556 karma · joined June 18, 2019
I’ve worked with two junior devs recently who are very good precisely because they’re trying to understand what they’re doing, rather than just producing code.
I’ve also worked with senior devs who basically gave up and stopped trying to understand the code. They became much worse engineers as a result. At this point I would much rather work with those 2 juniors.
They use it to explore things they don’t understand, ask me questions to clarify their reasoning, double-check assumptions. They generally use the tools available to increase their understanding.
It is just an amateur blog post about a small automation experiment. I am not selling anything or promoting a tool.
A new account doing anything unusual looks suspicious immediately. A 10yo account with normal history and real usage would probably be much less likely to get banned for the same behaviour.
I’ve edited it to focus more on the actual problem: the bot starts with a large visual search space and has to narrow it down to a small target.
Thanks for the feedback.
They likely have entire teams working on this, plus adversarial teams trying to break their own detection systems.
I'm blocking as much tracking as I reasonably can and they still caught it within a few days. So I doubt this failed because of one obvious browser API. It is probably a combination of behavioural signals that made the account look automated.
But that is part of what makes it interesting to me. Instagram has real anti-abuse systems, rate limits, weird UI states, etc. It's just a good place to experiment, everyone knows about it and it's something I was interested in.
I was totally expecting to get the account banned quickly.
Although I just added a disclaimer to make it clear.
> “I went from 4-5 parallel branches down to 2, sometimes just one. […] I had to stop to think things through.”
Where do you see me drunk on power here?
My point is that it isn’t faster because producing the code isn’t the bottleneck.
E.g: They kept predicting widespread fully autonomous vehicles. Significant progress has been made but a vehicle taking you from A to B without human oversight remain elusive.
Demand for great software developers isn't fading. Let's lean in and master our craft, not limit our growth by over-relying on AI.
Admittedly, I could be wrong. Maybe next year AI will be able to fully build software on its own and software engineering is dead. If you think that's true, why aren't you planning your career change right now?
If we land somewhere in the middle, what skills can I develop to become a better vibe coder which aren't just becoming a better software engineer generally? When I know exactly what to do, AI is just a faster typist. But when I'm stuck, if I outsource the solution entirely to AI, how do I actually learn and grow?
You crystallised the disconnect yourself: "with the help of Claude."
The article’s first line states: "LLMs are helpful." My stance is clear. I’m not anti-AI. When I wrote "I use Claude on a VERY short leash with a specific purpose" ... that is "with the help of Claude."
My experiment tested the opposite: "let Claude build it".
The 'telegram bot' is a fullstack app with non-trivial db schema, state management and 3rd party integration. Instead of dismissing it, show me a better alternative and explain why it’s superior.
You defend expertise in your domain (research synthesis), describing how you "constrained the robots" using your expertise to guide Claude. Yet you dismiss equivalent expertise in software engineering, reducing it to "$100k+ to write crappy JavaScript."
How did you "constrain" Claude without knowing what needed constraining, why it mattered and how to evaluate if the output was valid?
That is expertise. The salary is a symptom of it.
I'm trying to understand what I'm missing here. You describe these "AI whisperers" who can tamp down complexity, but doesn't that still require strong fundamental skills and deep knowledge of how to build software? Or am I thinking about this wrong?
If it eventually lifts this ceiling high enough that everyone can build with AI, how does this not simply become minimun wage work?
In which case, what's the unique value proposition?
The way I see it, either:
1. AI gets so good that everyone can build anything (in which case, why pay premium for whisperers?)
2. Or it stays limited enough that you need real engineering skills (in which case, we're back to valuing traditional expertise)
Where is the version of this story where retail clerks enthusiastically learned the UPC scanner system and then leveraged that knowledge into better careers? The inspiring tales of workers who "embraced the future" and thrived?
The technology was specifically designed to eliminate the need for worker expertise, not enhance it. They were learning to operate a system whose entire purpose was to make them replaceable.
I'm sure some clerks were skeptics back then. When technology did take over and replaced them, they probably got a big fat "I told you so." But everyone got the exact same outcome as everyone else: lower wages, fewer opportunities, less job security.
But imagine stores in the late 1980s realising their fancy new systems kept scanning steaks as bananas and designer jeans as dish soap. If databases crashed regularly and people were clueless as to what cost what.
Companies would have been desperately trying to retain and recruit back the very workers whose skills they'd just dismissed as outdated. The ones who knew every price would have become essential.
Unlike retail work, software engineering has multiple layers of skill. What happens if/when AI coding tools fail? Code can fail in subtle and catastrophic ways. Or what if it simply fails to live up to the hype and doesn't deliver?
The very tools that devalue coding skills might actually increase the value of deep engineering knowledge.
If AI replaces me, I'll be sad. I'll have to retrain, acquire new skills and find a new career. But at least that's a clean break, a clear transition point. My nightmare scenario is that it turns me into a button-pusher who prompts AI without understanding what it's producing, making me as replaceable as those scanner-operating clerks.
Today, I have to make a choice. I can decide to become the best clerk I possibly can, or I can learn how to use the scanner. What I would like to know is: people who think the scanner will be good enough, why bother and not plan your career change instead?
As for the broader responsibilities. Things like architecting systems, shaping their structure and behavior, planning their evolution and keeping everything clean with regular refactoring. These sound exactly like what I've been doing for the past 15 years. So what's actually different here?
Plus, I thought it would immediately shut down the strawman that I'm anti-AI. The generated image clearly shows I'm not.
Sorry it put you off. I hope you still found the text interesting.
I read the news: AI is taking over and we'll soon all be out of jobs. So I have to decide. Do I double down on software engineering, pivot to vibe coding, or try something completely different?
I need a sense of certainty to make this call, so I researched it. This post is the result. I might be wrong, but at least I'm choosing a clear direction instead of constantly switching and never getting good at either.
Vibe coding today doesn't deliver anywhere near the value of a competent software engineer. Rather than extrapolate from past progress, I looked at what it would take today. You asked about how we will turn a vibe-coded codebase into production-ready systems. I have no idea how we'll do that and I didn't find someone with a solid plan for it.
The logical conclusion here is that there's still plenty of runway for skilled software engineers. So I'm betting on becoming a better one with or without AI.
About "vibe coded codebases are garbage". If someone doesn't know how to build software (or quality doesn't matter) vibe coding is perfect. The code might be garbage but it beats having nothing.
These projects would otherwise be Excel spreadsheets or duct-taped tools. Now they have another option.
The problem is when people suggest vibe coding replaces developer skills, as if producing code was the bottleneck.
But just to make sure I got this right, I looked it up on Wikipedia:
> "Karpathy described it as "fully giving in to the vibes, embracing exponentials, and forgetting that the code even exists."
> "A key part of the definition of vibe coding is that the user accepts code without full understanding"
> "If an LLM wrote every line of your code, but you've reviewed, tested, and understood it all, that's not vibe coding in my book—that's using an LLM as a typing assistant."
What really sold vibe Coding to me was the idea that I could be significantly faster by working on multiple features in parallel. I'm an agent orchestrator. I send them off in the right direction, regularly check on them and correct if needed. They run all the tests and keep working until they all pass. Once they're done, I review their work and if I'm satisfied, merge it.
The hard reality is that properly reviewing code at this scale is extremely difficult and I inevitably let errors slip through.
Maybe I should spend more time reviewing it. But I'm almost certain it would take me more time and effort than writing the code myself and I'm not even sure I'd ever reach the same level of understanding.
If I ask Claude to do one thing at a time and monitor it, I need to get the prompt right, wait for it to respond with some changes and then review them thoroughly before accepting them. I can imagine this being an alternative way to build software. But I doubt it's faster.
But if it does happen, I'd rather be the one with enough technical skills to be able to jump in when agents aren't good enough or create new ones than the one managing them.
If it does happen and AI becomes good enough to replace a competent developer, vibe coders might still have a job after the dust settles. But it will be minimum wage work and the profession of software engineering as we know it is dead.
I guess this argument falls apart if you are already on (or close to) minimum wage.
Right now, I have a bug with my message queue. I didn't spend enough time thinking about the proper way to implement and test it. I guess I also didn't pay enough attention to the PR. But then, if I did, is it really vibe coding or classic software engineering while getting Claude to write for me with the added burden of reviewing it thoroughly?
The issue is that if a user sends a message to someone, quickly switches to someone else and sends a message, both messages get sent to the first person.
I'll have to figure out why it happens. My guess is that it's somewhere in the React app. Claude can't figure it out. What can I do to help it?