--- format: "aidr-story-markdown/v1" id: "9f6c4c31bcb05ee2212d7ab130a7c7696a19bf27b50bf5567d383d3edf668bb3" canonical_url: "https://aidr.today/9f6c4c31?lang=en" title: "Astra Resets System and Fixes Bugs After Context Experiment Hits 5,000 Users" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-09-12T05:48:38.000Z" category: "Products" topics: ["astra","llm","inference","product"] source_urls: ["https://huggingnews.com/ai/astra-resets-system-and-fixes-bugs-after-context-experiment-hits-5000-us-32206e0b","https://x.com/thsottiaux/status/2098612714704891959"] summary: "The removal of badly configured engines and outdated model skills targets quality degradations reported by Astra users. Astra disabled an opt-in context management experiment that caused early stops or replies to older messages for 4,000 to 5,000 users and scheduled a system reset for midnight today. Thomas Sottiaux stated the modifications address skills that triggered too frequently and a \"long tail of traffic\" affected by engine misconfigurations. The update introduces more consistent follow-through, better tracking of the latest user messages, and improved internal checks on the model's output to resolve recent performance issues." --- # Astra Resets System and Fixes Bugs After Context Experiment Hits 5,000 Users > [Open the canonical story]() **Published:** 2026-09-12T05:48:38.000Z **Category:** Products **Topics:** astra, llm, inference, product ## Summary The removal of badly configured engines and outdated model skills targets quality degradations reported by Astra users\. Astra disabled an opt\-in context management experiment that caused early stops or replies to older messages for 4,000 to 5,000 users and scheduled a system reset for midnight today\. Thomas Sottiaux stated the modifications address skills that triggered too frequently and a "long tail of traffic" affected by engine misconfigurations\. The update introduces more consistent follow\-through, better tracking of the latest user messages, and improved internal checks on the model's output to resolve recent performance issues\. ## Sources - [Story source]() - [Story source]()