Context Window Review is an independent publication on applied AI: evals, retrieval, agents, latency and cost, written for engineers, PMs and designers shipping AI features.https://sitemoment.com/demo/minimal-context-window-review/enhttps://sitemoment.com/demo/minimal-context-window-review/posts/designing-ai-features-users-forgive/https://sitemoment.com/demo/minimal-context-window-review/posts/designing-ai-features-users-forgive/A design lead on why her team stopped chasing accuracy alone, the edit-rate metric they track instead, and the small interface changes that halved complaints.Wed, 28 Oct 2026 00:00:00 GMTProduct and designAgentsInês Valentehttps://sitemoment.com/demo/minimal-context-window-review/posts/latency-budgets-for-chat-uis/https://sitemoment.com/demo/minimal-context-window-review/posts/latency-budgets-for-chat-uis/Users judge a chat feature in the first second and a half. How to split that budget across network, guardrails, retrieval and the model, and what to show while they wait.Thu, 24 Sep 2026 00:00:00 GMTProduct and designCost and latencyInês Valentehttps://sitemoment.com/demo/minimal-context-window-review/posts/prompt-caching-math/https://sitemoment.com/demo/minimal-context-window-review/posts/prompt-caching-math/Cached prefixes are cheap to read and slightly expensive to write. We work out the break-even hit rate, the traps that keep it at zero, and a prompt layout that helps.Thu, 27 Aug 2026 00:00:00 GMTCost and latencyAgentsTobi Adeyemihttps://sitemoment.com/demo/minimal-context-window-review/posts/agents-need-undo/https://sitemoment.com/demo/minimal-context-window-review/posts/agents-need-undo/An agent that is right 97% of the time still makes dozens of wrong moves a day. Reversibility, not accuracy, makes that tolerable. An argument, with a pattern.Thu, 06 Aug 2026 00:00:00 GMTAgentsProduct and designInês Valentehttps://sitemoment.com/demo/minimal-context-window-review/posts/why-rag-answers-drift/https://sitemoment.com/demo/minimal-context-window-review/posts/why-rag-answers-drift/Over six weeks an internal help assistant got quietly worse with no model or prompt change. The causes were all in the retrieval layer, and all preventable.Thu, 16 Jul 2026 00:00:00 GMTRetrieval and RAGEvalsField reportsMara Ekströmhttps://sitemoment.com/demo/minimal-context-window-review/posts/reading-a-model-card-skeptically/https://sitemoment.com/demo/minimal-context-window-review/posts/reading-a-model-card-skeptically/Model cards are marketing and documentation at once. A practical checklist for separating the claims you can rely on from the ones you need to test yourself.Tue, 30 Jun 2026 00:00:00 GMTSmall and on-deviceEvalsKenji Araihttps://sitemoment.com/demo/minimal-context-window-review/posts/small-models-on-a-laptop-field-test/https://sitemoment.com/demo/minimal-context-window-review/posts/small-models-on-a-laptop-field-test/We ran 3B, 7B and 14B open-weight models on two ordinary laptops to extract invoice fields. Speed, accuracy, memory and battery, measured rather than guessed.Thu, 11 Jun 2026 00:00:00 GMTSmall and on-deviceField reportsCost and latencyKenji Araihttps://sitemoment.com/demo/minimal-context-window-review/posts/the-real-cost-of-long-context/https://sitemoment.com/demo/minimal-context-window-review/posts/the-real-cost-of-long-context/Stuffing 120,000 tokens into every request is easy to build and expensive to run. We price it, time it and compare it against a 6,000-token retrieval setup.Thu, 21 May 2026 00:00:00 GMTCost and latencyRetrieval and RAGTobi Adeyemihttps://sitemoment.com/demo/minimal-context-window-review/posts/what-an-eval-suite-catches/https://sitemoment.com/demo/minimal-context-window-review/posts/what-an-eval-suite-catches/A 240-case eval suite for a ticket summarizer, a cheaper model swap, and an honest accounting of which failures the suite caught and which reached users.Thu, 07 May 2026 00:00:00 GMTEvalsField reportsMara Ekström