Hands-on accounts of building real AI systems, from training and quantizing small specialist models to running local LLMs on your own hardware. Expect detailed write-ups where the data, benchmarks, and failures get equal billing with the wins, plus the harder questions coding agents and AI-generated code raise for open source and software business models. The focus is on what actually happens when you ship, not what the marketing decks promise.
open-source, ai-development, code-review, software-maintainership, developer-tools
Open source is more important than ever in the AI era, but the contribution model built on drive-by pull requests is collapsing under AI-generated submissions. The fix is not to close the gates but to redesign how contributions are accepted, verified, and credited.
quantization, local-llms, small-models, ai-development
We shipped a 1.5B router that produced valid JSON 97% of the time. One quantization step took it to two. Here is the day we spent finding out why — and why the rules everyone repeats about quantization were written for models ten times the size. It started with a
training-data, prompt-engineering, local-llms, ai-development
We set out to teach a small model how to route developer prompts. The data taught us that our taxonomy — and our sources — encoded a picture of developers that doesn't exist. It started with a quota. I wanted a small, fast model that could sit in front of
coding-agents, open-source, business-models, ai-development
How a coding agent turned a capped open-source project into its full-featured paid equivalent in under two hours — and what that means for every company betting on artificial scarcity as a revenue model. It started with a monthly bill. I was evaluating a self-hosted communication platform — the
local-llms, gpu, self-hosting, ai-development
What happens when you throw 96 GB of VRAM at open-source models, optimize every last flag, profile CUDA kernels, redesign system prompts, and still end up reaching for a cloud API. It started with a simple premise: why pay per token when you have the hardware? I have four