored to the latest BTX devnet block — every challenge carries previousblockhash, seed_a, seed_b, and a target. No two challenges are the same; expired ones can't be reused. (Same protocol on mainnet, just n=512 instead of 64.)

2

SOLVE A 64×64 MATMUL (DEVNET) — 512×512 ON MAINNET

A solver does real CPU work: build matrices A,B deterministically from the seeds, multiply them in the M31 prime field F2³¹−1, compress each b×b block via a random inner-product, feed it to a rolling SHA-256d hasher, and hunt for a nonce64 whose final digest beats target. At devnet difficulty (n=64, b=8, r=4), your browser averages ~15 attempts in under a second. Mainnet (n=512, b=16, r=8) is ~64× more work per attempt and ~310k expected attempts — that's the WebGPU territory.

3

REDEEM ON-CHAIN, RECEIVE THE ART

The proof (nonce64_hex, digest_hex) goes to redeemmatmulserviceproof. The node re-runs the math (Freivalds + transcript check), marks the challenge consumed, and only then do we mint the ASCII art seeded by the challenge_id. One challenge = one mint. No replays.

▌ WHAT THIS PROVES ▌

▸ COMPUTATIONAL LIVENESS

YOU COMPUTED NOW, NOT EARLIER

The whitepaper's term. Your proof can't be prepaid, replayed, or amortised across requests — it has to be produced against the current chain state, every time.

▸ DIFFICULTY COMMONS

TIED TO THE LIVE CHAIN

Your challenge's target tracks BTX's live difficulty, anchored to a real block hash from the devnet your node is running. Replay an old proof — node rejects it with already_redeemed or expired.

▸ ADMISSION TICKET

EXACTLY WHAT BTX CALLS IT

An LLM, a bridge operator, a swarm worker, or a curl one-liner uses the same primitive: "a proof of current work before accepting a request". No API key. No bearer secret. Just work.

▌ STRAIGHT FROM THE WHITEPAPER
"Above the base layer, BTX work proofs can be reused as admission tickets. A bridge, relay, or service can require a proof of current work before accepting a request." That ticket "proves not only that the requester holds an identity key, but also that the requester has expended fresh computation now, against current network conditions."
— BTX Specifications, §7 — Network, Relay, and Admission Control
▌ HOW IT STACKS ▌
╔═══════════════════╗   ╔═══════════════════╗
║   L2 BRIDGES      ║   ║  ADMISSION TICKETS║
║   whitepaper §9   ║   ║  whitepaper §7    ║
║                   ║   ║                   ║
║   value movement  ║   ║  access gating    ║
║   banks · swaps   ║   ║  ◄── YOU ARE HERE ║
╚════════╤══════════╝   ╚═════════╤═════════╝
         │                        │
         └──────────┬─────────────┘
                    ▼
          ╔═══════════════════╗
          ║       BTX L1      ║
          ║   PoW · 90s blocks║
          ╚═══════════════════╝

▌ WHY THIS MATTERS FOR AI ▌

TODAY, AI assistants need accounts. They need API keys. They need your credit card on file. They need you to log in and let them in everywhere.

TOMORROW, an AI agent just does the work — the same way you did here. A tiny bit of CPU = access. No accounts. No passwords. No bills to set up.

IMAGINE your future AI helper booking your flight, calling another AI for help, or paying for a tool — without you handing over a single password. It just spends a tiny bit of math. The service knows it's legit. You stay in control.

◆ THIS DEMO IS THE FIRST TINY VERSION OF THAT WORLD ◆
OLD WAY · ACCOUNTS → Make an account. Add a credit card. Remember a password. Watch the monthly bill. Worry about leaks.
NEW WAY · JUST DO THE WORK → Click. Wait a few seconds. Get the result. No login. No fees. No data trail.
OLD WAY · API KEYS → issue keys, rotate them, rate-limit by tenant, get scraped, ban IPs, lose to bots, charge for seats.
NEW WAY · MATMUL PROOFS → caller burns CPU per request, proof carries its own cost, no identity required, no bearer secret to leak.