A sort-of followup to my "Show Me The Pull Requests" post (https://lnkd.in/dWpx6RrK) | Ben Evans

A sort-of followup to my "Show Me The Pull Requests" post (https://lnkd.in/dWpx6RrK) Sine posting that, people sometimes send me PRs as data points, just like I asked. And special thanks to Simon Willison for sending me this one! It's a compiler optimization for LLVM regarding the very-common ASN.1 data representation: https://lnkd.in/dHmCsEKT Alex, who wrote this PR, has a great write-up about it: https://lnkd.in/dHfhxQPU My biggest takeaways from this is that we are starting to see data points emerge for my initial challenge - that there are ways to use these tools to develop new capabilities. But there is a great deal more caution and guarded language from folks like Alex (who actually works at Anthropic) and Simon as compared to the LinkedIn hype and froth that gets posted every day. This is still very much a human-in-the-loop force multiplier as driven by an expert-level user. I would describe it as being more akin to using a REPL to speed up development than a wholesale replacement of the toolchain. My own experiments with it tend to align with this - I am getting some utility out of it but it needs much more insight and verification from the user than some people would have you believe, and the gains are force-multiplying, not game changing. I'm also not paying for any of the tools I'm using. I remain extremely skeptical about how much money there is actually to be made in this space and whether the current pricing is sustainable in the long-term. There also remain major questions about the ethics and resource utilization involved.

www.linkedin.com