Startup Ideas Bank
AI in Code Review? More Like a Recipe for Disaster.
AI roast score: 45/100 (F)
The idea
The Code Works. What Could Possibly Go Wrong?
Sylwia Laskowska
Posted on Jun 10
The Code Works. What Could Possibly Go Wrong?
# ai
# webdev
# discuss
AI as a tool rather than an authority
Would you treat a serious illness without seeing a doctor, relying only on whatever your favorite AI model suggested? Would you let AI take over your child's education?
Probably not.
So why are you willing to hand over your entire codebase to it?
JSNation is just around the corner, and as I mentioned before, I'll also be joining a discussion room called "Trusting AI Systems: How Much Is Too Much?" . So today, let's talk about exactly that.
Also, after an incredibly busy first half of the year, I think I'm officially entering vacation mode next week. Expect some JavaScript posts, popular bash commands, and occasional programming memes. 😄 I know many of you enjoy those too.
By the way, as I already mentioned, I'll be speaking at FrontKon in Prague this October! I have a feeling it's going to be one of the best conferences of the season. The organization has been fantastic so far. The agenda is already published, and I know my talk is scheduled for 3:30 PM. I haven't quite figured out which day yet, but don't worry, I should be able to sort that out before October. 😄
If you're into frontend development, definitely check it out. Apparently tickets are selling fast. You can also leave me a like here .
It will probably be my last conference appearance of 2026. Unless somebody invites me somewhere else. Which, as it turns out, is not entirely impossible. 😉 But that's a story for another day.
Does AI Lie?
Let's get back to the topic.
How much do you trust AI? And I'm not talking only about code. I'm also talking about knowledge.
I don't know how often you use LLMs outside programming, but I use them a lot. Really a lot. Sometimes I vent to them. Sometimes I ask for information, inspiration, or validation of an idea. And I've noticed an interesting pattern.
Remember school or university? General knowledge was easy to access. But when you needed something more specialized, you had to go to the library or dig through academic journals.
Models work in a surprisingly similar way. When I'm looking for general information, I trust them almost blindly. But the more I discuss topics I actually know well, the more nonsense I start noticing.
Yes, models hallucinate less than they used to. They no longer invent completely absurd facts every other answer. But do they really stop making mistakes? Not exactly.
Sometimes the facts are mostly correct, but names get mixed up. Sometimes two separate conversations will confidently give me two different explanations for the same medical issue. 😉
Of co
The roast
Your venture assumes developers will trust AI to review their codebase without serious oversight. This is not a small leap; it's a leap off a cliff. The market is skeptical, and for good reason. The differentiation and execution are vague at best. Stating that AI has issues with accuracy and expecting developers to pay for it seems contradictory. Moreover, relying on speed as your moat in an industry where trust and accuracy are paramount is fundamentally flawed. Your solo operation and lack of funding further compound the risk, making this a non-starter.
Red flags
- Trust issues with AI in critical tasks
- Contradictory market assumptions
- Solo founder with no funding
Verdict
Without addressing the trust and accuracy issues head-on, this venture is dead on arrival.
Roast your own startup idea →