Agentic RAGJun 2026· 9 min read
Why your RAG is hallucinating (and how we fixed ours)
Most 'RAG hallucinations' aren't model failures — they're retrieval failures. A walkthrough of the eval suite that finally caught ours.
ReadInsights
Field notes on building AI that ships.
Honest write-ups of what worked, what didn't, and what we'd do differently. No vendor talking points.
AI AgentsMay 2026· 7 min read
Designing agents that know when to stop
Tool loops, budget limits, and the underrated art of letting an agent fail gracefully and hand off to a human.
ReadVoice AIMay 2026· 12 min read
Sub-800ms voice agents on Twilio: the latency budget
Where the milliseconds actually go in a production voice loop — and the four optimizations that bought us the most.
ReadStrategyApr 2026· 6 min read
The AI ROI conversation enterprise buyers actually want
How to frame an AI initiative for a CFO without overpromising or underselling. A template inside.
ReadEvaluationApr 2026· 8 min read
Building a gold set you'll still trust in six months
Practical patterns for assembling, versioning, and rotating an eval gold set without it going stale.
ReadVoice AIMar 2026· 11 min read
Multilingual voice for emerging markets: lessons from 80k farmers
What we learned shipping voice AI in three languages for users on low-end phones over weak networks.
ReadGet the field notes by email.
One write-up a month. No fluff. Unsubscribe anytime.