For frontend, we don't need people anymore: backend, especially complex stuff, LLMs really waste our time by not getting it/going in circles so we just don't really do that anymore as plowing through 100s of lines of code or prs that are wrong, not conform to standards, has libs included we don't need etc. It is pretty useless as agents will just go in circles until it's so messy that it would've been many times easier just writing it with code. And often it's hard to get out besides just rollback.
Does this include UI design? We're finding tools like v0 decent, but nowhere near production design quality. Same for just using Claude or Gemini directly.
We've found that most, if not all, models are extremely bad at writing frontend code. I really hope you all know what you're doing here; you could end up with unmaintainable, incomprehensible AI slop...
I found ChatGPT can write terrible code for the simplest React components, but beautiful code for complex react-three-fiber components. I suppose that is because it was trained on beginner tutorials for basic components and advanced 3d modeling examples for react-three-fiber.
For basic components, I’ve found that by asking for more complexity (e.g. ask it to wrap your nav component in a react context or a custom hook) yields better overall code.
Java was taught far before that: we got taught the first public beta version in the 90s in uni. Thought it was nonsense compared to c, prolog, lisp and Haskell.
Beats at what? Not mission critical multi decade setups that can be easily repeated by hiring someone from postgres consultancy. Oracle handles every use case you throw at it: maybe not optimally, but that is not what you care about at that level anyway. They suck but what are the actual alternatives: not postgres or supabase for most orgs of significant size.
Whats a better way and mobile app? I tried a few but everything is pretty crap. Then a lot languages like Spanish or Portuguese are often the south American ones even though they (including duolingo) say they are not, which means it's fully unusable as no one will take you serious.
Last big outage here in Spain, we got a generator, wind turbine, solar and battery. All cheap Chinese stuff as it doesn't happen often, as a result, this event was barely noticeable to us. Especially the wind turbine did well the entire day and night until the power came back a few hours ago.
Not from traditional auto complete, but I have some LLM 'auto complete'; because the LLM 'saw' so much code during training, there is that magic that you just have a blinking prompt and suddenly it comes up with exactly what you intended out of 'thin air'. Then again, I also very often have that it really comes up with stuff I will never want. But I remember mostly the former cases.
reply