A single daily edition of announcements from the AI coding tools you use, read from a curated source list, summarised, and linked back to the original. No infinite feed, no account, no advertising.
Version 1.0, released 20 August, swaps the HTTP layer from httpx to httpx2, an API-compatible fork maintained by the Pydantic team. Tracing, APM instrumentation and HTTP mocking that hook httpx directly stop seeing the SDK's requests until they are repointed. The legacy Text Completions endpoint is gone, along with the HUMAN_PROMPT and AI_PROMPT constants.
OpenAI made Sol, Terra and Luna available in the AWS spec-driven environment on 24 August. Both companies attribute the saving to Kiro handing the model requirements, design documents and task context before generation starts, so it reaches working solutions in fewer iterations. The figure comes from their own joint testing on Terminal-Bench 2.1, not an independent run.
Release 2.1.243 adds per-loop metrics covering run count, total tokens, tokens per run and last run, alongside a modelPricing managed setting that feeds an organisation's own per-model rates into those cost calculations. The same release makes cache duration configurable across sessions through promptCacheTtl and subagentPromptCacheTtl, and lets a modelPicker setting reorder and relabel the model selector. Organisations that block API keys can sign in through the Anthropic Console instead.
