The reason the code is the specification is because people don't take specification seriously. Usually for good reason, but sometimes not. Real, long-lived RFC's can exist, and can have directional and corrective impacts on your LLM generated code. The trick is the RFC needs to be human written and maintained. The moment the org allows the LLM themselves to modify the spec... then yes the spec is no longer useful; or rather, "why" comments in code + a sliver of high level directional / summary content is probably all that is valuable.
A fun example can be having the LLM implement a well known spec (which it cannot edit), and then to use the spec in review to find mistakes and corrections. That's helped it really click for me. All these tools that let LLMs generate spec, even with iterative planning... I haven't found it useful for very long. It's great for building something complex in the very short term (e.g. days). But i haven't found keeping them afterwards to provide any benefit.
Their whole algo is good. I had it work on me recently on facebook marketplace, the only social media outside hn im still using. And its shockingly quick at finding niche things im interested in (somy cameras in a certain range, 30L osprey backpacks, im expecting it to infer my shoe size soon). and its fun. And also revealing in how it can make you want things you dont really need, if youre paying attention. Clearly bad for you. But definitely a strong (sub) product on their otherwise dismal facebook app.
Notably, I read this on both my iphone and mac (LG Ultrafine 5k). It is was enjoyable to read in a fun way on my phone, it did not really bug me as a fun blog font. I was more excited about the theme/design than the content. But on my main screen, which is hi res, it _did_ bug my eyes quite a bit. Interesting.
There is such a thing as difficult concepts that code can express elegantly and words cannot. Its far beyond the simple preferences you allude to. Recursion is the simplest example i guess. Perhaps maths is even more elegant, but im among many a non mathmetician for whom code is the only way ive experienced it thus far.
Especially when you combine and mix those concepts in expressive, simple, intuitive ways. Im just not sure what else is like it that our minds can percieve.
Narration is an obvious one. Theres also a huge risk to less qualified folks and jr employees. ie in most software shops, youll get more mileage giving your best engineers bigger opus budgets and a raise than jr engineers to train. Which means we meed maybe half or less as many, and at least half of that savings is floated up to sr folks and ai companies. That seems rather disruptive?
because you are now getting coded products written in large by people who do not have technical foundations, so the way they interact with and even prompt the model is different. We all know how to fix this scenario; be more specific, or diagnose the abstraction mix ups and straighten those out.
Ask for change A and get unwanted change B happens all the time with bad programmers and tradgedy of the commons (ie poorly architected, no restraint) codebases.
Yeah, a while back I did a small project with a stack I wasn't familiar with, and it was really non-critical. So I decided to vibe code it. The experience was pretty similar to this satirical example. But when I work in areas in which I'm paying attention and understand the stack better, I don't experience this nearly as much
Shutting down the isp doesnt let them read what you've sent. There wouldn't be intensive government efforts to eliminate private encryption if it wasn't effective.
very little intelligence that is the whole problem, really. actual intelligence wont likely nuke the species providing for its existence. But a highly capable sub intelligent model might.
Imagine the most AI pilled company imaginable. Then imagine openAI. Then imagine they are in an existential crisis and that failing may also take (part of) the American economy with it - that much on the line.
Then also remember before Anthropic was a leader, they were mostly derided lab of researchers that left OpenAI because they thought OpenAI didnt take alignment seriously.
idk. it all seems to be playing out as expected. i mean i guess i didnt imagine Trump 2 was at the helm of maybe the only apparatus that could help stop it. Quite a time to be alive.
A fun example can be having the LLM implement a well known spec (which it cannot edit), and then to use the spec in review to find mistakes and corrections. That's helped it really click for me. All these tools that let LLMs generate spec, even with iterative planning... I haven't found it useful for very long. It's great for building something complex in the very short term (e.g. days). But i haven't found keeping them afterwards to provide any benefit.
reply