Writing about AI-assisted development, Claude Code workflows, and building software with AI.https://adriacidre.com/https://adriacidre.com/blog/bounded-extraction/?utm_source=rss/https://adriacidre.com/blog/bounded-extraction/?utm_source=rss/Confidence is what the model thinks of itself. Bounds are what your system thinks of the model. They are independent signals; use both.Tue, 16 Jun 2026 00:00:00 GMThttps://adriacidre.com/blog/dont-delete-the-row-delete-the-key/?utm_source=rss/https://adriacidre.com/blog/dont-delete-the-row-delete-the-key/?utm_source=rss/Append-only history and GDPR-style erasure look incompatible. They are not, if you put the secret in two places.Mon, 18 May 2026 00:00:00 GMThttps://adriacidre.com/blog/there-is-no-workflow-engine/?utm_source=rss/https://adriacidre.com/blog/there-is-no-workflow-engine/?utm_source=rss/My agent pipeline runs for hours, survives Ctrl+C, and resumes from where it stopped. There's no Temporal, no queue, no database. The state lives in YAML files and commits.Fri, 15 May 2026 00:00:00 GMThttps://adriacidre.com/blog/multi-tenant-isolation-vs-awareness/?utm_source=rss/https://adriacidre.com/blog/multi-tenant-isolation-vs-awareness/?utm_source=rss/Every system I built had a tenant_id column and a discipline problem. This is the first one where the database itself refuses to return the wrong rows.Wed, 13 May 2026 00:00:00 GMThttps://adriacidre.com/blog/voice-agent-perceived-latency/?utm_source=rss/https://adriacidre.com/blog/voice-agent-perceived-latency/?utm_source=rss/Two agents with identical latency can feel completely different. The gap is fixable at the orchestration layer, without touching the model.Tue, 05 May 2026 00:00:00 GMThttps://adriacidre.com/blog/maintenance-agent-14-refusals/?utm_source=rss/https://adriacidre.com/blog/maintenance-agent-14-refusals/?utm_source=rss/A maintenance agent filed 559 bugs and fixed 412 on its own. The interesting number is the 14 it refused to touch.Sun, 26 Apr 2026 00:00:00 GMThttps://adriacidre.com/blog/parallelism-was-a-workaround/?utm_source=rss/https://adriacidre.com/blog/parallelism-was-a-workaround/?utm_source=rss/Feature branches and merge queues were how we coped with slow humans. When an agent finishes a task in ten minutes, the overhead is the bottleneck.Tue, 21 Apr 2026 00:00:00 GMThttps://adriacidre.com/blog/self-healing-parsers-instead-of-vision/?utm_source=rss/https://adriacidre.com/blog/self-healing-parsers-instead-of-vision/?utm_source=rss/I stopped feeding hospital PDFs to a vision model. When the layout changes, the AI fixes the parser instead — and production never sees a token.Sun, 19 Apr 2026 00:00:00 GMThttps://adriacidre.com/blog/voice-agent-turn-taking/?utm_source=rss/https://adriacidre.com/blog/voice-agent-turn-taking/?utm_source=rss/Most builders assume end-of-turn detection is a silence threshold. That model breaks in production. The fix is architectural: four probabilistic events, speculative reasoning, and everything downstream cancellable.Wed, 15 Apr 2026 00:00:00 GMThttps://adriacidre.com/blog/explorer-agent-breaks-things-on-purpose/?utm_source=rss/https://adriacidre.com/blog/explorer-agent-breaks-things-on-purpose/?utm_source=rss/An 8th specialist that touches the browser — navigating, clicking, and typing to find bugs that scripted journeys miss. How exploratory testing with Playwright catches what verification alone cannot.Mon, 23 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/specialist-agents-different-eyes/?utm_source=rss/https://adriacidre.com/blog/specialist-agents-different-eyes/?utm_source=rss/Seven specialists, each with their own checklist — QA, UX, UI, Security, Performance, Data Leakage, Language. How splitting evaluation into focused agents catches more.Wed, 18 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/batch-ai-orchestration-overnight/?utm_source=rss/https://adriacidre.com/blog/batch-ai-orchestration-overnight/?utm_source=rss/The leap from AI-assisted coding to autonomous batch processing. Fresh context per task, filesystem locks, model routing, and orchestration that runs while you sleep.Wed, 18 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/reliability-patterns-ai-systems/?utm_source=rss/https://adriacidre.com/blog/reliability-patterns-ai-systems/?utm_source=rss/Exponential backoff, dual timeouts, SSE heartbeats, idempotency caches — the unglamorous patterns that keep LLM-powered systems running at 3am.Thu, 12 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/qa-run-ai-driven-qa/?utm_source=rss/https://adriacidre.com/blog/qa-run-ai-driven-qa/?utm_source=rss/Real browser, predefined journeys, specialist agents checking every step. How QA findings feed back into the issue tracker to close the development loop.Sun, 08 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/work-issue-autonomous-implementation/?utm_source=rss/https://adriacidre.com/blog/work-issue-autonomous-implementation/?utm_source=rss/The AI picks up a planned issue, implements it, tests it, commits, and marks it done. How headless batch mode lets me plan in the afternoon and wake up with code in the morning.Wed, 04 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/semantic-cache-sub-10ms-ai-responses/?utm_source=rss/https://adriacidre.com/blog/semantic-cache-sub-10ms-ai-responses/?utm_source=rss/Users ask similar questions in different words. Semantic caching with pgvector turns repeated intent into instant answers — no LLM call, no embedding, no retrieval pipeline.Wed, 04 Mar 2026 00:00:00 GMThttps://adriacidre.com/blog/context-window-management-long-ai-conversations/?utm_source=rss/https://adriacidre.com/blog/context-window-management-long-ai-conversations/?utm_source=rss/Context windows fill up fast in long AI conversations. Sliding windows, progressive compression, and token budgeting — the patterns I built before I knew their names.Thu, 26 Feb 2026 00:00:00 GMThttps://adriacidre.com/blog/plan-issue-collaborative-planning/?utm_source=rss/https://adriacidre.com/blog/plan-issue-collaborative-planning/?utm_source=rss/Vague instructions produce vague implementations. How I sit with the AI to refine rough ideas into concrete, implementable specs — before any code gets written.Wed, 25 Feb 2026 00:00:00 GMThttps://adriacidre.com/blog/ai-assisted-development-loop-not-chat/?utm_source=rss/https://adriacidre.com/blog/ai-assisted-development-loop-not-chat/?utm_source=rss/The bottleneck isn't the AI — it's me. Three slash commands that form a loop: Plan, Work, QA. How structured phases replaced ad-hoc chatting and made AI development manageable.Wed, 18 Feb 2026 00:00:00 GMT