| ▲ | jaykru 2 hours ago |
| Have you actually tried vibecoding something not firmly on distribution? I'll give an anecdote. I've been building for the past few weeks a fancy text editor for myself to work with AI in a more sustainable and sane way. I don't have much time for this, so I've entirely vibecoded it: I don't look at the code and I've provided only some initial design and engineering decisions, debug hints, etc. on top of natural language feature requests. For about a week using the strongest Claude models (Fable 5.2 and Opus 5.5) this went great, and I had a minimally functional tool that worked very nicely. It seems that I hit some sort of complexity ceiling for those models though. The second week, every feature I try adding creates a new type of jank that can (at least by me) be only vaguely characterized. Most recently inserting into a line (even in the middle of the viewport) causes the viewport to jitter about. Our work will change, we will spend less time on tedium, we will become more ambitious. Barring an architectural paradigm shift we will still need people with strong engineering and design instincts, and even more importantly the ability to write reference code and formal specifications codifying desired behaviors. |
|
| ▲ | cautiouscat 2 hours ago | parent | next [-] |
| I feel like comments like these, are from people who don’t work with complex systems. There is absolutely a ceiling to vibecoding and I don’t think it’s technical. When systems get complex, someone needs to know them and know where to burn the tokens. You need to know what events need to be produced because you have specific consumers. You need to know your business invariants and why. If you’re working on something that isn’t complex, yeah I can see why you’d think you can be replaced. If it’s just a front end? Yeah I can see that. A basic CRUD service? Sure. A business critical domain service with events, data migrations, complex business logic? I don’t see it. |
| |
| ▲ | sbysb 2 hours ago | parent | next [-] | | It is because the average developer has outsourced all product thinking to an external team. The gravy train might actually be over for the Software Engineer that takes a Jira ticket and produces a PR, but in my opinion the demand for Product Engineers is only going to increase. However it will be similar to cybersecurity in that there will be no 'entry level' product engineer. It will probably have requirements having worked in customer service or finance or some other adjacent field to get a larger picture of what you are delivering. You will also not need 3000 of these individuals anymore (did we ever really??) I can see a world where you go from "AI Shepard Customer Success rep" -> some middle role with increasing 'product' responsibilities -> Product Engineer. The new roles will also probably be more generic - instead of backend and frontend you have 'Backend' (Infra + API) and 'Frontend' (Frontend + Design) and both of these roles will require genuine product thinking | |
| ▲ | kalcode 30 minutes ago | parent | prev | next [-] | | Many LLMs are working on huge complex codebase. I'd look inwards at engineers unable to get it to work on large codebases. No different than sending a contractor in. Do you tell the contractor "understand the entire codebase, business context and documentation all at once!" No, and it's the same with leads, they can be given a system they understand but need to reference it. I'm very certain these are people brushing up against their lack of processes, workflows, or engineering practices. The context are large enough. You do not need to understand the world to build a chair. I see the very opposite in many different domains happening now, and some codebases are monsterous. They are figuring out ways that do indeed work well. The engineers that are figuring out these problems when working with a new tool, will have jobs. Those who continue to doubt, and distance themselves are the ones that will be behind. | |
| ▲ | garrickvanburen 2 hours ago | parent | prev | next [-] | | even somewhat complex systems. currently working on an initial product for a startup and while I'm substantially further along that I would be without LLMs, I'm at the point where the LLMs are now randomly breaking things. | | |
| ▲ | kannanvijayan 23 minutes ago | parent [-] | | I'm similarly working on a complex project and have run into these issues. There's typically a combination of issues that are occurring when the codebase goes off the rails. One is that cruft builds up. Large vibecoded codebases need regular "review cleaning". Where you review the whole thing top down for hygiene, contradictions, bad comments, style, redundancy, design. It's a standard thing for me now. A full review and cleanup pass once every couple months. The second thing that tends to go wrong is that after a certain complexity threshold, the models cannot keep track of the relevant details for a task very well. They lose sense of project directionality. Every task gets disconnected from other tasks and they start pulling the system apart at the seams, exacerbating the cruft issue above and accelerating collapse. I've found that separating my documentation into three distinct lobes helps: goaling docs, guardrail docs, and implementation docs. Goaling docs describe in prose the goals of the project which includes scope, certain design targets, and other things that you are aiming towards. Guardrails are for design and implementation choices you have explicitly ruled out, the "what not to do" part. Impl docs describe the state of the system as it is. The top-level agent documents refer to these and make their purpose explicit. You'd think a ticketing system might capture this, but it doesn't really provide the richness of expression a collection of documents in a directory does. These approaches aren't perfect. The process still has a tendency to wander off the rails, but it happens less often now and I catch it more quickly. Every session misses little bits of the goaling alignment, but they also find and address small issues missed by previous sessions. Overall it does a better job of keeping the codebase hovering around some threshold of coherency. I'm having a lot of fun to be honest. Yes it's bizarre and nothing like what development used to be, but it finally feels like there's some new things to learn now. | | |
| ▲ | kalcode 14 minutes ago | parent [-] | | Yes! Codebase health review, drift and realignment are big! These days they know how to handle their own context in review. So I will use that to find improvements too in retrospectivea and a small usage script hook it runs each push that alerts if high token usage. Hint the subagents are getting excessive context, which is also good hint of drift. So then I can measure when it starts happening typically via that script that Claude checks and readjusts when one goes over budget. |
|
| |
| ▲ | apexalpha 2 hours ago | parent | prev | next [-] | | What makes you think there won't be a (specific) model for that, too? | |
| ▲ | djha-skin 2 hours ago | parent | prev | next [-] | | This is more of a "table saw" moment. Table saws didn't obsolete carpenters, but it did mean now you only needed one. | |
| ▲ | cmrdporcupine 2 hours ago | parent | prev | next [-] | | I agree and it's why I find it hilarious to find places still trying to hire based on whiteboard coding and skillz testing nonsense. It's not that I don't think those skills are important, but they betray a fundamental misunderstanding of what SWE is, which is about managing systems, juggling complexity, and having some wisdom about how those things should fit together. Hopefully we start to better formalize some more of that now that the grunt work can be taken out of the equation a bit. And teach it in schools, find ways to measure it in interviews. etc | |
| ▲ | dboreham 2 hours ago | parent | prev | next [-] | | For me the experience is very similar to what it was like to manage a software operation with humans working for me. Actually for the first year or so it was like that where I was the senior engineer and AI was the junior devs. But then it got good enough that it could also be a trusted senior level advisor. So I'm just setting the project scope and general direction and making the odd tweak here and there now. I'm truly baffled by posts saying that AI is terrible in late 2026. My very heavy use experience is quite the opposite. | |
| ▲ | paul7986 2 hours ago | parent | prev | next [-] | | All digital forms of work are a dying field. Designing and building applications is now similar to writing a song with SUNO. Pumping out slop and anyone can do it even my friend working in an Amazon warehouse (never coded anything and a zero technical background) is pumping out apps. What we the people need to do is gain control and hold all the power of AI in our hands - not the filthy rich's hands! Devise systems that flip the power into our hands and create economies and jobs around it (i have ideas and surely 1000s to more do to around this). | |
| ▲ | bothers 2 hours ago | parent | prev [-] | | [dead] |
|
|
| ▲ | semiquaver 2 hours ago | parent | prev | next [-] |
| Take this as genuine advice, not criticism: The fact that you felt the need to “characterize” the jank and verbally describe it is evidence of your mistake. The only way for these new development models to function is to create a tight machine-only validation loop. Every aspect of an LLM-generated text editor must be drivable fully by the LLM itself so that it can independently observe the problems and fix them (and so those observations can be mechanically translated to regression test cases) LLMs won’t always guide you to do this correctly so you need to understand how important it is from the start and always enforce it. |
| |
| ▲ | qlte 2 hours ago | parent | next [-] | | I'm not seeing any indication that OP didn't do that. The jank they're describing sounds exactly like something that could slip past the harness' validation loop (weird intermittent graphical anomalies). | |
| ▲ | titzer 2 hours ago | parent | prev | next [-] | | Yeah, this is really important. Having a tight test loop is already great for humans, but doubly-so for agents. A clunky manual process sucks for humans and agents alike. | |
| ▲ | jaykru 2 hours ago | parent | prev | next [-] | | I appreciate the friendly tone :-) and I mean this not as a jab: I think maybe you missed the point of my comment above, which wasn't that one can't build functional software with LLMs, but rather that it requires the diligent involvement of a technical worker. > Barring an architectural paradigm shift we will still need people with strong engineering and design instincts, and even more importantly the ability to write reference code and formal specifications codifying desired behaviors. The people building those specifications and the tight validation loops you describe are the technical software workers of tomorrow. | | |
| ▲ | PhilipRoman an hour ago | parent [-] | | I generally agree with your idea, however you can also say "Hey Claude, I want to vibe code XYZ, set up an AI friendly automation environment" and it will have no problem throwing something together with qemu, etc. And of course it doesn't have to be you saying that, it could be a N+1 level agent, who loaded his /ai-project-management-best-practices skill. |
| |
| ▲ | the_real_cher 2 hours ago | parent | prev | next [-] | | I'm not sure you're correct. If the software was capable of creating high complexity software I, or anyone, should be able to replace and build anything I want with just some prompts and guidance. Anyone who has worked with LLM heavily knows this is far from the truth. | | | |
| ▲ | groestl 2 hours ago | parent | prev | next [-] | | > always enforce it. Here be lions, since these things tend to lie to you or themselves when skipping the enforcement part. | |
| ▲ | asdfman123 2 hours ago | parent | prev [-] | | I find LLMs produce very good results if you just prompt it into good software engineering practices. Write thorough unit tests, minimize diff noise, try to re-evaluate all work to see if it can be done more elegantly and simply. Then you can run an adversarial review agent to catch any bugs it missed. Same goes for AI slop comments. I can just prompt AI into more or less writing like me. |
|
|
| ▲ | microtonal 2 hours ago | parent | prev | next [-] |
| Our work will change, we will spend less time on tedium, we will become more ambitious. Barring an architectural paradigm shift we will still need people with strong engineering and design instincts, and even more importantly the ability to write reference code and formal specifications codifying desired behaviors. It is hard to say how things will develop the coming years, but this sums it up for me today. Almost constantly, LLMs (e.g. Opus 5.5/Sonnet 5.5) make choices that a junior programmer would think is ok, but I know that will create technical debt, subpar designs, etc. Note that I'm not saying that LLMs program like a junior programmer, in many ways they program like senior engineers who have happened to memoize all API docs as well. But they are often missing the larger context or don't understand the time dimension (what will be hard to maintain, what will introduce an API that we can not easily backtrack from, etc). So, I am not surprised that vibecoding breaks down as a project becomes larger. At the same time, current LLMs make me a lot more productive (which was not true for me ~one year ago). I can do in 30-60 minutes what would take me a day 1-2 years ago. In many ways, programming also became more interesting, because I can focus on design, the really challenging parts, while the LLM can quickly implement it. LLMs really feel like a different, in many ways complementary intelligence for development. |
|
| ▲ | jonah 2 hours ago | parent | prev | next [-] |
| I find myself working more as a technical project manager which is fine by me. My other firsthand observation is seeing the difference between what quality of software that I, as someone with many years of experience, can produce versus my coworker, a talented designer without SW experience who is very AI pilled and excited to be making things, is able to produce. |
|
| ▲ | flutas 2 hours ago | parent | prev | next [-] |
| At $DAYJOB I was working on something novel on mobile devices (so much so other companies we consulted with said it was something they wouldn't even try to do), and ran into a lot of the same issues. The model was quick to try and do quick patch fixes on various parts of the system to make it work, result in a janky but "usable" experience. Through a combo of me and it over about a month we got it working well enough to roll out though. The entire time I had to hand hold the model through the changes though, and it loved to try and throw it away for quick fixes that would break something else. |
| |
| ▲ | DANmode 39 minutes ago | parent [-] | | Don’t use the same chat/session this long. Dump your progress in meaningful ways, begin a new session. |
|
|
| ▲ | dist-epoch 2 hours ago | parent | prev | next [-] |
| There is a skill in using these LLMs. For example I don't see any mention that you ran debt reduction loops, code simplifications loops, and in general asking the model to clean up the code after it's initial phase of "organic growth". |
| |
| ▲ | Choad_20261104 2 hours ago | parent [-] | | A whole cottage industry could grow up that bridges between those who know how to craft the LLMs into customized personalized utilities for individuals and small businesses. For sure, the guys who lost out when automobiles displaced horses were the ones who were certain they only knew how to make buggy whips. Adapt and thrive. |
|
|
| ▲ | scottLobster 2 hours ago | parent | prev | next [-] |
| I ran into a similar issue with Gemini 3.8 in an even simpler context. Spent several hours asking it why my new module wasn't working and each time after several minutes of thinking it would exclaim "I FOUND THE ROOT CAUSE!" and despite implementing its fix the problem would still persist. After about a dozen failed iterations I reverted to the last known working commit (albeit with the new module only partially implemented), isolated the issue as best I could, then asked Gemini again and this time it correctly identified a latent timing bug that was the cause of all the problems. In short, I had to manually reduce the complexity of the problem for this magic infinity-dimensional-processor to deduce what the problem was. Maybe it would have gotten there eventually, but without AI I would have just reverted the commit a lot sooner and likely figured it out on my own with some basic debug logging. It's still a net gain, as it's successfully one-shotted coding errors and bugs that might have taken me hours to find before, but it can miss basic problems as well. |
| |
| ▲ | husseinhb 2 hours ago | parent [-] | | That's great you find the root cause, as a way that i using it that helped me a lot which is first of all understand what is the bug without touching the code. Then after i understand it i can make a targeted investigations for that issue and after that it will be obvious for the AI, and all of those we still didn't change even one line in the code. After that, we fix that issue (which is the root of the cause), release and congratulation the bug was fixed this way works even in lowest powerful models. |
|
|
| ▲ | foolserrandboy 2 hours ago | parent | prev | next [-] |
| How did everyone miss the actual thrust of this post? Maybe we really are cooked. |
|
| ▲ | mlmonkey 2 hours ago | parent | prev | next [-] |
| Today's models seem to lose focus as the conversation goes on. But, as @sama said, these models you see today are the worst they'll ever be. It's only a matter of time before someone figures out how to maintain focus and drive a product to a polished end state over many iterations. |
| |
| ▲ | 123987569 2 hours ago | parent [-] | | Lol no. The enshittification phase hasn't even started yet. Wait for your quantized "premium" models with less compute and less used GPU memory at higher prices. | | |
| ▲ | Tumblewood 2 hours ago | parent [-] | | How can these companies "enshittify" when open source models are not that far behind the frontier? Not to mention the closed-source competitors who would be happy to take market share. | | |
| ▲ | drTobiasFunke 2 hours ago | parent [-] | | Existence of open weight models doesnt mean anything to unit economics and shareholders seeking more profits because someone has to pay for training those models and someone still needs to host them at scale. |
|
|
|
|
| ▲ | aurareturn 2 hours ago | parent | prev | next [-] |
| It's still a skill issue. LLMs have massively improved the quality of our company's main app. However, someone less skilled at our company tried to vibe code an internal tool and it was pure crap. |
|
| ▲ | 0_____0 2 hours ago | parent | prev | next [-] |
| You're on the beginning of the slope of how your career as a SWE will change. Hope y'all been working on people skills. |
|
| ▲ | Jtarii 2 hours ago | parent | prev [-] |
| You are assuming models don't get significantly better than they currently are. |
| |
| ▲ | layer8 2 hours ago | parent | next [-] | | Models have been getting significantly better, but what hasn’t changed is that you ultimately can’t trust their judgement and always have to double-check the important details. | | |
| ▲ | packetlost 2 hours ago | parent [-] | | It's less their judgement and more they don't actually know what you're asking for. They "guess". | | |
| |
| ▲ | bigstrat2003 5 minutes ago | parent | prev [-] | | They haven't gotten significantly better yet. That's a good indication they won't in the future, either. |
|