borplk

The idea that users "make their own tool with AI" is a nerd fantasy that does not exist in the real world. Some use cases get absorbed into products like ChatGPT because Jim no longer needs a special web-based tool/calculator/product and can just ask ChatGPT to do it for him. But normies can't and don't want to build or maintain products. They wouldn't even have the patience to write the prompt to ask the AI to build it for them. Vibe-coders are mainly wannabe programmers and nerd hobbyists.

If you ask AI to build product "A" it will say ok and build it for you. If you then ask it to build product "B" which is in a similar domain, but total polar opposite of "A", it will also build it for you. So ultimately you have to decide what you want it to do for you, what to build, how to build it, what not to build, etc... Different people attempting to build the same product with AI will still produce wildly different products. In that sense, competition and differentiation will never go away.

Here's my simple theory. Increasing the level of quality is not rewarded and often actively discouraged. Almost all software is stuck at the "barely tolerable minimum level of quality". The quality level remains constant (or goes down) regardless of any productivity gain. Any and all productivity gain is spent for expanding scope, chasing more features, etc... Volume of work is increased so that the development team is perpetually understaffed. I have never once seen a team that had what you could remotely call a healthy or appropriate level of staffing. It has always been "the absolute minimum we can possibly get away with and never more". This leaves zero room for any kind of maintenance or quality improvement. Every day is a fight just to keep the lights on. I have seen people queitly sneaking-in maintenance work because that was the only way it could be done.

0.7M USD annual revenue for Jetbrains? That can't be right. Their annual revenue is 400M+ USD from my quick search.

I guess you'll have to do it with browser automation. I assume that's possible?

This is cool but personally I never really understood the interest in turning a barcode into a piece of art or trying to embed images and logos into it.

I'll take the 300x slower non-vibe-coded pg, thanks!

If it's good enough to replace plain Software Engineers it's good enough to do DevOps/SRE too. There's no meaningful distinction between those categories in this context.

Gimmickmaxxing?

I don't remember 2001. But 2008 tech job market was way wayyy better than today.

Unfortunately nerds are easy to trick. Good luck convincing any other discipline to do free labour for optional donations, vibes, and "the community".

Can anyone suggest or come up with viable "use cases" of a custom LLM like this? I wouldn't mind giving it a try but ideally I'm looking for something that is not just a toy.

That doesn't justify it because the user should still be able to decide when to initiate it. People are ok with not interrupting the update if they get to choose when it starts.

I find it shocking (not really) that among the many BILLION dollar companies built on the back of Postgres there isn't enough sense to pay the salary of one dude to keep a project like this going forever.

Site appears to be down intermittently with a Django error

If author sees this: Turn off Django debug mode