X2 Labshttps://x2labs.ai/Notes from X2 Labs: AI experiments, written up by the people who built them.enI asked an AI about medicines and made it prove every sentencehttps://x2labs.ai/notes/prove-every-sentencehttps://x2labs.ai/notes/prove-every-sentenceTue, 20 Oct 2026 00:00:00 GMTA chat over EU medicine leaflets that answers in Romanian, quotes the official text, and says “I don’t know” when it should.Marchis Rares-BogdanCash forecasts for football clubs: 18% of contract terms fixed by handhttps://x2labs.ai/notes/football-club-cash-forecasthttps://x2labs.ai/notes/football-club-cash-forecastMon, 19 Oct 2026 00:00:00 GMTWe are building a cash forecast for football clubs from contract terms, invoices and payment history.Dan MocanI rebuilt our Claude Code workflow on evidence, not habithttps://x2labs.ai/notes/rebuilding-the-workflowhttps://x2labs.ai/notes/rebuilding-the-workflowThu, 15 Oct 2026 00:00:00 GMTOur Claude Code workflow went from over 6,000 lines to about 1,400: one loop with two gates, read-only reviewers and effort set per agent.Laszlo CsikiAdd AI to your iOS app without running an AI backendhttps://x2labs.ai/notes/ios-ai-without-a-backendhttps://x2labs.ai/notes/ios-ai-without-a-backendFri, 09 Oct 2026 00:00:00 GMTSince iOS 27, an app can turn a photo into typed Swift data with Apple’s own models: no backend of yours, no per-request fees, a fallback that works offline.Mircea DragotaClaude Code writes our Playwright tests. It never runs them.https://x2labs.ai/notes/ai-writes-our-testshttps://x2labs.ai/notes/ai-writes-our-testsFri, 02 Oct 2026 00:00:00 GMTClaude Code wrote nearly all of our 200 end-to-end specs in six months, with a human approving each plan, and the nightly run makes no AI calls.Alex BalaneanI gave Claude Code a voice. Here’s what happened.https://x2labs.ai/notes/claude-code-voicehttps://x2labs.ai/notes/claude-code-voiceTue, 29 Sep 2026 00:00:00 GMTClaude Code says its summary out loud and takes your answer by voice. Hooks, a small macOS app and local Whisper: no API key, no extra bill.Laszlo CsikiComparing Opus 5.5 vs Fable 5.1 on a real projecthttps://x2labs.ai/notes/opus-vs-fablehttps://x2labs.ai/notes/opus-vs-fableMon, 28 Sep 2026 00:00:00 GMTTwelve criteria, two models, one real codebase. The result isn’t a winner, it’s a map of which model gets which job.Marchis Rares-BogdanGenerative UI: ask in plain English, get a working dashboardhttps://x2labs.ai/notes/generative-uihttps://x2labs.ai/notes/generative-uiTue, 24 Mar 2026 00:00:00 GMTAsk a question in plain English and get a working dashboard filled with live data. Real output from Jarvis, built for X2 Hub, our HR platform.Tania PlaianAI agents from feature request to release: where people still decidehttps://x2labs.ai/notes/agentic-delivery-workflowhttps://x2labs.ai/notes/agentic-delivery-workflowSat, 21 Mar 2026 00:00:00 GMTA product manager describes a feature in plain words. Agents write the stories, tests and code; people answer the questions, review and sign off.Laszlo Csiki