Ask HN: Why would this be a good idea?

https://news.ycombinator.com/item?id=47418297

by ZLStas • 4 months ago

1 5 4 months ago

news.ycombinator.com

I packaged programming books (Clean Code, DDIA, Effective Java, etc.) into Claude Code skills that apply each book's practices when reviewing or generating code.

  https://github.com/booklib-ai/skills

  Why would this actually be useful?

bediger40002• 4 months ago

it would not be useful. Even a large number of programming books will not overcome the weighting of all the terrible source code and pitifully inadequate and often wrong explanations of programming topics that an LLM has.

Beyond that, all the LLM will do is recommend best practices. Sure, it's great to keep best practices in mind, but you have to realize that best practices really lags current practice, and do not even come close to excellent programming. Religiously follow "best practices" without care and you will have garbage programming. I give you the example of large numbers of website asking you to create a password with really exotic character combinations. In an HTML "password" field that shows you dots, so you're doing it blind. Twice, to make sure you typed what you thought. It's bullshit, and it's a "best practice". There's also no incentive to rid ourselves of no longer relevant, or mistaken, best practices. Something screws up when you didn't follow a mistaken best practice? You're at fault for not following a best practice, no matter how horrible.

ZLStas2• 4 months ago

Fair point that blind rule-following is dangerous — but that's true of any checklist, not specific to this tool. The goal isn't religious compliance, it's surfacing relevant principles at the right moment so the developer can make an informed decision to follow or consciously break them. Also — if LLM weighting on bad code is the core problem, isn't that an argument against using LLMs for coding altogether? Yet here we are, and they're useful anyway.

bediger40002• 4 months ago

Your heart is pure. But people are going to "let the AI do it", just like people let calculators do it in the 1970s, or let computers do it in the 1990s. An LLM with a very small additional bit of weight is still probabilistic, and the weights favor "best practices" and median program texts. You're encouraging blind rule following

My point is that LLMs weights are full of misinformation, misconception and median code. Putting a few books in isn't going to change that.

Ask HN: Why would this be a good idea?

Related Stories