Agent Vault vs Auto-Arch Tournament

Agent Vault is an open-source credential broker from Infisical that solves one of the nastiest unsolved problems in AI agent security: AI agents are non-deterministic and vulnerable to prompt injection attacks that could trick them into leaking secrets. The solution is elegant — Agent Vault never gives credentials to the agent at all. Instead, it acts as an HTTPS proxy, intercepting the agent's outbound API calls and injecting credentials at the network layer. The flow is simple: give the agent a scoped session token and set HTTPS_PROXY to Agent Vault's local server. The agent calls APIs normally; Agent Vault transparently swaps in the real credentials before the request leaves the machine. The agent literally cannot leak what it never had. AES-256-GCM encryption with optional Argon2id password wrapping protects the vault, and all proxied requests are logged (method, host, latency) without recording sensitive bodies. Works out of the box with Claude Code, Cursor, Codex, custom Python/TypeScript agents, and any HTTP-speaking process. Infisical is a credible backer — they already run one of the most popular open-source secrets managers. This is MIT-licensed with enterprise features planned. For teams deploying agents in sandboxed environments, this is the missing security primitive.

Auto-Arch Tournament is an autonomous research system where an AI agent iteratively proposes, implements, and validates microarchitectural improvements to a RISC-V CPU. Starting from a standard 5-stage pipeline, the loop runs hypotheses in parallel, each going through formal verification (53 symbolic checks), cycle-accurate simulation, multi-seed FPGA place-and-route, and CoreMark CRC validation. Only hypotheses that beat the current champion get merged; everything else gets discarded. Starting from 301 iterations/second, the system hit 577 iter/s (+92%) across 73 attempts in 9.8 hours — producing a design 26% faster and 40% smaller in LUTs than the baseline. The insight the author drives home is that the real innovation isn't the AI agent — it's the verifier. The orchestrator is hardcoded to prevent agents from manipulating their own evaluation gates, a simple but critical design constraint that turns a creative process into a trustworthy one. Without a rigorous verification harness, agent-driven optimization becomes a confidence trick. This is early but fascinating proof that AI-driven hardware design loops can produce commercially meaningful gains. The repo uses Claude Code or Codex as the coding agent, SystemVerilog for the RTL, and standard open-source EDA tooling (Yosys, nextpnr, Verilator). It's a compelling template for anyone building agentic optimization loops where correctness matters.