# Acorn Labs

> A small studio for rapid prototyping, discovery, and taking agentic AI from design to production — with deep software platform, machine learning, and product experience.

Acorn Labs partners with teams to move quickly from idea to working software, with a focus on agentic AI, applied machine learning, and production-grade systems. We work end-to-end: discovery, design, prototyping, and shipping.

## Pages

- [Home](/): What we do, our approach, and how to get in touch.

## Field notes

- [The best AI agent framework is the one your on-call can debug at 3am](/notes/agent-framework-comparison-guide): A production-lens comparison of CrewAI, AutoGen, LangGraph, and plain function calling.
- [The case for a 3B model on your desk](/notes/3b-model-on-your-desk): Why small local models often beat frontier APIs for narrow product jobs.
- [Refusals are a product surface, not a safety afterthought](/notes/refusals-are-a-product-surface): Treating model refusals as design moments instead of policy enforcement.
- [Why every editor we work with hates the word 'leverage'](/notes/leverage-and-other-words-editors-hate): The vocabulary tells of LLM drafts, and how to fix tone.
- [Evaluating tone is harder than evaluating truth](/notes/evaluating-tone-is-harder-than-truth): Why tone evals need pairwise comparison, rubrics, and calibrated judges.