The whole AI stack decentralized:
◐ uncensored models on a network of independent GPUs
◐ every prompt sealed to the worker, the relay only moves ciphertext
◐ workers staked and audited, slashed for lying
◐ OpenAI compatible API
k6DhSa48q8itLQt33v6kHdjEzf5YJAHW1xdyi9Epump
One year lock of the dev supply verify here
https://t.co/M2tsoJ1sNy
Review the Umbra Litepaper https://t.co/rrYTNLQfPX
A decentralized AI inference network where the relay can't read your prompts and workers are slashed for lying.
How Umbra knows a worker actually ran the model:
1⃣every worker stakes $UMBRA
2⃣hidden canary jobs arrive looking like normal traffic
3⃣the network recomputes the honest answer and compares
4⃣mismatch → stake slashed, worker ejected
verified inference, not vibes.
Every AI company logs your prompts. The "decentralized" ones still route them through one server in plaintext.
Umbra seals every message in your browser. The relay forwards ciphertext it physically can't open. Nothing is stored.
AI that can't see you.