InITNEXTbyDavid Gilbertson·Jun 25Using a behaviour index to help AI coding agents explore your codeThis article introduces the ‘behaviour index’, a document that helps AI coding agents and humans navigate your application.A response icon1A response icon1
InITNEXTbyDavid Gilbertson·May 13Comparing AI Providers: OpenAI, Google, Anthropic, and xAIWhen comparing AI providers, the headlines tend to focus on benchmark scores. But when actually building applications that use AI, there’s…
InITNEXTbyDavid Gilbertson·May 6A system to keep you informedI’d like to tell you about a system I’ve built. I find it very useful, and I’d like to know if you’d find it useful too.A response icon2A response icon2
David Gilbertson·Apr 6For AI coding agents, not all file search methods are equalI wanted to know if there was any meaningful difference between the different ways an AI coding agent can explore a codebase.A response icon1A response icon1
InITNEXTbyDavid Gilbertson·Jan 28Shrinking a language detection model to under 10 KBTL;DR: AI models are usually too large to be sent to a user’s device, but for some tasks they can be made surprisingly small.A response icon1A response icon1
InITNEXTbyDavid Gilbertson·Dec 4, 2025Cloudflare vs Firebase: which is best for side projects?In this post, I’m going to compare Cloudflare and Firebase, to see which is the best choice for a specific niche: a developer’s collection…A response icon3A response icon3
InITNEXTbyDavid Gilbertson·Nov 25, 2025How to read an LLM’s mind; it’s easier than you thinkWhen a Large Language Model is responding to input, we can tap into its ‘inner thoughts’ to know what the model’s thinking about, before it…A response icon3A response icon3
InITNEXTbyDavid Gilbertson·Nov 16, 2025How to run a background service in WSLSo, you’re using WSL and you’ve got something you want to run in the background at startup. Maybe you want to serve a website, or spin up…
InITNEXTbyDavid Gilbertson·Sep 25, 2025A practical approach to AI safetyYou may have heard that AI can perform real-world actions, but can be misaligned (e.g. deceptive), so it poses new and unique risks.A response icon4A response icon4
InITNEXTbyDavid Gilbertson·May 14, 2025How I found mistakes in OpenAI’s HealthBench using AII was quite excited when I saw that OpenAI had released a benchmark for health. Medicine is so vast and complex that humans — with their…A response icon3A response icon3
