InDesign Systems CollectivebyCristian Morales Achiardi·1d agoThe autonomy vs alignment dial for AIBy treating your harness as a composition dial, you can leverage agentic silence to produce better, more creative user interfaces.A response icon1A response icon1
Saichithra Swaminathan·1d agoMost AI product specs I read are wishes written as requirements.“The assistant should give helpful, accurate, on-brand answers.” Every PRD says something like this, and you can’t test a word of it. With…A response icon1A response icon1
Dealonai·Oct 26What Are AI Evals? A Guide to Types, Methods, and Best PracticesImagine that a customer comes up and a support chatbot tells them that they can request a refund within 90 days when the actual policy…
Sandeep Chakravartty·Oct 25Your Test Suite Is Green.The skills that made you a great tester are exactly the ones the AI era is short on. You just need to point them at a new kind of system.
Micheal Lanham·Oct 23The Eval Ladder: Routing Agent Grading From Code Checks to Jev to LLM JudgesEvals Are Your IP, Day 3 of 10: grade each criterion on the cheapest rung that can decide it, and pay a person to reread the confident…A response icon2A response icon2
Micheal Lanham·Oct 21Model-Swap Testing for AI Agents: Pinned IDs, pass^k and a Non-Inferiority GateEvals Are Your IP, Day 1 of 10: models and even fine-tuned weights retire on someone else’s schedule, so the asset you own is your…
Priyam·Oct 18are evals the new prd?my thoughts on evals, vibes and shipping ai products that don’t embarrass you…
InTowards AIbyRick Hightower·Oct 16Langfuse VIII: The Playground Looked Fine. That Is Not a Langfuse Experiment.LangFuse Part 8: Scoring live traces is how you notice a miss. A Langfuse experiment is how you refuse to ship the next prompt change until…A response icon1A response icon1
Riyas P·Oct 16Fine-Tuning a Small Decision Model: A Gmail Experiment with Laya and JEVI’ve been experimenting with simple AI-based classification for my Gmail inbox.
