AI has already changed everything about how I write code, think about code, budget for code, debug and refactor code, etc. If you're doing that, too, I know the HN audience would welcome hearing from you.

My stuff is niche on my personal blog[0] (indie game dev), but I write about bigger AI topics on Medium[1]. All of this stuff has got to eventually filter down to individual use cases, but perhaps the philosophy of software is changing?

Is it too soon to call out some early victims? I don't think we need any more new languages or data formats. Ever. Text is going to be MD so there go proprietary text markup formats. Data is JSON, full stop. (Speaking for web apps here).

Also going away are endless tweaks to UI frameworks as customers (dev and non-dev) will be "drawing" and defining their own dynamic UIs through the prompts they write. On the backside, those prompts will turn into multiple tool calls to perhaps unexplored APIs via MCP -- a complete inversion of the client-server control model we're used to.

If there's a better place to write or read about this stuff, I'd love to hear about it, too!

[0] https://davidbethune.com/blog

[1] https://medium.com/@mimixco/list/ai-library-e2fcb2e18159

The basic process is fetching https://hnrss.org/newest then piping each item through gpt-4.1-mini with just the URL and title (I think fetching each actual page is overkill and a bit greedy) and asking it to figure out if something is likely to be development related. If it's a confident yes, that item then gets into the final feed. It cuts the firehose down quite a bit just with that simple approach.