Gemini Decisions API: confidence scores need calibration checks
Google introduced the Gemini Decisions API on November 11 for bounded multimodal choices. The announcement describes scoring candidate options rather than generating a free-form answer.
The developer’s AI workbench
AI moves fast.
Build with a clearer head.
Coding news, useful research, and a little orcish curiosity. Follow the changes that matter to the work you do.
Worth your attentionGemini Decisions API: confidence scores need calibration checksAgent operations need a working stop buttonClaude Fable 5.2 and Mythos 5.2: read the access distinction
On the workbench
All stories
Keep digging
Browse the archive
AI crawling preferences deserve a deliberate publishing policy
OpenAI announced support for a shared robots.txt training-preference signal on October 30. Publishers should distinguish discovery from model training.
AI secret detection: the useful metric is a safer repository
GitHub described a context-aware secret-detection model on October 13. Its report focuses on detection quality and scanning efficiency.
Claude Haiku 5.5: smaller models need explicit boundaries
Anthropic introduced Claude Haiku 5.5 on October 7 for high-volume work including summarization, subagents and browser use. Routing remains an engineering decision.
ReviewBench: what a code-review benchmark should answer
GitHub introduced ReviewBench on October 5 as an open benchmark for AI code review. The important question is whether review catches consequential defects.
September’s Google AI recap: separate releases from adoption
Google’s October 2 recap collected its September AI announcements, including Gemini 4 Argon. A monthly recap is a useful discovery map, not an adoption checklist.
GPT-6.1 Sol: evaluate the work, then the price
OpenAI introduced GPT-6.1 Sol for coding, computer use and professional tasks on September 29. The release makes model selection a workload question.
Every story has a source.
We follow original announcements, distinguish claims from tests, and explain what a release could mean for your workflow.
Read our editorial standardsstory = {
source: "primary",
claims: "attributed",
curiosity: true
}
source: "primary",
claims: "attributed",
curiosity: true
}