10 Posts
Tag: Ai Agents
BY eric
Jul 09, 2026
The Model That Aced Olympiad Math Can't Write a Tweet
The 4 GB series found a 3B that reasons like a giant. So surely it can write? It can't — and how it fails is the interesting part. VibeThinker-3B chokes on a 120-word story while two plain instruct models a fraction as clever nail it, and a language twist decides which one you want.
BY eric
Jun 17, 2026
The 4 GB Card You Already Own Can Reason Now
The dispatch experiment ended on a question: route the hard tail to a more capable model — but which one, if you do not want to reach for the cloud? VibeThinker-3B, a 3B reasoning specialist, runs entirely on the 4 GB laptop GPU you probably already own, solves competition math the workhorse cannot, and sips about 50 watts doing it.
BY eric
Jun 15, 2026
The Small Model's Real Job Is Dispatch, Not Work
We pushed a 4 GB laptop model past coding into reasoning, extraction and writing to find its limit — then measured a better use for it entirely: not doing the work, but routing it. A reasoning-tuned 1.7B model covers 90% of a general workload, and a perfect router is 74% cheaper at equal quality.
BY eric
Jun 14, 2026
What a Native Tool Call Actually Is
An LLM only ever outputs text — so how does an AI agent run a tool? The answer is the native tool call, a precise, trained, parseable protocol that most people hand-wave over. Here is exactly what it is, what happens on the wire, and why a small model that knows what to do still cannot make one.
BY eric
Jun 14, 2026
The 4 GB Experiment: What Actually Makes a Small Model a Good Coding Agent
We stopped theorizing and ran ~900 trials on a 4 GB laptop GPU to measure what makes a small local model a good coding agent. The result overturns intuition: the edit format and the editor model decide almost everything, the clever agentic scaffolding mostly hurts, and the simplest setup won.
BY eric
Jun 13, 2026
How a Coding Agent Actually Writes Code
The first four posts in this series took the telescope view of AI agents. This one takes the microscope: how a coding agent turns an instruction into a working, tested change — how it edits, compiles, runs, tests and debugs — and why Claude Code behaves like a native *nix programmer while others reach for a throwaway script.
BY eric
Jun 13, 2026
The Coding Agent Shootout and the Best Harness We Could Not Read
After reading the source of five open coding agents — OpenAI's Codex, Pi, OpenCode, Kimi Code and Gemini CLI — here is how they actually stack up, strengths and weaknesses, which one wins, and why the best harness of all belongs to a tool we deliberately left out of the code study: Claude Code.
BY eric
Jun 13, 2026
Skills and Memory: How AI Agents Actually Learn
Self-improving AI agents sound like marketing — until you read the code. This is how agents capture knowledge as skills and memory, why a human still has to write some of it down, and what self-improving actually means once you strip away the hype: bookkeeping, an LLM review pass, and a human gate.
BY eric
Jun 13, 2026
What an AI Agent Harness Actually Does
We read the source code of six AI agent harnesses — Pi, Hermes, OpenCode, Kimi Code, OpenAI's Codex CLI and Google's Gemini CLI. This is what the harness really does for the model, why almost every agent failure is a harness failure, and why the engineering lives there, not in the model.
BY eric
Jun 13, 2026
Codex, Pi and Hermes: Three Agents, Three Species
Comparing OpenAI Codex, Pi (pi.dev) and Hermes Agent by Nous Research — an odd trio at first glance, but their differences map out the whole landscape of AI agents: what each type is, what it can do, and which one can help you.