The "pelican benchmark" by Simon Willison arbitrates Kimi K3
An SVG illustration test acts as a quick progress gauge: Simon Willison publishes his reading of K3 on July 16 and 17.
5 h ago 3 3
"Claude Code: Anatomy of a Misfeature" - when public review becomes the real QA
Olaf Alders publishes on July 17 a well-argued critique of a Claude Code feature - the "public post-mortem of a misfeature" format is becoming a standard of hype fatigue.
5 h ago 4 5
The AI context gap: enterprise AI's real problem isn't retrieval, it's trust
VentureBeat raises the question that has been lingering in CIOs' minds: enterprise AI projects fail less on the technical aspects (retrieval, embedding, RAG) than on the trust in what they return—and on the governance of their sources.
yesterday 11 8
Google Genkit v.Agents: detached turns and human-in-the-loop are now in preview
Google releases a preview of the Genkit Agents API for TS and Go, with detached turns (the agent survives disconnection) and HITL with anti-forgery validation - two building blocks that post-mortem agents have been requesting for twelve months.
Jul 14, 2026 at 15:44 8 9
AWS multi-region: the pre-flight discovery call that blocked global failover
InfoQ publishes the analysis of a pre-flight discovery call, present in every customer session for years, revealed as an obstacle to global failover during AWS regional outages. A lesson in systems more than a patch.
Jul 13, 2026 at 17:27 10 7
Malaika: Understanding malware through "triangulated" agent-based reasoning
A new arXiv paper proposes a multi-agent system that reconstructs malicious behaviors from partial evidence - no hype, the real utility of LLMs in malware analysis.
Jul 13, 2026 at 09:13 12 8
Migrating a production agent to GPT-5.6: 2.2× faster, 27% cheaper - the real post-mortem
The first quantified feedback from a frontier→frontier migration. A textbook case for those preparing their next switch.
Jul 13, 2026 at 02:19 9 7