Archive: 2026/09

September 18 posts

• Entity Deduplication From 50 Million Pairs to Thousands Sep 20
• Qwen3.8-Omni-Flash vs Gemini Flash Real Cost per Task Sep 19
• The Claude OpenAI Security Incident, Decoded for Builders Sep 18
• Gemini 3.8 Live Extended Thinking Needs a Latency Budget Sep 17
• LLM Prompt Caching Can Cut Input Costs Up to 90% Sep 16
• Open Weights vs Frontier Models Is a 4.4 Month Call Sep 15
• Async GRPO Training on Serverless GPUs Cuts RL Cost Sep 14
• AI Agent Monitoring Beyond the Dashboard Sep 13
• Devin GPT-6 Astra Self-Testing Moves Review to Evidence Sep 12
• Training a Diffusion Model From Scratch in 3.5 Days Sep 11
• OpenAI Navier-Stokes, a Reported $40M Lesson in Verification Sep 09
• Netflix MAPS Teardown and What Actually Transfers Sep 08
• Meta's AI Second Brain Runs on a Knowledge Supply Chain Sep 07
• AI Text Watermarking in Python, Three Families Tested Sep 06
• KV Cache Math for Million Token Agent Runs Sep 05
• GPT-6 Astra Pricing at $6 an Hour, Audited Sep 04
• Prompt Dependency Graphs That Shrink Your Retest Set Sep 03
• TontaubeV1 Review With Serving Math You Can Run Sep 01