The creator of Ruby on Rails just said programmers may be worse at AI coding than people who cannot code. Lex Fridman asked David Heinemeier Hansson directly. His answer: 100%. Because once the AI handles the implementation, knowing how to write a loop is no longer the rare skill.... The valuable skills become: What should this product do? Who is it for? What should version one include? What needs to be removed? What should the AI build first? A traditional programmer may tell the agent exactly how to solve the problem. That experience can become a trap. The agent follows the old path instead of finding a better one. Meanwhile, someone who understands users, products, priorities, and taste can describe the desired result and let the agent choose the route. Programming knowledge still matters when the system breaks. But AI is moving the bottleneck away from writing code. The advantage now belongs to the person who knows what should exist, not merely how to implement it. WATCH FULL PODCAST ON Lex Fridman YT CHANNELshow more

LOL. That's assumes that AI agents perfectly implement what you tell them, and if you just give them a general ask they will precisely fulfill your wishes and the language we speak is perfectly clear with no ambiguities and all words have a single clear sense.

Yes… that‘s why ruby-on-rails was always nice but slow as hell! It is NOT enough to specify what the user wants, you also have to specify the technical details sometimes down to registers and database indices.

Very true and accurate

Skill issue... Next he will blame a hammer for not knowing how to drive a nail into wood.

🤣 you are talking about a legend here

😂🤣😂 I know but since 2016 I have been ignoring medals one might have and looking at what is right before me. You would be surprised how many times that has helped me.

yes, i can see where you are pointing at. Got it

I call this hyper-vibe ai evangelism.

Are you aware of DHH?

A am aware of ridiculousity of his claim. AI is 5 years from replacing whole SWE. It's too early now. This short phase is the phase of AI boosted fake confidence.

there are good product guys who are also programmers.

I agree for the coder if they're a jira factory worker. But a good software developer has always been the one with an architectural eye and that is still needed. Along with knowing what product to build and what makes it great. Architecture has no value on its own. It's what lets you evolve a great product.

This is so confusing. You are saying that programmers are worse at AI because they can't do the product design? But even before AI, programmers were not doing product design. You are mixing up product design and programming. And adding AI here out of nowhere like wth.

Totally agree! 💯 Working with my nephew, a programmer, on construction takeoff software. I'm the estimator, and I've used other software. He's full of technical jargon about all his implementation ideas – some great, some not so much. Ultimately, as the user, I know what needs to be built. 👷‍♂️💻

This starts to break down when you are unable to describe the problem since you don’t have the pre-requisite knowledge needed

This is the most retard take that I've seen.

I get his point, my take is micromanagement. But I think I'd like my AI surgery to be overseen by a "surgeon" and not a "prompt surgeon".

He forgot that the coding agents were trained on human written code... which is average...

yes and no. For an excellent result you need to know what is going on behind the seams. Also, once the product reaches the masses you NEED to understand the system and tech stack, one mistake and your db is gone, you get hacked, get denials of services, nasty bugs etc.

People with good taste will create the best products.

Well, that guy makes an awesome guest

You do realize ai generated code gets it wrong too sometimes? Somebody has to review generated code. Lol

Any evidence so far?

hahahahahaha we've finally made it! so the nerds they employ are totally smart enough to use it, just not the other billion nerds out there. it's almost as stupid as the "opus 5 isn't bad you just have to purge everything from your environment for it to be good" stuff from the other month lol

Yes and no. As an experienced developer I know software is complex and AIs also may add bugs or sometimes subtly incorrect behaviors that take valuable time to check. Just assuming some code works is a recipe for trouble. Software is a lot more complex than just "writing loops".

Product owners are kings

@ihteshamali honestly, it's wild to think about. feels like we're all just learning how to speak the AI's language now

Meu vizinho, dev senior, obeso, mora com a mãe em Sapopemba, 45 anos, disse que ele é insubstituível e que em breve vai ficar rico arrumando bugs em códigos feito por vibe coders.

Absolutely. Now writing code is not a moat. Understanding domain knowledge, knowing what should be the expected behaviour, what are all flows needs particularl change, all the domain level context became much important.

Final someone not afraid to tell the truth

Code is becoming the easy part

I am a Solution Architect and I agree with DHH

Many would agree.

It’s not like product managers with excellent product understanding have made unreasonable requests for decades already. The problem with today’s coding agents is that they won’t say no, and will move mountains of code to satisfy any architectural disaster.

This tracks hard. I don’t win because I can code better than a senior. I win because I know the bar, the rush, the boring problems that actually matter. AI can implement. I decide what deserves to exist.

Great for people who understand intuitive UX and product but also used to code.

A lot of non tech founders are realizing this and jumping in the pool themselves

What advantage now belongs to the person ….

Twisting his words, he said "there are problems" (not all problems), which any good programmer knows already.

Absolutely nailed it

i see his point first you make code work like an average joe but after that you also need to challenge it....LLM will usually contradict itself in longer discussion ...you give it few more lines of new context and it may totally abandon previous strategy