https://entalpa.com/en/guides Long-form pieces on spec-driven development for AI coding agents, built from what developers said about their own projects and from incidents in ours. en https://entalpa.com/en/guides/prd-template https://entalpa.com/en/guides/prd-template Fri, 25 Sep 2026 00:00:00 GMT A product requirements document template written for the agent that reads it: requirement IDs, testable acceptance criteria, explicit out-of-scope, and open questions that stop the build. Copy it, fill it, hand it to Claude Code, Cursor or Codex. https://entalpa.com/en/guides/spec-driven-development-with-claude-code https://entalpa.com/en/guides/spec-driven-development-with-claude-code Tue, 15 Sep 2026 00:00:00 GMT What Claude Code already gives you for building from a spec (CLAUDE.md, plan mode, MCP), the lightest loop that is actually spec-driven, the three places it breaks, and how to hand Claude Code a spec it cannot rewrite. https://entalpa.com/en/guides/spec-driven-development https://entalpa.com/en/guides/spec-driven-development Tue, 15 Sep 2026 00:00:00 GMT Spec-driven development, defined in one paragraph; the tools that carry it; the two Ask HN threads asking whether anyone still does it; and the one complaint that outlived the hype: the agent still builds things that are not in the spec. https://entalpa.com/en/guides/agent-guesses-requirements https://entalpa.com/en/guides/agent-guesses-requirements Mon, 31 Aug 2026 00:00:00 GMT Agents get called sycophantic for agreeing with everything and shipping the wrong feature. The failure is not agreeableness — a model cannot object to a constraint nobody stated, so it fills the hole with the most typical answer. https://entalpa.com/en/guides/reviewing-the-spec-not-the-diff https://entalpa.com/en/guides/reviewing-the-spec-not-the-diff Mon, 31 Aug 2026 00:00:00 GMT 45% of developers say debugging AI code takes longer than writing it. Generation sped up 100x; reading unfamiliar code did not speed up at all. The reviewable unit has to move upstream of the diff. https://entalpa.com/en/guides/what-your-agent-breaks https://entalpa.com/en/guides/what-your-agent-breaks Mon, 31 Aug 2026 00:00:00 GMT An agent sent to fix a bug in our own repository standardised behaviour that was deliberately non-standard, because nothing recorded that it was deliberate. What that failure actually is, and what stops it. https://entalpa.com/en/guides/the-vibe-coding-wall https://entalpa.com/en/guides/the-vibe-coding-wall Mon, 31 Aug 2026 00:00:00 GMT Vibe coding is excellent at the prototype and bad at the second month. Five specific ways it breaks, quoted from developers describing their own projects, and what genuinely helps against each one.