These models generally are not capable enough to do one-shot vibe coding. They are pretty good as coding assistants if you tell them exactly what you want and let them focus on a specific aspect/part of code, not the whole thing at once.

Using an agent framework (I like Kilo on VSCode but there are many others) you can start with a planning session to let the model find out what you want to build. Then you let it write that gist into AGENTS.md and double check if that is what you want. AGENTS.md will be loaded into the context automatically so the model has a solid base of understanding for everything you do afterwards. Once you have that, building in vertical slices on top of that skeleton is much easier. Another neat trick is to ask the model a few questions about the current code base (if there is any…) at the beginning of a session, e.g. "How does feature x work in the current code? ". This primes the model for what you are about to do. All this is obviously a bit more work than just vibe coding away, but it lets you keep in control of the code and helps in being alert for errors these models (and all LLMs in general) will inevitably produce.

You could use the 35B MoE model, tune it a little bit and get much better results. I have a 5060 ti and 70-80 tok/s are the norm

I switched to a US keyboard with EURkeys keyboard layout (alt-u for ü etc) and I’m very happy with it. Common keys for programming like backslash, square/curly brackets, pipe, tilde and of course the backticks are just horrible to use on a German keyboard.

An air conditioning device that needs no pipe to the outside you say? It defeats the laws of thermodynamics for only 99€? Shut up and take my money!

/s

Yeah, I can relate. I’ve been doing consultant work and I always tried to avoid being pushed into the “architect” role. My background for over 25 years was software development and we had a lot of success with agile methods (doing it right is hard but viable and it produces quality software) My experience is that a good dev team does not need an “architect” to tells them what’s best. But all of this is now gone anyway, or at least taking an extended break. At the moment no one is investing in development teams, and the prospect of being able to fire all the developers because AI can do their job now is making CEOs giddy everywhere. Not going to happen (and this should be obvious for anyone who can judge the quality of code or the effectiveness of processes in terms of reliability, quality, cost, performance etc.) but that doesn’t stop them from trying.

Interesting perspective. Software development works different where I come from, the dev part is still very much in control of the developers themselves. However the management lives in delusionville and the expectations are pretty much insane compared to the reality of the actual turnout. My comment above was more about educating the general public… AI results can look pretty compelling unless you decide to have a closer look, and a lot of non-dev people are falling for it. Either because they are dumb and gullible, lack analytic thinking or because they have no real contact with AI and are just exposed to the hype through the media.

Or educate the people around you that “AI” is not the magical fairy dust machine that can do anything imaginable. Might be better than giving them any money at all.

There is a crucial difference though, vibe coding enables people who are neither interested nor capable of understanding what is involved in maintaining a piece of software. Before vibe coding it didn’t even come to this situation: if you can’t write software you don’t need to maintain it. While there were bad coders and bad maintainers before, the numbers have now increased dramatically because everyone can pretend to be a software dev these days.

Just Imagine, some day he will not wake up any more.

If you do, at the very least negotiate a much higher salary. They will be desperate, make it hurt them.

“Mark Zuckerberg is realizing there’s a limit to ruthless efficiency”.

Cost cutting is not synonymous with efficiency. It’s sad that a business focused magazine doesn’t see through the bullshit.

• Losing 70+ billion on the failed VR experiment and laying off workers related to VR has zero to do with efficiency. It’s just stopping losses without having any other positive effects. No efficiency gain here because the whole thing is mothballed.
• Putting AI bros in decision making positions and demoting highly payed engineers who actually know how the platform works to AI training babysitters: Not very efficient so far because the fuck ups are getting costly quickly.

There is more, but why should I give a fuck about Meta. Let this dumpster fire burn.