I'm in a #github internal group for high-profile FOSS projects (due to @leaflet having a few kilo-stars), and the second most-wanted feature is "plz allow us to disable copilot reviews", with the most-wanted feature being "plz allow us to block issues/PRs made with copilot".

Meanwhile, there's a grand total of zero requests for "plz put copilot in more stuff".

This should be significative of the attitude of veteran coders towards #LLM creep.

@IvanSanchez @leaflet …just wondering, do you see added value for AI in coding? Or not?
Study finds AI tools made open source software developers 19 percent slower

Coders spent more time prompting and reviewing AI generations than they saved on coding.

Ars Technica
@IvanSanchez ...for the total inexperienced programmer (me) that is surely not true, much better value , also as a tool to explain code for me, it is also useful.
@IvanSanchez ..also if you want to make machinereadable code from laws/regulations, Claude can do amazing things I have witnessed , in Dutch GitHub - MinBZK/poc-machine-law: Proof of Concept RegelRecht https://github.com/MinBZK/poc-machine-law
GitHub - MinBZK/poc-machine-law: Proof of Concept RegelRecht

Proof of Concept RegelRecht . Contribute to MinBZK/poc-machine-law development by creating an account on GitHub.

GitHub

@ErikJonker @IvanSanchez While it is impressive that Claude can (partly) automate this work, I think this type of task *shouldn't* be automated. I don't trust law code that was written by a word-guessing machine of which the inner workings cannot be fully traced. Human experts could verify all the code, but then they are spending more time than if they were writing the code themselves.

I also find it shocking that my government is using a theft-based, power-hungry tool for this purpose.

@ErikJonker Talk to @jdelacueva about project Kelsen. This stuff (and its limitations) are not new.

@ErikJonker @IvanSanchez Useful and easy sure, but you end up learning less and you will stay dependent on the (energy-hungry and theft-based) tool unless you learn to do it all by yourself.

https://www.researchgate.net/publication/392560878_Your_Brain_on_ChatGPT_Accumulation_of_Cognitive_Debt_when_Using_an_AI_Assistant_for_Essay_Writing_Task

You can learn to do it the hard way, millions of people have done it before you. Once you've done that, the AI tool will be something that slows you down and reduces the quality of your work. Steeper learning curve but greater rewards and more autonomy.

@ErikJonker
Read the fucking paper. LLMs give users *the impression* of being useful while adding negative value. There are powerful biases at work.