Join the discussion
Write your take first — we'll ask for email only when you're ready to publish.
- Hacker News
- You can't one shot a perfect app with AI. You definitely can create a pretty complex and beautiful production-ready app with AI in a couple of days or weeks depending on what exactly you're building.
I now have my own link catalog, read-latter app and an RSS reader. Tailored to work exactly how I like. Hardened, with automated backup, and external users for the RSS app. It works. It takes learning, some knowledge of terms and very high-level practices, plus design thinking, but I haven't written a line of code for these.
by ymolodtsov - aggressively proactive
I'd say that's the correct way to describe frontier models. They were trained with reinforcement learning based on human feedback. And obviously, humans prefer the bug-free variant. That's why models are now super verbose and spam tests like crazy. In their training environment, tokens were effectively free. And the humans that got asked never saw the price. If you ask people to choose the better offer and both are free, you end up with bloat. It's like people over-filling their plate at a buffet, then leaving leftovers. Except in this case, it's AI models burning through your wallet.
by fxtentacle - I've found that LLMs make throwaway software better than I ever did.
They handle edge cases, catch bugs, and write tests that I'd never write.
Even if, however, this leads to the average piece of software improving, this one-shot complexity has the same issues as any large project. The more code, the longer it takes to steer the ship.
This "rising tide lifts all boats" mentality will make exceptional software even rarer than it is today.
Excited for the Roller Coaster Tycoons of tomorrow[0].
[0]https://en.wikipedia.org/wiki/RollerCoaster_Tycoon_(video_ga...
by jumploops - This really resonated for me. It's like the smarter the model gets, somehow the more tokens get burned? Same failure mode whether you’re on Claude, Codex, or Cursor: the harness will spend the whole pool if you let it. I'm building my own Harness on top of pi that is add supported (www.freepi.ai) mostly because pi is so much more efficient with tokens. (That said, it tens to be slower and vastly more verbose with information I don't need to know). But yeah, since I'm trying to offer free ad supported inference the vibe tax would kill the business model. I've even been thinking about installing the 'caveman' skill to reign in token costs.by freepiai
- This article somewhat reflects my experience with autonomous agentic coding. I've run several experiments with similar results: the agent burns through all my tokens while making very little progress, or produces something unacceptable.
So I'd rather micromanage the process step by step. It takes more of my time, but the result is much, much closer to what I actually wanted.
by dzhar11 - Hyperbolic, but I'm seeing hints of this – Models refusing to do pair work with an engineer and trust their input, instead mandating having full control over something. Friends switching back from Fable/Opus 5 to Opus 4.8 just so they can have some input.
Anthropic especially right now seem to be optimising for doing the whole task with no input. That's fine when that's the only task, and it's fine when you don't care how the sausage is made, but it's not fine for actual software engineering.
by danpalmer - Yeah I'm getting this feeling too, that Opus 5 collaborates better with other Claudes, but that some of the older Opus models collaborated with people better.by TOMDM
- I’ve never had an agent fail to write the actual implementation. Has it done so badly, yes, but not nothing but tests. This sounds to me like a rare case that doesn’t generalize.
If the general idea is that these agents write too many tests, sure I guess? ‘Too many tests’ doesn’t sound like a failure case of engineering to me; typically software has had too few tests. Also, a lot of the power of these agents is their ability to self-verify and correct, which the test loop is a part of.
Nobody is making you pay this supposed tax. Just tell it not to write tests.
by markbao - I tried, but I’m not sure I understand. The vibe tax is caused by the model trying to one-shot everything and doing so requires unnecessary tests? How are vibe coders training the model over months? Do you mean their sessions and preferences are being fed back into the RL?by alehlopeh
- I'll try to explain my experience with this. I've noticed the AI has a tendency to overengineer scaffolding. For instance, I asked it to help me with a refactor, and it erected this massive 100kloc function registry, and then caused GitHub CI to verify the contracts every single commit, which took upwards of 30 minutes (I suspect this proclivity is widespread and has contributed to their recent issues).
As if this wasn't bad enough, it also was not smart enough to regenerate the evidence in these contracts as it changed the underlying source code. So it would get in a loop where it would update code -> commit -> 15 minutes later CI would error citing the contracts weren't updated -> it would fix the contracts -> 15 minutes later CI would error because the fix was wrong -> it would fix the fix and commit -> 15 minutes later contracts would fail -> contracts were fixed again and this time maybe 30 minutes later it would pass, maybe it errors again.
This loop could go on all day every day if someone wasn't paying attention because the agent has no concept of time or wasted work. It's an AI livelock of sorts, but it will eventually converge in my experience. It'll just take 10x longer (literally like 20+ hours) than if you just intervene and tell it knock it off, so it feels like lighting money on fire (hence the tax).
That's why I feel like this vibe coding stuff has to actually be monitored, like a Tesla system -- because like a Tesla system it cannot be trusted to not crash into the proverbial code wall.
by ModernMech - Forgot where i saw it discussed; If you observe recent model benchmarks over the past year; the performance is slowly climbing, but if you divide by the token count; the score per token is dropping.
The current trend in state-of-art LLM coding agents is giving more output, thinking longer and checking the results more to catch mistakes. Be it an economics inventive to make users burn through their quota or show increase in usage for shareholders, or a market demand of users liking the ability of models to do independent work without intervention or oversight; the result is what the article seem to call the Vibe Tax.
I myself asked Claude code recently to review a somewhat large PR, to see what it would find. I didn't expect much, but also didn't quite realize how the model would interpret my request; I burned $20 in 3 minutes in API usage, as it ran 2 sub-agents which themselves spun up 5 more each. Most sub-agents were manually checking for things clang-tidy would catch without actually calling clang-tidy. This behavior rose as i changed from sonnet/opus 4.6 to 4.8 and now 5.0.
I don't want to run a agent independently in this way; i ask targeted questions about specific things and review the result. But model development is targeted towards a more hands-off "vibe" workflow, because that's where the money and hype is. As a result, i find the models more frustrating, less trustworthy and more costly to my work. (I've even started using haiku more, since it remains to-the-point without steering away from what i ask)
- I feel this, yes.
In effect, I’ve always wanted a pair programmer agent, not a zero to one programming agent. Unfortunately models these days are mostly of the latter kind and it has caused a major disruption in the way I work. I’d much rather appreciate a small model making fast and specific edits that I ask if it, rather than ingesting 20 files to make changes, and then starting to write tests, etc.
- Wasn't that what the first iterations of cursor were?by devrob
- I would love, love an even better Zeta model for that reason.by Lalabadie
- Just hook vanilla pi to Sol and you’ll have the pair programmer agent you crave. Make sure never to install a subagent extension. That’s it!by jdkoeck
- I've found writing small well defined tickets and getting Claude to work on them works well for this type of workflow.by beezlewax
- I just had a sunday afternoon request for a solution to a trivial but annoying spammer pattern on a stackoverflow-like site I maintain.
The software has a plugin API. I asked Muse Glimmer to recommend a plugin — it found one but I tested it and it didn't work for unclear reasons (among other things the software installed version is old, the plugin older). I then asked it to outline how to implement a simple word filter, it gave me an overview of some hooks that looked right from dim-and-distant-past recollection of reading the docs when I installed it. I asked it some questions, it did the research.
I then set it off generating the skeleton of a filter plugin, went to the shops to buy food, came back and worked through filling it in and finishing it off. There was a bug. It found the solution.
It's only about 100 lines of code but it is a random old webapp and it had to look stuff up to finish it, and I think it did rather well. All on my Mac.
I am deeply cynical of the one-shot code, "nobody codes anymore" hype culture idea and that distaste put me off AI and agentic coding for ages. Like you, I want an assistant but as a freelancer I have to stay in control. I have no interest in the "just specify loops" BS and it will be bad for my business anyway.
I worry about code that I don't have a good working overivew of, and I worry that I might forget what I have done (I have pretty bad issues with focus and memory). But in this particular case, I don't really care if I forget, because there's documented code and I have no intention of specialising in this app. So it was a nice little test case.
I also don't really want to sit around waiting for Qwen 3.8 27B on this machine. Muse Glimmer is fine, actually. Gets to the solution as quickly as Qwen 3.6 35B-A3B.
This gives me a little hope that local AI will give me the sort of responsive developer sidekick I actually want.
by dofm - Just said something similar myself in another thread. I'm either writing things by hand (and using LLMs for research, or double checking an idea), or having an LLM spit out something I treat as an external dependency. Its still too tedious for me to use them to write code when I care how it works or there's not obvious invariants the code needs to holdby lilbigdoot
- I have found what works well is to modularise the code as much as possible and get an agent to work within a very limited scope. Break everything down well using SRP with well defined interfaces and let the agent work on small problems. Then when it shits the bed, there’s a smaller blast radius and you can strip back and try again.
I think seasoned developers, over time, learn how to work a code base and design components with well defined interfaces, where the implementation is isolated in small well contained classes. SRP etc. more junior programmers can work on those smaller components/services in isolation.
For me this also seems to be a productive way to work along side an agent. Break up functionally into well defined chunks, and let the agent work on each small problem. Take more of a lead in the architecture I suppose.
by dave_sid - Most, if not all, of the code in the company I work in is written by AI. Our tests are useless. We have tests that make sure that mongoose schemas are creating the collections defined in them. We have tests that check that zod shcemas parse objects currently. Every PR, even if 1 line change, will drag 25 file changes because our tests are so ad-hoc, so verbose, and so incompatible with each other. Imagine how the codebase is looking... This is a badge of incompetence for the company I work with, and prob to the entire tech world.
The funny part is? This makes our managers and investors proud, every PR is bigger, we make x3 more PRs (wheres the promised x10).
When they say this is the death of software engineering, this is what they mean.
by localhoster - What a disgusting mess. Hopefully at some point people will be willing to listen to arguments in the vein of: “we spent $x dollars writing y loc that was then reviewed using z hours that did nothing, added no revenue and we then spent w additional budget removing it”by FuckButtons
- A project I was working on for a couple years until this spring started to turn like this when another contractor started committing AI work. There was a test literally called “test_imports.py” that, you know, tested that you could import every module of the project. I asked if we could just run `python __main__.py` to test that the imports worked, but he said no, we need this for big Claude-powered refactors to make sure nothing breaks.by eloisius
- i'm not sure why people expect agents to one shot everything to perfection with just a prompt.
There's a reason why we talk about software development lifecycle, design, architecture, testing ... It's because it's been the most reliable way to build and ship software. We shouldn't expect discard this and expect agents to perform well outside of this.
I'm treating LLM agents as junior devs who happen to have vast knowledge of software engineering. As their team leader i make them go through planning, implementation, bug sweeping cycles using strict workflows. And it works quite well, i've been working on several large projects (1M+ LOC java,typescript,c/c++) and by any measure the projects are healthy. Sure the code isn't that beautiful, sure i'd have written things differently but it's pretty good nonetheless.
Shameless plug here: i've been also working on https://kodfactory.com, the code factory i've built to work on these large projects with workflows, reviews, etc ... I'm cleaning things up to open source it later.
by guybedo - I can one shot a prompt if I write down a nice spec file, Claude can do a lot in one shot. I test it every few months. With enough detail Claude will know what to do.
- If you explicitly ask the agent to make the perfect architecture for the problem and write it down in to a spec and have the developer agents follow it they will. Its just that coding agents have a hard time coding at think about architecture at the same time.by perarneng
- > i'm not sure why people expect agents to one shot everything to perfection with just a prompt.
They do often enough that it's not a surprising event, depending on prompt quality, context available, ability for the result to be objectively judged and iterate on by the agent, etc. For frontiers on very high settings at least.
by jimmaswell - > i'm not sure why people expect agents to one shot everything to perfection with just a prompt.
because that's the end goal? and for simple small stuff they're already there?
by redox99 - > i'm not sure why people expect agents to one shot everything to perfection with just a prompt.
Every time you see a benchmark for "how long the agent can go without asking for human intervention", that's encouraging vibe coding.
by JoshTriplett - Using AI is kayfabe. What I mean is, you create interaction patterns that resemble how humans work. This is because it is what the models are trained on but also because we've all been trained to interact in this way. So it manipulates you into providing more useful prompts.
But I don't really want to play a part in a simulation, trying to cajole my scene partners into saying the lines I need them to say. I want to use a tool the same way I would use any other tool. If this is AI it should just do the thing. Anything else is an imperfection of the technology.
But at the same time, language is a vague communication medium. We have a precise language for describing forms of computation, but that's code so we're back at square one. We still haven't nailed the right amount of follow up and correction and interrupt-ability of these coding agents.
And we may never figure it out. It may simply be impossible. But it doesn't mean this weird anthropomorphization of AI is something I want to do. If I wanted to be a manager, I would be a manager.
by CoolestBeans - > i'm not sure why people expect agents to one shot everything to perfection with just a prompt.
because that's how agents are marketed.
by 0x457