// HACKER NEWS — CYBERSECURITY
AI Can Make You Suck Faster Too
Not too long ago, I shared a post that I really liked by Disesdi Shoshana Cox. They went through the math of how many revolutionary tech companies we should have by now thanks to AI. With a 10x improvement on development speed, the technical aspects of launching an app are no longer a limit to reaching a target market. It seems like AI has finally provided the silver bullet that solves the problem of programmers moving so slowly.
Borrowing Disesdi's logic in their arithmetic (which is hard to argue against), after four years of open source LLMs, we should have three AirBnBs,1 two Stripes,2 and three Dropboxes3 thanks to the power of AI.
The biggest new tech companies in the world since the advent of GenAI are, well, GenAI companies. OpenAI, Anthropic, and High-Flyer (developer of DeepSeek) are being valued at record-breaking sums of money. But the biggest disruptions they're causing is to the global economy. The most noticeable contribution to society from GenAI is making poorly worded emails the norm, flooding social media with garbage, and ruining the quality of internet search engines.
Or else admit this is a dopamine game that makes you feel like The Universe's Most Special Programmer™️ when it's really just gamified mass-scale intellectual dependency.
Disesdi's post struck a chord with me because I've witness the same underwhelming output of software during this AI rage. These tools, in the hands of non-technical people who have not put in the time and effort to learn about the nerdy stuff, are simply not able to generate working apps that will persist and be security-hardened against run-of-the-mill attacks. And when experienced software engineers use them, the impact feels like a limp handshake.
The massive blindspot here is that writing lines of code isn't the part of software development that drains the most time. This may seem like I'm contradicting myself on the argument that we should have at least a handful of AirBnBs or Dropboxes by now. But there isn't a contradiction because leaders of companies and startups truly believe that LLMs have solved the bottlenecks in product delivery. It's the perception that all technical issues in software development no longer exist because you can "just get Claude to do it".4
I recently attended a tech conference where there was a panel discussion on the main stage that was titled something like "No code, No problem". The panel comprised four startup founders who vibe coded their way through MVP and, assumedly, are taking money from customers.5 The moment that made me walk out of the session was when one of the founders said he didn't have, and won't ever need, a CTO because he can just use Claude to solve his technical problems.
Imagine that. You don't need software expertise in your software company. You don't need a structural engineer to design the structure of an office building. You don't need a surgeon to perform surgery.
I've been told by people who are much, much better developers than I am that they've sped up their productivity in writing software by using Claude and other coding agents. So I thought, what the heck? When you run a consultancy, you can actually convert that type of development speed into money! I had already used a few different AI agents to help write the boring stuff - boilerplate code, repetitive SQL, etc. But I wanted to see if I can get in on this 10x magic. I decided to put my money where my mouth was and I bought $10 worth of DeepSeek credits to use with a project I was working on.6
The process was incredibly infuriating. The chatbot recommended some of the dumbest shit you could possibly do. And let's be clear here: the code DeepSeek wrote would run, but it was a clown car rolling around with wheels held on by duct tape. Now, you might be thinking, "Well this fucking guy doesn't know how to prompt. If he was just better at prompting, he wouldn't suck so much." And you might be right. But I've also shipped a product before, which is inf