Engineering notes, architecture decisions, and build logs from the team at Ode.https://www.ode.com/engineeringenhttps://www.ode.com/engineering/evals-for-document-extractionhttps://www.ode.com/engineering/evals-for-document-extractionLearn how to evaluate LLM-powered PDF extraction using exhaustive ground truth, synthetic documents, partial labels, and output validation.Thu, 03 Sep 2026 00:00:00 GMThttps://www.ode.com/engineering/stop-designing-chatbotshttps://www.ode.com/engineering/stop-designing-chatbotsOn defaulting to chat as the interface for AI products — and why it's holding you back.Mon, 30 Mar 2026 00:00:00 GMThttps://www.ode.com/engineering/artifact-driven-ai-creationhttps://www.ode.com/engineering/artifact-driven-ai-creationHow artifact-based thinking makes AI-assisted creation precise and cumulative.Fri, 27 Mar 2026 00:00:00 GMThttps://www.ode.com/engineering/if-i-could-turn-back-timehttps://www.ode.com/engineering/if-i-could-turn-back-timeHave you ever exclaimed after a few hours engaged in some fruitless activity, "Well, that's X hours of my life I'll never get back"?Thu, 19 Feb 2026 00:00:00 GMThttps://www.ode.com/engineering/your-evals-have-a-rotten-tomatoes-problemhttps://www.ode.com/engineering/your-evals-have-a-rotten-tomatoes-problemA single eval score tells you that something regressed, not what regressed. Scoring outputs along specific dimensions — as close to each individual LLM call as you can get — turns an unhelpful number into a diagnosis.Thu, 12 Feb 2026 21:30:09 GMThttps://www.ode.com/engineering/the-virtues-of-showing-your-work-do-llm-explanations-actually-helphttps://www.ode.com/engineering/the-virtues-of-showing-your-work-do-llm-explanations-actually-helpIf you've spent any time prompting large language models, you've probably heard of Chain of Thought (CoT) reasoning—the technique of asking an LLM to...Thu, 05 Feb 2026 00:00:00 GMThttps://www.ode.com/engineering/do-what-i-meanhttps://www.ode.com/engineering/do-what-i-meanTests are crucial to reliable code. I’ve found I do a better job writing tests if I think of them as specifications and write them before I write code.Wed, 05 Feb 2025 00:00:00 GMThttps://www.ode.com/engineering/using-historical-codebase-to-build-an-eval-suitehttps://www.ode.com/engineering/using-historical-codebase-to-build-an-eval-suiteHow thousands of pull requests let Ode automate annotating a ground truth set for a new product.Mon, 11 Nov 2024 00:00:00 GMThttps://www.ode.com/engineering/fine-tuning-quirkhttps://www.ode.com/engineering/fine-tuning-quirkI was working on content moderation for Change.org. Our goal was to automatically identify violative content to flag for reviewers.Tue, 27 Aug 2024 00:00:00 GMThttps://www.ode.com/engineering/insisting-on-known-knowns-using-evaluators-to-drive-reliabilityhttps://www.ode.com/engineering/insisting-on-known-knowns-using-evaluators-to-drive-reliabilityLearn how evaluators can enhance the accuracy and reliability of LLM-driven GenAI projects through rigorous correctness measurement techniques.Thu, 08 Aug 2024 00:00:00 GMThttps://www.ode.com/engineering/customizing-cursors-code-generation-and-personalityhttps://www.ode.com/engineering/customizing-cursors-code-generation-and-personalityOne of our rules at Ode is “overuse AI”. This is a bit tongue-in-cheek, but what we mean by it is that even within our own work, we occasionally suspend...Sun, 29 Oct 2023 00:00:00 GMT