Setup. Same project as the broken MMLU questions: before trusting any benchmark number, check what it is actually measuring. That post was about the data layer. This one is about the harness layer:...

Setup. MMLU is one of the standard benchmarks LLMs are scored on. While doing an error analysis on its physics subjects, every “the model got this wrong” first had to survive a check the benchmark ...

Updated 2026-07-01: batched inference, the scaled 16-item sweep, and a fourth defective MMLU item (r80). The updates are marked inline. Setup. Gemma-4-E2B (small) vs Qwen3-32B (reasoning model), o...

Okay, quick test. You know how a bicycle works. You’ve seen thousands of them, you can probably ride one, and if I handed you a pen you’d be pretty sure you could sketch one from memory: two wheels...

Most of what I ship now starts as a draft a coding agent wrote from a spec I gave it: a planning doc for a project, a proposal, a blog post, a design note. The agent is good at the first pass, but ...

I spend a lot of my time on one question: do language models know what they know? Can a model tell the difference between a fact it holds and a guess it is about to make? It is a practical question...

This one is for Dareen. If you are curious about how web development works, what shows up in your browser, the code that powers it, and where the data lives, you are in the right place. We will bui...

Taming Your Gmail Inbox: Automating Unread Email Management Are you tired of logging into your Gmail account only to be greeted by an overwhelming number of unread emails? Perhaps they’re updates ...

Project URL https://github.com/ranawaqas-ai/local-llm Ollama Docker Compose Setup Welcome to the Ollama Docker Compose Setup! This project simplifies the deployment of Ollama using Docker Compos...