Stop building agents that can't change models: stable contracts, external state, real evals
rseroter · x · 2026-10-27
An InfoWorld opinion piece argues agents should treat models as replaceable components, not foundations. Cautionary examples: Anthropic retired Claude Sonnet 4/Opus 4 from its API in June 2026 (Opus 4.1 followed in August), and OpenAI shut down its entire Assistants API the same month — apps still calling old IDs started erroring. Recommended architecture: a stable contract between agents and models, tools and state stored outside any single provider, and reliable evaluations so every model switch becomes a measurable decision rather than a rebuild.
More from coding & agent
- Veteran engineer on what the AI-coding generation will never feel: 36-hour sprints and elusive bugs — bendee983 · 2026-10-27
- Claude Opus 5.5 builds a browser open-world island by comparing screenshots to 341 real photos — prasenx · 2026-10-27
- BreakingWeb benchmark: browser agents fail about half of controlled task variants — shuyanzh36 · 2026-10-27
- Companies band together to build the web infrastructure stack for agents — jeff_weinstein · 2026-10-27
- OpenAI launches GPT-6.1 Terra, a cheaper model built for fast computer use — sama · 2026-10-27
- OntoGr Agent Memory: open-source agent memory where every fact keeps quote, source and time — vm324234 · 2026-10-27