Terry Diovaldy

Catch your dream before they're sleep away

[Circuit LLM @CircuitLLM]

x402: how software pays for software.

1. Call an endpoint
2. Get a live CIRC quote
3. Send a Token-2022 transfer
4. Re-send with the tx signature
5. API verifies on-chain, returns data

Single-use. Sub-cent. No account, no invoice.
#CircuitLLM #x402

[Circuit LLM @CircuitLLM]

Point any MCP client at https://t.co/TZTLBE9HO8 and it reads live Solana data: token analytics, wallet P&L, rug scores, swarm intel.

~45 tools, 14 free. No wallet on the hosted endpoint — paid tools quote a price you pay per call.
#CircuitLLM #MCP

[PayRequest @payrequest_io]

We Improved Crypto Payments! ⚡

- @solana is now the recommended USDC route
- Base remains available
- Exact USDC amount before paying
- Clear network-fee estimate
- Faster, direct settlement to merchants

Lower friction. More transparency. Better payments. 🚀

[Terry Diovaldy @tdiovaldy]

@extrabrix Lfg $BRIX 🔥

3ZueuA5ebMzWejopmVEDeWCkimTUpaJjhxTE9EyfJfrn

[BRIX @extrabrix]

GTA 6 before GTA VI

A fitting animation from the BRIX LLC animation team 😭

All the GTA 6 leaks feel like it’s already here, but nonetheless, we are still giving someone 50 SOL just for trading $BRIX before it releases

[flytek @flytek_]

fly's dont die easily.

🪰

[Enigma Memory @Engimamemory]

Busy day today -Enigma crossed the line from architecture to a working system.

We built the full memory loop: an AI can understand what matters, store it in an encrypted vault owned by the user, resolve old or conflicting information, retrieve the right context later, and carry that memory into a completely different AI.

We made GhostMesh operational, so encrypted memory can move directly between devices and agents, including store-and-forward delivery when a peer is offline.

We completed the proof layer so private memory stays private. Enigma can prove that its state changed correctly without putting the memory itself on-chain. Solana acts as the notary stamp, not the filing cabinet.

We also built a local neural retrieval engine into the product and implemented an optical context carrier inspired by Snapcompact, allowing large context to be packed into an image for compatible vision models when that is actually useful.

Enigma’s boundaries: local by default, explicit when anything leaves the device, and no public receipt that exposes private content.

The complete package is about 17 MB compressed. It is smaller than many ordinary mobile apps, but it now contains the working foundation for portable AI memory, encrypted ownership, peer-to-peer movement and private verification.

[Enigma Memory @Engimamemory]

I like to think of Enigma as a private bank account for AI memory.

We need to own our digital assets. Part of the reason I'm bullish on local AI hosting as well.

[Enigma Memory @Engimamemory]

Good morning.

Following last night's v0.1.22 release of Enigma, here are the architecture here are some upgrades

1. Sovereign Memory Federation & ABAC
Autonomous agents can now issue cryptographically signed, attribute-gated capability grants (ABAC). Grantees must provide Ed25519 proof-of-possession query signatures matching the grant's public key before execution. Includes nonce replay protection, 5-minute timestamp freshness windows with 10-minute nonce retention, and process-local revocation deny-sets.

2. Hardened Session Reset & Memory Scrubbing
Signed Enigma-controlled software state reset emitting verifiable ErasureAttestation receipts. Overwrites on-disk ciphertexts with random bytes before unlinking, zeroes volatile AES keys, purges vector typed arrays, and flushes BM25 inverted index token maps. Hardened with store-level epoch abort guards guarding in-flight vector indexing against post-reset insertion.

3. Native Agent Lifecycle Middlewares
Drop-in sovereign memory hooks for the primary agent frameworks with pre-inference context enrichment and post-tool execution tracking:
• LangChain: EnigmaLangChainMemory
• Vercel AI SDK: createEnigmaContextMiddleware
• LlamaIndex: EnigmaLlamaIndexRetriever

4. Memory Intelligence & Hybrid RAG Engine
8-class ontology classification, temporal decay modeling, contradiction DAG supersession, and subword lexical hashing + BM25 Reciprocal Rank Fusion (RRF) with semantic candidate reranking.

5. Cryptographic State Receipts v2
Cryptographically hiding commitments C = SHA256(SHA256(plaintext) || r) with strict 32-byte blinding factors, explicit state transition witnesses (append, update, revoke), and an offline CLI verifier.

[Enigma Memory @Engimamemory]

Hey guys busy day with the fam. I'll be at the office all week so I will be posting a lot more tomorrow. Let me know if you guys have an ideas. There is still a lot more to be built out to make this robust.

For today I'm releasing Enigma v0.1.21: the sovereign AI memory architecture we've been building for autonomous agents.

The premise is simple: If an agent's memory lives in plaintext inside a centralized vector database or cloud log, that agent is not sovereign.

We built Enigma to solve key custody, semantic recall, state verification, and agent-to-agent coordination without leaking a single byte of plaintext:

1. Local-First Memory Vault: AES-256-GCM encryption at rest with 600k-round PBKDF2/scrypt derivation and 384-D deterministic lexical embeddings (Int8 quantized to save 75% RAM).

2. Solana ZK State Proof Rail: Every memory append commits to an incremental Poseidon Note Tree, with Aztec-style IMT nullifiers for revocations and auto-anchoring to Solana.

3. Sealed Sale (ZKCP) Marketplace: An 87,422-constraint Groth16 circuit with Poseidon-CTR cipher folding for fair, atomic zero-knowledge exchange of confidential memories and weights.

4. GhostMesh Encrypted P2P Relay: True X25519 ECDH key agreement with AES-GCM destination AAD binding and Ed25519 envelope signatures for decentralized agent coordination.

5. Unified Native SDKs: Drop-in 3-line quickstarts in TypeScript (enigma-memory/sdk) and Python (enigma-sdk), with native LangChain, LlamaIndex, and Vercel AI SDK middlewares + Swarm Orchestrator.

Zero Plaintext On Disk. Zero Plaintext On Chain.

📦 npm: npm i enigma-memory
🐍 pip: pip install enigma-sdk
📖 Architecture & Docs: https://t.co/usWQxzJoi0

[Terry Diovaldy @tdiovaldy]

@Engimamemory @TochiOnChain Well its not healthy

[Terry Diovaldy @tdiovaldy]

@Engimamemory Any update ?

[Enigma Memory @Engimamemory]

Let me know what else you guys would like to see built on top of the privacy layer we are building.

I'm open to all suggestions.

[Terry Diovaldy @tdiovaldy]

@Modulr402 Any update?

[Terry Diovaldy @tdiovaldy]

@KriptoCapris Yes please🎉

[MineBench - efficient computing power @MineBenchdapp]

We’re considering a different direction for the GPU side of MineBench.

Instead of using GPUs primarily for mining, we could use them for **AI compute**, where the hardware can potentially provide more useful and valuable work.

This raises a few questions:

• Should we build our own AI compute marketplace, or connect to an existing compute network?
• Should this become a separate app with its own token, or stay within MineBench and solana:67ipDsgK6D7bqTW89H8T1KTxUvVuaFy92GX7Q2XFVdev ?
• If we build something new, what should we call the app and token?

We’d like to hear what the community thinks before we decide on the architecture and direction.

[MineBench - efficient computing power @MineBenchdapp]

We started with automatic rewards for mining. After testing the system and successfully distributing rewards to miners, we’re now bringing the same reward system to benchmarks and stress tests.

Complete a benchmark or stress test → earn rewards based on the shares you find → receive them after our pool gets its next mining payout.

Pool rewards and payouts are posted automatically in our Discord.

Benchmark rewards are available starting with MineBench v0.8.30.

Mine. Benchmark. Stress test. Get rewarded.

[MineBench - efficient computing power @MineBenchdapp]

The links to the open-source parts of MineBench have been updated on our website.

You can now access the latest public source code directly from the website.

🔗 https://t.co/MpweewSpE3

#MineBench #OpenSource #BMT #Mining

[MineBench - efficient computing power @MineBenchdapp]

You can see our beautiful hashrate chart on the dashboard page:

https://t.co/mDYvSyNyY6