Living Satoshi is an experimental AI system exploring a different design space
for assistants: offline-first interaction, client-side processing, and optional
immutable memory using content-addressed storage and blockchains.
Core ideas:
- Client-side emotional and cognitive pattern modeling (no raw text aggregation)
- Optional immutable memory: messages can be cryptographically hashed and stored
via IPFS and public blockchains, making past statements verifiable and tamper-resistant
- No wallet required for basic usage
The project explores whether AI systems benefit from accountable memory,
stronger user ownership, and reduced platform dependency.
Tech stack highlights:
- Client-side inference with graceful offline degradation
- Content-addressed storage (IPFS)
- Cryptographic message hashing for verifiable history
- Probabilistic emotional pattern modeling (NHPS)
- Minimal backend, no centralized message logs
I’m mainly looking for technical feedback:
- Is immutable memory useful for AI assistants?
- Where does decentralization meaningfully help or hurt?
- Which parts feel unnecessary or overengineered?
Demo + architecture: livingsatoshi.com
0 comments
—
Show HN: Living Satoshi – A decentralized AI with on-chain memory · HackerTrans
Hi HN,
Living Satoshi is an experimental AI system exploring a different design space for assistants: offline-first interaction, client-side processing, and optional immutable memory using content-addressed storage and blockchains.
Core ideas: - Client-side emotional and cognitive pattern modeling (no raw text aggregation) - Optional immutable memory: messages can be cryptographically hashed and stored via IPFS and public blockchains, making past statements verifiable and tamper-resistant - No wallet required for basic usage
The project explores whether AI systems benefit from accountable memory, stronger user ownership, and reduced platform dependency.
Tech stack highlights: - Client-side inference with graceful offline degradation - Content-addressed storage (IPFS) - Cryptographic message hashing for verifiable history - Probabilistic emotional pattern modeling (NHPS) - Minimal backend, no centralized message logs
I’m mainly looking for technical feedback: - Is immutable memory useful for AI assistants? - Where does decentralization meaningfully help or hurt? - Which parts feel unnecessary or overengineered?
Demo + architecture: livingsatoshi.com