/log
The InferHaven log.
Engineering notes, design decisions, and the occasional shipping party. Long-form, infrequent, unfiltered.
-
How I built a tutor that refuses to write your code
The never-write rule sounds like a prompt. It turned into a grading contract, two execution paths, and a pile of things that broke in production. Here is the whole build.
#CodeTrain#AI coding#Pyodide#sandboxing#prompt design#engineering read → -
CodeTrain: the AI that won't write your code
An AI trainer that teaches you on your own codebase and never writes the code for you. Free, no card, live today.
#CodeTrain#AI coding#training#coding bootcamp#socratic#learning read → -
Privacy as the baseline, not the upsell
Local models on hardware you control are how you build AI workflows that don't break when a vendor does. Here's the case for local-first development, and the mission InferHaven ships against: make privacy and security the default, not a feature you pay extra to unlock.
#local LLMs#privacy#mission#self-hosting read → -
Building haven bench in the open, and the flaky CI ghost it flushed out
I shipped a tokens/sec benchmark for local models. The unit tests were green, and then CI turned red in a way that looked like my fault. It wasn't. Here's the whole hunt: a chown that raced the tide, set -e, and a zsh lock file that vanished mid-sweep.
#build in the open#debugging#ci#local LLMs read → -
InferHaven's source code is live!
Today inferhaven-core goes public on GitHub under FSL-1.1-ALv2. Why we picked fair source instead of MIT or Apache, what that means for you in practice, and what comes next on the road to InferHaven Cloud.
#release#licensing#fair source#roadmap read → -
The private AI coding box: why not just use XYZ?
A common question is why run InferHaven instead of Ollama, or Continue, or Tabby, or a cloud IDE? Most of those things run on InferHaven. Here's the honest map of what we are, what we aren't, and when you should pick something else.
#local LLMs#comparison#self-hosting read → -
Why InferHaven runs your AI offline by default
Using a cloud API sends your repo to someone else. InferHaven flips that default: your code stays local, and the network is an opt-in, not an opt-out.
#local LLMs#security#privacy read →