This was actually the only point in the essay with which I disagree, and it weakens the overall argument. Even 2 years ago, before agents or reasoning models, these LLMs were extremely powerful. The catch was, you needed to figure out what worked for you.
I wrote this comment elsewhere: >>44164846 -- Upshot: It took me months to figure out what worked for me, but AI enabled me to produce innovative (probably cutting edge) work in domains I had little prior background in. Yes, the hype should trigger your suspicions, but if respectable people with no stake in selling AI like @tptacek or @kentonv in the other AI thread are saying similar things, you should probably take a closer look.
Sure, but I would argue that the UX is the product, and that has radically improved in the past 6-12 months.
Yes, you could have produced similar results before, manually prompting the model each time, copy and pasting code, re-prompting the model as needed. I would strenuously argue that the structuring and automation of these tasks is what has made these models broadly usable and powerful.
In the same way that Apple didn't event mobile phones nor touchscreens nor OSes, but the specific combination of these things resulted in a product that was different in kind than what came before, and took over the world.
Likewise, the "putting the LLM into a structured box of validation and automated re-prompting" is huge! It changed the product radically, even if its constituent pieces existed already.
[edit] More generally I would argue that 95% of the useful applications of LLMs aren't about advancing the SOTA model capabilities and more about what kind of structured interaction environment we shove them into.
But I think my other point still stands: people will need to figure out for themselves how to fully exploit this technology. What worked for me, for instance, was structuring my code to be essentially functional in nature. This allows for tightly focused contexts which drastically reduces error rates. This is probably orthogonal to the better UX of current AI tooling. Unfortunately, the vast majority of existing code is not functional, and people will have to figure out how to make AI work with that.
A lot of that likely plays into your point about the work required to make useful LLM-based applications. To expand a bit more:
* AI is technology that behaves like people. This makes it confusing to reason about and work with. Products will need to solve for this cognitive dissonance to be successful, which will entail a combination of UX and guardrails.
* Context still seems to be king. My (possibly outdated) experience has been the "right" context trumps larger context windows. With code, for instance, this probably entails standard techniques like static analysis to find relevant bits of code, which some tools have been attempting. For data, this might require eliminating overfetching.
* Data engineering will be critical. Not only does it need to be very clean for good results, giving models unfettered access to the data needs the right access controls which, despite regulations like GDPR, are largely non-existent.
* Security in general will need to be upleveled everywhere. Not only can models be tricked, they can trick you into getting compromised, and so there need to even more guardrails.
A lot of these are regular engineering work that is being done even today. Only it often isn't prioritized because there are always higher priorities... like increasing shareholder value ;-) But if folks want to leverage the capabilities of AI in their businesses, they'll have to solve all these problems for themselves. This is a ton of work. Good thing we have AI to help out!
Building a mental model of a new domain by creating a logical model that interfaces with a domain I'm familiar with lets me test my assumptions and understanding in real time. I can apply previous experience by analogy and verify usefulness/accuracy instantly.
> Upshot: It took me months to figure out what worked for me, but AI enabled me to produce innovative (probably cutting edge) work in domains I had little prior background in. Yes, the hype should trigger your suspicions[...]
Part of the hype problem is that describing my experience sounds like bullshit to anyone who hasn't gone through the same process. The rate that I pick up concepts well enough to do verifiable work with them is literally unbelievable.
Maybe? Social proof doesn't mean much to me during a hype cycle. You could say the same thing about tulip bulbs or any other famous bubble. Lots of smart people with no stake get sucked in. People are extremely good at fooling themselves. There are a lot of extremely smart people following all of the world's major religions, for example, and they can't all be right. And whatever else is going on here, there are a lot of very talented people whose fortunes and futures depend on convincing everybody that something extraordinary is happening here.
I'm glad you have found something that works for you. But I talk with a lot of people who are totally convinced they've found something that makes a huge difference, from essential oils to functional programming. Maybe it does for them. But personally, what works for me is waiting out the hype cycle until we get to the plateau of productivity. Those months that you spent figuring out what worked are months I'd rather spend on using what I've already found to work.
One big problem with Claude Code vs Cursor is that you have to pay for the cost of getting over the learning curve. With Cursor I could eat the subscription fee and then goof off for a long time trying to figure out how to prompt it well. With Claude Code a bad prompt can easily cost me $5 a pop, which (irrationally, but measurably) hurts more than the one-time monthly fee for Cursor.
Learning how to use a tool once is easy, relearning how to use a tool every six months because of the rapid pace of change is a pain.
While I agree with the skepticism, what specifically is the stake here? Most code assists have usable plans in the $10-$20 range. The investors are apparently taking a much bigger risk than the consumer would be in a case like this.
Aside from the horror stories about people spending $100 in one day of API tokens for at best meh results, of course.
Anyway, if you've tried it and it doesn't work for you, fair enough. I'm not going to tell you you're wrong. I'm just bothered by all the people who are out here posting about AI being bad while refusing to actually try it. (To be fair, I was one of them, six months ago...)
A thing being great doesn’t mean it’s going to generate outsized levels of hype forever. Nobody gets hyped about “The Internet” anymore, because novel use cases aren’t being discovered at a rapid clip, and it has well and throughly integrated into the general milieu of society. Same with GPS, vaccines, docker containers, Rust, etc., but I mentioned the Internet first since it’s probably on a similar level of societal shift as is AI in the maximalist version of AI hype.
Once a thing becomes widespread and standardized, it becomes just another part of the world we live in, regardless of how incredible it is. It’s only exciting to be a hype man when you’ve got the weight of broad non-adoption to rail against.
Which brings me to the point I was originally trying to make, with a more well-defined set of terms: who cares if someone waits until the tooling is more widely adopted, easy to use, and somewhat standardized prior to jumping on the bandwagon? Not everyone needs to undergo the pain of being an early adopter, and if the tools become as good as everyone says they will, they will succeed on their merits, and not due to strident hype pieces.
I think some of the frustration the AI camp is dealing with right now is because y’all are the new Rust Evangelism Strike Force, just instead of “you’re a bad software engineer if you use a memory unsafe languages,” it’s “you’re a bad software engineer if you don’t use AI.”
People have all these feelings about AI hype, and they just have nothing at all to do with what I'm saying. How well the tools work have not much at all to do with the hype level. Usually when someone says that, they mean "the tools don't really work". Not this time.
I also think hype cycles and actual progress can have a variety of relationships. After Bubble 1.0 burst, there were years of exciting progress without a lot of hype. Maybe we'll get something similar here, as reasonable observers are already seeing the hype cycle falter. E.g.: https://www.economist.com/business/2025/05/21/welcome-to-the...
And of course, it all hinges on you being right. Which I get you are convinced of, but if you want to be thorough, you have to look at the other side of it.
But even if we look at your notion of stake, you're missing huge chunks of it. Code bases are extremely expensive assets, and programmers are extremely expensive resources. $10 a month is nothing compared to the costs of a major cleanup or rewrite.
But none of that really matters; I'm not so much engaging on the question of whether you are sold on LLM coding (come over next weekend though for the grilling thing we're doing and make your case then!). The only thing I'm engaging on here is the distinction between the hype cycle, which is bad and will get worse over the coming years, and the utility of the tools.
I think that is one interesting question that I'll want to answer before adoption on my projects, but it definitely isn't the only one.
And maybe the hype cycle will get worse and maybe it won't. Like The Economist, I'm starting to see a turn. The amount of money going into LLMs generally is unsustainable, and I think OpenAI's recent raise is a good example: round 11, $40 billion dollar goal, which they're taking in tranches. Already the largest funding round in history, and it's not the last one they'll need before they're in the black. I could easily see a trough of disillusionment coming in the next 18 months. I agree programming tools could well have a lot of innovation over the next few years, but if that happens against a backdrop of "AI" disillusionment, it'll be a lot easier to see what they're actually delivering.
I have no reason to care whether you use AI or not. I'm giving you this advice just for your sake: Consider whether you are taking a big career risk by avoiding learning about the latest tools of your profession.
AI company execs also pretty clearly have a politico-economic idea that they are advancing. The tools may stand on their own but what is the broader effect of supporting them?