[Brendinooo]

>e.g. front-end web development

It's kind of funny that you say this, because I am a frontend developer and I tend to see the state of the art as being very good at doing the boring behind-the-scenes plumbing that I don't care about, and not great at doing the kind of bespoke design work that my day job's clients want.

I'm not saying that either of us are definitively right or wrong, and I agree that having a more generalist skillset is probably the best way to succeed in this new era; I'm just pointing out that LLMs don't really own any part of the stack so thoroughly that specialists in that segment will just go away.

[tracerbulletx]

Yeah everyone thinks its great at the things they don't personally know that much about or appreciate and I think it's kind of embarrassing to proclaim its going to do someone else's speciality great but not theirs and just reveals an underlying ignorance.

[technomoloch]

I'm solidly a generalist, I custom create designs for products and implement them, and also work on backends and large scale production ml systems. I would actually have to agree with the person you are replying to - simply because backends are text problems, (an LLMs domain), and frontends are visual - llms are just still not quite as good at seeing details as humans, whereas they can scan a large codebase for possible problems much faster than any human could. Both areas need careful supervision and feedback loops, adversarial reviews etc - but for the frontend, I find myself having to do much more manual work actually checking myself, because an llm just doesn't get symmetry if it doesn't perfectly correlate to margin being 16px on both sides of a box, etc, ie symmetry you can see in code. Or whether a design "feels" nice to look at, etc. But with backend, you give it proper guidance to create tests, do benchmarking, follow sane design patterns, etc, and its very effective.

[JimDabell]

I’ve been writing front-end code for more than a quarter of a century. I can assure you I’m not saying this because I’m unfamiliar with the field.

[jaggederest]

Honestly I think it's great at the things I do know about. I've been doing this stuff soup to nuts since javascript was released, and it's tactically better across the board - presentation, ux, frontend ui, api, backend, databases, even systems and devops.

It's taste can be atrocious, so we're not replacing engineers entirely yet, but it's clear that it's almost hands off for any task I would have done as a consultant in 2012, for example. And, contrary to my opinion a couple months ago, I think taste is a pretty shallow moat, ultimately. Many of my clients when I was operating a consultancy had plenty of taste, if that's all they required, and I think it'd be foolish to assume frontier models won't acquire taste eventually.

I do think that, ultimately, the tippy top of the pinnacle for things like truly original design work, truly original work of any kind, will take a long time to replace. But most software engineering isn't moving the boundaries of the possibilities of humanity, it's making sure that we can turn $0.10 of infra spend into $2 of revenue reliably.

[tripleee]

I feel people saying this are imagining front-end web dev is all just HTML tables and centering divs. AI does better with back-end in my experience

[sdellis]

I agree with you. Maybe AI can generate designs that look moderately good and aesthetically pleasing for UIs that solve known problems in a prototypical way.[0] That is useful, especially for simple utilities, internal tools, or hobby projects. However, I have yet to see AI solve new design problems, improve on old problems, and create a unique design style that defines a brand and separates it from competition.

[0] https://research.google/pubs/the-role-of-visual-complexity-a...

[JimDabell]

Most organisations need a high quality design system and to be able to assemble pages from it. Totally bespoke design will start with the designer using an agent. So the way I see it, the everyday stuff will be the first to go, followed by designers taking the rest of it. Front-end developers have their area eroded from the bottom then the top.

As far as frontend vs backend, there’s a greater scope for fuckups when dealing with the backend. Frontend problems tend to be more transient. So the stakes are lower, which means that the accountability of humans has less value.

[nec4b]

The frontend feedback is much harder for AI. AI needs to be able to see the page with all the moving parts and sounds and have a bit of sense for the layout and taste in colors etc. I think it much harder for AI to recognize what feels good for the human user. Backend is devoid of that personal touch. By definition you are building for the machine. You have APIs, databases and infrastructure which is logical and easy for LLM to interact with. "All it needs" is curl and it can run in a loop the whole day improving backend software.

[avgDev]

Codex has been great for me for backend wiring, mapping and creating boiler plate code in C#. However, it seems when I go in to fix things its 60% front-end.

Idk, I like AI when it works, but it drives me insane when it keeps making errors. I've had a few errors which I figured out from documentation fairly quickly, provided said docs but the AI would still mess it up somehow.

[cloverich]

> provided said docs but the AI would still mess it up somehow.

The AI is not intelligent. Its really hard to grasp cleanly. But it can't do anything logically like we do. Its pattern matching. It has to be a pattern its seen; then it can assemble them. If there are competing patterns - it'll trip up being consistent. Long established libraries and languages that change the least, it'll be best at. Anything newer it'll be bad at - even with documentation. The only way out is to give it tests, then it can loop over several simpler problems, where the errors (failed tests) match well onto the more basic primitives that don't really change (wrong string, wrong type, wrong structure, etc)

[throw-the-towel]

You know what else is pattern matching? Experience.

I think I finally have a way to describe how AI works: infinite experience, none of the smarts.

[some-guy]

At $LARGE_ENTERPRISE_COMPANY, I've found that if you have:

1) A designer that uses Figma correctly (using well defined components / design systems) 2) A front-end framework as close to HTML / CSS as possible for the visuals (I have success with Web Components / Lit) with Figma MCP

The front-end is usually one-shot using frontier models. However in my experience, designers are all over the place with using Figma correctly.

[evilduck]

I'll second this, with the caveat that I've not yet tried to build anything with Fable.

Every engineer can now produce things in the front end that doesn't look like complete and utter garbage, sure, but everyone is also producing the new-era of Twitter Bootstrap pages. It all has the same touched-by-AI look and it might as well be customer kryptonite from everything I've experienced at my workplace with customer surveys and collaboration. It has raised the floor substantially for internal tools and admin pages though.