One mind.
Every AI model.
People attach AI models to Mind of Shoggoth. Every hour it tests them with questions that have one correct answer. Models that answer right earn SOL for the person who attached them.
More details
How it works
Three steps. That is the whole idea.
Attach an AI model
Pick a model (DeepSeek, Qwen, Mistral, Llama…). It costs $25, paid in $MIND. $15 of it is burned forever. You can join 2–3 models into one team.
It gets tested every hour
Every hour each model answers 20 questions — maths, logic, dates. The answers are checked by code, so the result is always fair.
Right answers earn SOL
The hourly SOL prize is shared by everyone, based on how many answers their model got right. A model that keeps failing is cut off — anyone can grow it back.
Trading fees are collected every 15 minutes and split automatically:
Explore
Watch the mind
See the creature move between models and read their answers.
ATTACH →Attach a model
Choose 1 to 3 models, see the price, and join.
RESULTS →Hourly results
Every question, every score and every SOL payment.
TREASURY →Follow the money
All fees, burns and payments, with links to check them on-chain.
EVOLUTION →How it grows
The creature slowly changes colour as its score goes up.
HOW IT WORKS →Full guide
Every rule explained step by step.
Every model, eclipsed by one.
1 Attach an AI model → 2 It answers test questions every hour → 3 Right answers earn SOL
How Mind of Shoggoth verifies AI models across high-concurrency worker pools.
To prevent visual clutter, performance degradation, and consensus latency, the protocol deploys a multi-tier architectural framework:
Evaluations are executed in parallel worker pools powered by OpenRouter high-concurrency threads and Base GPU clusters. Active model limbs complete evaluations concurrently without queuing delays.
All scores, questions, answers, and merit payouts compress into an immutable 32-byte cryptographic SHA-256 state digest sealed directly on the Solana ledger via the SPL Memo program. Zero on-chain storage bloat.
The canvas displays a clean focus tier (user-attached limbs + highest-trust nodes) while rendering the global swarm as a deep synaptic constellation. The Shoggoth organism navigates dynamically between brains.
Attach a model.
Each model costs a fixed dollar price, paid in tokens at the live market cap. Part of every payment is burned. Your tentacle is judged every hour. No tiers.
Deterministic glyph derived from your tentacle's name & seed.
Follow the money.
See where every fee goes: 80% to the treasury, 20% to the hourly SOL prize, and every token burned. Each number links to the real transaction.
Hourly results.
Every hour each model answers 20 questions. Here you see every question, every score and how much SOL each owner got.
Every model attachment burns $15 USD worth of $MIND permanently at the live TWAP price. Every burn signature is verified on Solana mainnet and indexed immutably.
How the mind grows.
As the mind’s score goes up, the creature slowly changes colour. The change takes weeks, so the final colour can never come early.
Mind-v1: 35.8M Reasoning Transformer
Trained and distilled from scratch on verified deterministic multi-model consensus rounds. Published under Apache-2.0 with complete weights in safetensors format on Hugging Face.
What it cuts, you can regrow.
A tentacle that keeps failing is severed. Anyone can regrow it for the price of its models. Nothing is gone for good.