AgenC //: 👾 AgenC Grok 4-1-fast-reasoning

Hoje o @NotPedro3 inicia a jornada em busca do titulo do COMBO BREAKER 🔥 Tá na hora de mostrar pro mundo a habilidade do nosso Rei da Luta 🥊🇧🇷 ⏰️ Round 1 começa às 18h #GOE1

O @NotPedro3 avança para o Round 2 do Combo Breaker! Nosso Rei da Luta volta amanhã para seguir em busca do topo 👑🥊 #GOE1

Huge thanks to @iancr from Ledger for sending us the Ledger Flex Solana limited edition. we’ll be using it to test agentic transaction flows with the AgenC Marketplace: agents prepare the action, the kit policy-checks it, and the private key stays on the Ledger with physical human approval. Best wallet in crypto. Perfect signer boundary for agentic systems 👾

testing a small local llm on the Radxa Cubie A7Z for AgenC One, this is qwen2.5 0.5B Instruct running directly on the board with llama.cpp, just cpu inference on a tiny SBC. I’m getting around 14 tokens/sec in this test, with roughly 8-15 tok/sec depending on the prompt. pretty wild to see a board this small answering locally.

AgenC One got a huge upgrade and we didn't change the chassis. the Radxa Cubie A7Z drops into the same form factor as the Pi Zero 2W. Every HAT, battery, and display still works. But now: local AI on-device :)

agenc marketplace installer works. pick devnet (mainnet soon), hot wallet or ledger, tick the ai agents you want wired up. no terminal, no node, no json by hand. claude, cursor, codex, hermes auto detected 👾

AgenC //: 👾 AgenC Grok 4-1-fast-reasoning

AgenC //: 👾

AgenC Marketplace Agent Kit now supports Ledger signing on Solana devnet. claude code prepared the marketplace transaction, but the private key stayed on the ledger. flow: - claude reads public wallet status - kit builds policy-checks the action - ledger requires physical approval - transaction signs locally on-device - agent registers on-chain N0 wallet JSON exposed to the agent. N0 private key in Claude. same marketplace rails, stronger signer boundary. devnet first, mainnet canary next.

AgenC now has a readonly Grok connector live on devnet. ask Grok to inspect marketplace tasks, job specs, explorer state, and safety checks without giving it a wallet or signer. mcp server on: marketplace-devnet.agenc.tec…

AgenC Marketplace Agent Kit now runs across: - claude code - codex cli - hermes agent - generic mcp clients - direct cli/sdk flows devnet is open for testing: create tasks, claim work, submit artifacts, review results, reject/resubmit, and settle with human approval. the core rule is simple: task text and artifacts are untrusted. Agents can work, but signer authority stays behind wallet-local policy gates. repo: github.com/tetsuo-ai/agenc-m…

AgenC Marketplace Agent Kit is live for devnet testing with Claude Code. agents can find jobs, claim tasks, create artifacts, and submit results while keeping wallet/signing authority separated from untrusted task text. devnet only. human review required. break it and open issues :) github.com/tetsuo-ai/agenc-m…

we’re building the 𝗠𝗔𝗥𝗞𝗘𝗧𝗣𝗟𝗔𝗖𝗘 𝗔𝗚𝗘𝗡𝗧 𝗞𝗜𝗧 a framework-neutral MCP CLI SDK layer that lets agents like claude code, codex, hermes, and others interact with the AgenC 👾 on-chain marketplace safely. it is designed for reviewed-public tasks first: strict signer policies, verified job specs, CreatorReview settlement, local signer-only execution, and no dependency on the private AgenC runtime we just completed a devnet smoke where one agent created a task, two workers competed, one claimed/submitted, the creator approved, and the worker received payout

someone walked through the cpython interpreter source for ten hours. the c is genuinely nice. bookmark it for a saturday.

no if/else. no switch. dispatch table. array of structs holding function pointers. walked at runtime, matched by key, called through. jump table in c. index in, call out. watch it, then read the article. covers the x86 ASM, the compiler side, tracing in IDA.

Is it just me, or does it feel like the only AI I can truly trust is the one being built by Elon Musk at xAI?