Writing · Nº 041065 essays
The writing
Ideas from inside the work. Explore technology, entrepreneurship, and the practice of building things that last.
45 essays found · Page 2 of 4
Clear filtersEngineeringHow to Design a Product Around a ModelProduct design when the model is the core mechanism: shape the workflow around what it does reliably, design for non-deterministic output, and build the recovery paths.ThesisBolted-On AI Is Fine for Internal ToolsAI-native purity is for products you sell. For internal tools, bolted-on AI is often the right call. When a quick model integration beats a proper AI-native rebuild.StrategyWhen Bolted-On AI Is Actually the Right CallAI-native is not always the answer. When bolted-on AI is the right call, when a wrapper is enough, and how to tell if your product really needs a rebuild.ThesisAI-Native Legal Research Answers, Boolean Search Returns DocumentsAI-native legal research answers the legal question. Boolean search returns a list of cases to read. Here is why answering beats retrieving for real practice.StrategyWhich Feature to Put AI In First When RetrofittingRetrofitting AI into an existing product? Do not start with the flashiest feature. How to pick the first AI feature by value, tolerance for error, and data you already have.ThesisAI-Native UX Is Not Just Adding a ChatbotAI-native UX is not a chatbot bolted onto your app. Why a chat box is the laziest AI interface and what designing the experience around the model really looks like.ThesisThe AI-Native CRM Updates Itself, the Bolted-On One Asks You ToAn AI-native CRM updates the record from the conversation. A bolted-on CRM still makes reps log everything by hand. Here is why that gap decides adoption.EngineeringFall Back to Deterministic When the AI Is UnsureThe strongest AI-native systems keep a deterministic path underneath the model. When to route to rules instead of the model, and why a hybrid beats pure AI in production.EngineeringThe Technical Debt Hiding in Bolted-On AIBolting AI onto old software creates technical debt that compounds fast. Where the debt hides, why it grows, and what it costs when you finally have to rebuild.ThesisTrust Is a Feature, Not a SloganTrust is a feature you engineer, not a slogan you print. Here is how to build trust into an AI product with logs, limits, and proof buyers can verify.ThesisHaving a Data Moat Does Not Make You AI-NativeProprietary data is not the same as being AI-native. Here is why a data moat means nothing if your product workflow was still built around forms and fields.AIDesign the Fallback Before the Model Fails YouThe model will fail, time out, or return garbage. AI-native failure design means building the fallback path first, so a bad model call degrades instead of breaking.