Developments that matter in production
What changed, how it works and what it means for the people running reliable systems.
46 stories · 13–24
How we reportA bilingual answer needs to connect the runbook’s condition with the incident’s facts. Aleph Alpha’s Kolibri offers that comparison, with a memory footprint larger than its active-parameter count suggests.
Background reading: AI in productionWhen a desktop app has no useful API, Copilot can interact with its interface. Two similar test records show how to distinguish access, selection and a saved change.
Background reading: AI in productionAn old SDK example and a newer removal notice can both turn up in search. Cloudflare’s beta provides a common route; the useful answer connects the right page to the installed version.
Background reading: AI in productionThe latest timeout says 800 ms. Bigtable’s retained cells can show older values, but the next question is which one the slow instance actually loaded.
Background reading: AI in productionAn assistant can discover a rollback tool without being allowed to run it. API-key protection for MCP discovery adds a check to that first request, while invocation keeps its own rules.
Background reading: AI in productionAn incident update and a historical export can share a queue without needing the same retry policy. The handler still decides whether returned work is wanted or already done.
Background reading: Platform engineeringA silent assistant might be waiting on its model or a log-search tool. Apigee’s separate views help narrow the question, provided you keep the calls and time window straight.
Background reading: ObservabilityThe same request can read a laptop file or a remote copy. Follow a README question through Harness profiles, plugins and model routes to understand the answer it returns.
Background reading: AI in productionStreaming can replace an empty screen with useful opening text. Google API Gateway's preview changes delivery, while gateway and backend deadlines still determine whether the ending arrives.
Background reading: Platform engineeringFifty successful attempts and a 61.1% graded score measure different things. A small defensive parser example shows what a team still needs to find out for itself.
Background reading: AI in productionAn overnight assistant can carry maintenance across conversations. The useful handoff is a patch whose installed version, checks and unfinished decisions the next engineer can understand.
Background reading: AI in productionPrepare the toolchain once, then let each task keep its own work. That useful separation also explains why republishing an environment does not update an unfinished task.
Background reading: Platform engineeringDevelopments since our original reporting, with dated updates and supporting sources.
New model documentation makes retained instructions and thinking-block compatibility part of the compaction migration review.
Original report: Claude compaction: keep the reason the worker is waiting