https://willianpinho.com/blog Thoughts on software architecture, AI/ML systems, production engineering, and building products as a solo founder. en-us Mon, 05 Oct 2026 21:03:28 GMT https://willianpinho.com/blog/recall-is-not-enforcement https://willianpinho.com/blog/recall-is-not-enforcement Tue, 18 Aug 2026 14:00:03 GMT Palantir published this discipline in 2017 and Anthropic files it under debugging rather than auditing. I borrowed the protocol from Atomic Red Team and ran it against all ninety guard hooks in my own harness: 28 of the 31 that can refuse were watched refusing, two cannot fire at all, and two more block correctly while not being armed. ai agents guardrails process https://willianpinho.com/blog/api-layer-that-was-fake https://willianpinho.com/blog/api-layer-that-was-fake Sat, 15 Aug 2026 13:00:03 GMT A generated client with a baseUrl nothing ever reads, and no transport anywhere. I wrote the retry and error layer the usual advice asks for, ran it against that boundary and then against a real one, and counted what the switch turned on: three dead branches, one that was wrong in a way nothing local could reveal, and a timeout that cancels nothing. Then six classifiers instead of one, because the first three defects might have been mine. architecture api integration testing https://willianpinho.com/blog/guard-that-blocked-its-own-investigation https://willianpinho.com/blog/guard-that-blocked-its-own-investigation Sat, 08 Aug 2026 13:00:04 GMT Deny-by-default is the right posture for agent tool access and the wrong one for a workflow guard. The difference is which error you cannot recover from, and a hook that blocked cat, grep and git diff on the very script it was guarding is what taught me to check. devtools automation ai agents https://willianpinho.com/blog/do-your-agent-system-prompts-do-anything https://willianpinho.com/blog/do-your-agent-system-prompts-do-anything Thu, 06 Aug 2026 14:00:03 GMT A null-prompt ablation across 19 agent configurations: 13 survive the noise floor, four say more about the tests than about the prompts, and two stay unresolved because their assertion graded a literal output signature rather than behaviour. Anthropic runs the same kind of ablation on Claude Code and reports the opposite direction. The difference is scope. ai llm testing agents https://willianpinho.com/blog/success-code-that-hides-a-failure https://willianpinho.com/blog/success-code-that-hides-a-failure Sat, 01 Aug 2026 13:00:04 GMT An HTTP 200 means the request came back, not that the work happened. A rotated token, an alerting path that went silent for days, and why the same trap gets worse once agents are wired to real tools. observability api ai agents https://willianpinho.com/blog/multi-llm-routing-production-failure-modes https://willianpinho.com/blog/multi-llm-routing-production-failure-modes Wed, 29 Jul 2026 16:14:03 GMT Multi-LLM routing looks clean on a whiteboard. In production the real risk is the cost math that hides its own downside, latency treated as a single number instead of a distribution, and the silent failures that return a clean HTTP 200. llm routing ai production https://willianpinho.com/blog/fail-close-agent-tool-access-default https://willianpinho.com/blog/fail-close-agent-tool-access-default Wed, 15 Jul 2026 14:45:18 GMT Why AI agents wired to real tools through MCP should default to deny, and how to build the per-operation, audited RBAC that makes fail-close real. mcp ai agents security https://willianpinho.com/blog/mcp-in-production https://willianpinho.com/blog/mcp-in-production Fri, 03 Jul 2026 16:38:59 GMT What survived contact with real usage maintaining an MCP server across three clients I never built for: fail-close defaults, replayable observability, and tool descriptions as the real interface. mcp agent-engineering software-engineering observability https://willianpinho.com/blog/mcp-gateway-production-readiness-audit https://willianpinho.com/blog/mcp-gateway-production-readiness-audit Tue, 30 Jun 2026 20:40:31 GMT I ran an MCP-gateway production-readiness audit on a popular open-source LLM gateway. Here's what it found. mcp llm ai security https://willianpinho.com/blog/pre-mortem-as-plan-debugger https://willianpinho.com/blog/pre-mortem-as-plan-debugger Tue, 23 Jun 2026 17:00:48 GMT The pre-mortem is the cheapest way I know to catch the preventable failures before they ship. The 25-minute version I actually run, and when not to bother. engineering planning productivity career https://willianpinho.com/blog/mcp-gateway-scan-production-readiness https://willianpinho.com/blog/mcp-gateway-scan-production-readiness Tue, 09 Jun 2026 12:00:00 GMT A free MIT CLI that scans a repo for seven classes of AI-agent-gateway anti-patterns — tool-access/RBAC, fail-close, supply-chain pinning, observability, multi-LLM routing, secrets, production-readiness — runs as a CI gate, and exposes itself as an MCP server. mcp ai security devops agents https://willianpinho.com/blog/the-cross-origin-observability-tax https://willianpinho.com/blog/the-cross-origin-observability-tax Fri, 05 Jun 2026 12:00:00 GMT When one team adds a request header, the backend CORS allowlist becomes a coordination point. Here is what that tax actually costs, why it fails silently, and four patterns for handling it. observability cors otel reliability distributed-tracing https://willianpinho.com/blog/the-agent-runtime-event-model https://willianpinho.com/blog/the-agent-runtime-event-model Sun, 24 May 2026 12:00:00 GMT Four runtime events, four policy-attachment surfaces, and the hook coverage at each event is what separates a demo from something a regulated buyer will sign for. Framework selection is a distraction. ai agents security architecture mcp https://willianpinho.com/blog/large-file-mcp-intelligent-chunking https://willianpinho.com/blog/large-file-mcp-intelligent-chunking Wed, 15 Apr 2026 12:00:00 GMT A production-ready MCP server that makes working with large files in Claude Desktop and AI assistants effortless — smart chunking by file type, line-level navigation, regex search, and streaming for files up to 10 GB. mcp ai claude typescript developer-tools