If you need to fine-tune a model for your codebase, your code is bad. Agree or disagree? We talk AI hot takes in our latest GitHub Podcast episode. What would you add to the list? Listen wherever you listen to podcasts or find it here

nah, plenty of good codebases just have conventions no public model has ever been trained on

I'd first separate missing documentation, domain-specific rules and actual code complexity. Needing more context doesn't, by itself, tell us which of those problems we're solving.

This is definitely not the case. Good codebases are not easy to come by so just yoloing will statistically not give you a good codebase. Quite frankly I am disappointed in the side you guys are talking. But..... To each his own.

Your “finished” code is simply an input to LLMs and neural networks. If fine tuning was useless; then 🦥. Take note @UnslothAI

Disagree as a blanket rule. When is fine-tuning actually better than retrieval, repo instructions, and evals for a codebase-specific task?

Disagree. Needing to fine-tune is a scale signal, not a quality one. A perfectly clean codebase larger than the context window still cannot be reasoned about in one shot, and fine-tuning is one way to compress that corpus into weights.

A fine-tune is a fork of your codebase. Rename a module, move a boundary, refactor an API, and you now own a second artifact whose training data silently drifts out of date. Your code was never the problem.

Disagree. Fine-tuning can encode stable local conventions, but it is the wrong fix for missing architecture. If the agent needs weights updated to learn who owns a module, which APIs are forbidden, or why a migration is half-finished, the codebase is missing machine-readable boundaries and decisions.

The take confuses "the model never saw our code" with "our code is bad." Plenty of clean repos speak a dialect no base model was paid to learn. Fine-tuning is a distribution problem, not a shame signal.

Mostly disagree. A specialized codebase can need domain adaptation even when the architecture is clean.

Preach! 😂 If you need to fine-tune a model, your code is due for a serious intervention 🚨

every codebase looks bad to whoever didn't write it. models included.

Interesting debate. Good software design should reduce the need for heavy customization, but fine-tuning can still be valuable for adapting models to specific domains, workflows, and unique codebases.

感觉这类讨论最有价值的地方，是把“代码会不会被生成”拉回到“团队怎么验证和维护”。没有测试、审查和清晰边界，生成得快反而只是把返工提前。

Agree, as good code can still carry years of business decisions no prompt has seen.