Tag: agents

All the articles with the tag "agents".

• Production AI Is Mostly Boundary Design

  Published:Jun 16, 2026 |  at 03:00 PM

  Production AI works when the model owns the ambiguous middle, not the whole workflow. The lesson shows up in a lead dedup case study, but the real story is boundary design.

• Agents Shouldn't Grade Their Own Homework

  Published:Jun 9, 2026 |  at 03:00 PM

  GitHub built an impressive stack to validate non-deterministic agents — a prefix-tree graph, dominator analysis, a multimodal LLM. I needed the same thing for Grove's agents and found the hard part didn't exist. The reason is the whole argument for structured trajectories.

• Opus 4.8: The Model Learned to Doubt Itself

  Published:May 31, 2026 |  at 04:00 PM

  First impressions of Claude Opus 4.8 from real work: the honesty upgrade is the real story, the benchmarks are a sideshow, and the one place it still trips is — ironically — knowing when to do one thing at a time.