znnajdla

You've got to start somewhere. You peel each layer off the dependency stack one at a time.

When I try to show my non programmer friends what AI can do, I always get embarrassed by how bad Claude and Codex are at non-programming work. For something that is supposed to be world changing and something that I find unbelievably exciting as a programmer, it's just embarrassingly bad at other basic work. Then I realized it's the harness that matters more than the LLM for all practical use cases.

Most people outside of the techie bubble simply cannot understand what all the hype is about AI because to them it's just a slightly more advanced version of Google, and a weird friend to chat with in the computer. Yes, most normies use ChatGPT every day. But very few people actually get productive use out of it.

One of the reasons why I think the software engineering profession is not going anywhere is because it's going to take a lot more visionaries like Steve Jobs to completely reinvent the computing experience of AI for so many different professions. As far as I know, most professions have not had a Claude Code moment like we programmers did. And it will take engineers to build those harnesses for millions of different use cases. And that's why the engineering profession is not going anywhere. Because even though LLMs may be commoditized (they already are), harnesses cannot be.

What is LE? Let’s Encrypt?

It does make a difference because the only reason OpenAI cares about these problems in particular versus any other random problem in math is because of the prestige associated with official recognition, not just claiming something as marketing.

I don't know how this is not obvious to anyone, but the only way to slow down AI progress is an actual world war.

Have you actually used the latest Chinese models? Because, in my experience, they're not just cheaper, they're actually better in many cases. In one recent experiment I did a few days ago on a task that I need in production at scale at my company, Qwen 3.8 and GLM 5.3 Flash were not just cheaper, the results were significantly higher quality than GPT 5.6 Sol.

Even if China doesn't continue to release this stuff for free, I think somebody will. Eventually, maybe Europe or maybe a smaller country that picks up this knowledge will. Or maybe just some random philanthropic billionaire.

”Fuck it, make it anyway“ is absolutely the best advice I can give to programmers going through the crisis. Just make something that people want and you may find the joy of building again.

> And yes, it will reduce your value in the job market. Eventually it will be a dramatic reduction.

Maybe not. Look up Jevons paradox. Maybe software engineering was a bit overvalued before, anyway. And now it'll come down a bit. But I don't think the job is going to go away. It will just become a little bit less high paid.

Couldn't disagree more. Here's how I look at it as a person who wrote my code by hand for 15 years. I'm the owner of my own small bootstrapped startup.

1. I've always been enjoyed delivering solutions more than the actual process of programming, and AI just helps me deliver more of that faster. And for the actual process of programming, I found a way of working with AI that makes it enjoyable. I've built my own AI agent coordinator and harness which works the way I want it to. I enjoy working with AI in that harness.

2. It is completely false that you cannot prompt the AI any better than anybody else. The likes of Terrence Tao can prompt ChatGPT to solve Millennium Prize problems in math. But I'm not a professional mathematician, so I cannot do that. Generally speaking, the more deep technical knowledge you have, the stronger prompts you can write because an amateur doesn't even know what to potentially improve. Actual real world productivity with AI is still strongly correlated with deep technical knowledge and experience, and I expect it to stay that way. This has nothing to do with AI capabilities, which I expect will keep gettinf better. This has to do with the fundamental definition of a problem requires expertise to handle it.

3. Jevons paradox suggests that the fact that building software has become cheap now may only increase demand, not decrease it. Sure, traditional software engineering jobs may not look the same. But AI has made it possible to build so many kinds of software that I would have never considered before. Like, for example, I built my own email client, which works exactly the way I want it to. Imagine every company starts building their own operating system, email client, and CRM. I can easily see that happening, and I can actually see it as a net positive as well as a fun thing to watch. And only an increase in job opportunities.

Correction: OpenAI takes credit for what it's model spits out in response to other people's prompts. That's even worse.

Sounds like you just copy-pasted from AI without even understanding what you're talking about.

Based on what you're saying, you're claiming this is Grossman's work, not Einstein's. Why don't we rewrite scientific history too based on your copy-pasted AI slop?

It's so pointless talking to idiots who don't what they're talking about when they use AI, just because they think AI does everything, that reflects their own experience, not the experience of people who actually do real work. Some people are driven by AI, others drive it. As for those who are driven by it, they don't have sufficient imagination to think otherwise.

The difference is that Einstein didn't literally have someone prompting him towards his result.

My point is if you're not a mathematician on the level of Terence Tao you wouldn't have been able to get this result.

Thus, AI cannot be credited with the result. Because it's not AI that did it, it's the human that used it as a tool to get the result.

I think you're not aware of the specifics of this case. A mathematician worked on the prompts for over a year to get this proof. Look up the details.

Mostly low-level performance optimization and new programming language design.

So, what if they surpass me? It's still a tool under my control. Many engineers work under the authority of a CEO, even though they're smarter than them. Just because AI might be more intelligent doesn't mean it's still not a machine that is completely under my control.

I don't get this take at all. Do you feel inferior that bears are stronger than you? Or that a machine can lift more weight than you? Or that a calculator can calculate faster than you?

These AI things are machines which are completely under our control. Even if they become more intelligent than us, that doesn't mean that it has authority over us. There are plenty of very intelligent humans which work under the authority of much less intelligent humans. Many engineers are smarter than their CEOs.

What makes you think that just because an AI is more intelligent that it will not be completely subject to our authority? It's just a tool in our hands.

My point is, the AI could not have done what it did without the skilled mathemetician prompting it. It was just a tool in the hands of the human that did it. And that was a very skilled human. Unless you're a professional mathematician, you cannot get the same result even if you use the same AI model. That's the proof that it's the human's work, not the AI.

Why don't you try getting a Millennium Prize then? If you think it was the AI that did it, you have access to the exact same ChatGPT.

> That an AI was able to synthesize all the work and bring it forward

But that's not what it did. A mathematician literally prompted it towards that, helping it every step of the way for years.

And only the mathematician was able to make the AI do that. Anybody else with the same AI could not do the same thing.

If you think it was the AI that solved a Millennium Prize, then why don't you try using ChatGPT to get a Millennium Prize?

I think the author is not aware of how much the Millennium Prize proof was actually driven by a human working for years towards that problem. The AI did not solve this on its own. A world class mathematician prompted it towards the proof. In my view, this is still a human achievement, not an AI achievement.

If I design a bulldozer to push a five ton rock, did I push the rock or the bulldozer?

If the AI really did get the Millennium Prize, then why can't you get a Millennium Prize when you have access to the exact same model in ChatGPT?

I don't think humans necessarily become subpar when AI can do most of the work. Taking my own personal example, I have far more intellectual curiosity and improved my skills in programming far more with Claude Code than for 15 years of programming without AI simply because I was bogged down by boilerplate and grunt work. Now that AI handles most of the boilerplate and grunt work and can handle harder and harder problems, I have the time and space to work on unexplored frontier problems.

So, no, my skills have not become subpar, but have only become stronger because of the presence of AI.

It's hard even for a lot of tech people to use AI at its peak power. Many of the programmers I know are not using it nearly as much as they could.

Maybe it's a lag, or maybe it's just that other parts of the world have their own trajectory.

There is a well known problem with LLMs that if you feed it its own output, it gradually gets worse and worse.

We haven't really understood what the limitations of LLMs are. And I do not know. But as a person who uses Fable and Sol regularly to design my own new programming language, they suck at the task of defining new systems coherently. So far no AI I have tried is good at defining new coherent systems well. I suspect it's because of the recursive problem of using it on its own output.

So, a human is still needed to define and write the sort of first genesis of the system, and then AI can take it once the problem has been defined. But defining the problem and using it on itself is what AI is really bad at (for now).

This is a problem with every technology product. Before Facebook and Gmail had strong institutional controls, any employee could just open anybody's email and private messages on a whim. Now things are a bit more strict for random employees, but if the company as a whole wants to do something, you bet they can.

Never underestimate a small dedicated underdog. You never know who might catch up later in the race

I can think of one use case for this: Dashboards to answer one-off questions. Think PowerBI, Tableau, etc.

The most interesting part about this for me is that they decided to create their own language or DSL for the task at hand. So it's not just a large language model; it's an LLM with its own language.

I have a feeling that the best AI systems to come will, in fact, be a complete package like this: a harness, a DSL, and an entire package designed to produce certain outcomes cheaper and faster.

And producing that complete package is why software engineering will not be obsolete.

This is an article about declining literacy in the West. But this is notably a Western decline problem. In some other parts of the world, literacy is improving.

One of the first things I noticed when I moved to Ukraine in 2020 was that even the political jokes that you see on television have real highbrow wit and humor, which assume a high degree of literacy to understand. Even talking to some kids who were 14 years old, I was surprised at how they had memorized poetry in multiple languages and could write beautifully in cursive. You don't find that anymore in the West.

Sure, kids over here doomscroll themselves on TikTok as well, but they are not illiterate.

Sure, I already do that at my company. Before I go to sleep, I put Opus 5 in bypass permissions mode and give it SSH access to my server and tell it to restart it or take necessary actions to keep the server online while I sleep. I could easily put it in control of more things. But at that point, AI is just managing a well-oiled machine and harness with tools that I have spent years building. It still didn't build a company. Any idiot can manage an established company (how many idiots have we had in the White House?), but very few can build one. And our original discussion was about AI building a company.