Hacker News Reader

Comment by simonw | Hacker News Reader

One big difference I have with Ed Zitron is I look at companies suddenly getting worried that they're spending millions of dollars on tokens and think "wow those AI vendors are going to make SO MUCH MONEY".

The best price for a product is what I call the "suck air through your teeth" price. You want your customers to suck air through their teeth... and then pay the full amount anyway.

Uber set their per-developer token allowance to $1500 per developer per tool. That suggests to me that they think they can get at least that much ROI out of AI tooling.

Selling $1500/employee/month plans to companies is a great business to be in.

View on HN

Well hold on. "They" also thought tokenmaxxing was a good idea, until they didn't.

Whether companies pay will come down to the bottom line once the hype around the country club dies down a bit and for that it's still TBD on the actual product cost impact.

Yeah, tokenmaxxing was obviously stupid. Setting a limit is what happens when the adults in the room stop being stupid.

Not sure about that. As I replied elsewhere, it's a forcing function:

> People misunderstand the point of the tokenmaxxing time period, it was to force people to use AI so as to not have them stuck in their way, as some people are, and then to evaluate how it can help the company.

https://news.ycombinator.com/item?id=49047448#49047826

I expect everyone understood that.

The obvious second effect is that if you have a leaderboard people will compete to top it, by running dumb loops or whatever.

That impacts how much you can learn, and costs you more money than you wanted to spend.

Not everyone did understand that, you'll see some comments expressing that they've never heard that perspective before, interestingly enough. But yes Goodhart's Law applies.

The excessive spending was part of the point. The market is watching AI spend very closely. AI spend is a proxy for AI absorption. Expanding your MS contract to include Copilot for your rank and file doesn't move the meter much. Letting your engineers run up 100k/month shows extreme absorption. The share price increase potentially pays for the entire token bill once Wall Street goes through your earnings.

That is a truly commendable steelman. But let's be honest, there's basically no chance that was true. Management is frequently stupid enough to not think through the second order effects of their policies, and that is likely what happened this time as well.

At least at my workplace this was exactly what we were told, that it's a short term methodology to get people acclimated to thinking about what in their work could be automated, as honestly most do not think that way, they're not like engineers.

It's hard to read this take seriously, when you promoted a company spending $1k per engineer per day on LLM usage.

You mean Strong DM? I stand by what I said about them at the time: they're an interesting case-study in extreme AI-assisted software development, and they have a bunch of ideas that are worth learning from even if you're not going to spend that (absurd) kind of money.

One of the hot topics since then has been how to validate software without reviewing every line of code by hand, which is exactly the kind of thing they were exploring before anyone else.

Yes, Strong DM. How is that different from tokenmaxxing? Because of the article you wrote, a few companies went to them and starting adopting their methods and guess what happened in a few weeks into that process? Upper management started tracking engineers who weren't using AI enough (~1k a day). A few engineers got fired because they were slow to adopt StrongDM's process.

Have you really seen engineers fired for failing to spend $1,000 on tokens a day because their bosses read my piece about Strong DM? https://simonwillison.net/2026/Feb/7/software-factory/

I would hope that my commentary on those patterns would be less likely to lead to bad decisions than if people just read Strong DM's own description of their software factory approach: https://factory.strongdm.ai/

> Have you really seen engineers fired for failing to spend $1,000 on tokens a day because their bosses read my piece about Strong DM?

Bruh, come on. Not directly is the best answer I have for that. Unfortunately, they didn't read your article carefully, because the idea of productivity gains was all the candy they needed to jump in the van.

It sounds to me like you are blaming imagined job losses on anyone who wrote about an interesting experiment that a company did.

Damn, not blaming you for job losses. Indirectly influencing upper management isn’t your fault. That’s on them. I’m saying if you look at what strongdm preached it leads to tokenmaxxing.

this content was brought to you by *Insert Sponsor name*

"the last decision these people made was obviously stupid, but trust me bro that this one is smart"

Very true and I don't think you can completely discount the end consumer market either. A large number of the people not currently paying for a subscription are becoming dependent on AI enabled search results. I imagine monetization of those users is in its infancy and will rival the search result ad business eventually

that $1.5k is ostensibly not coming out of productivity gains from the AI though.

it's coming out of the salary budget from what i've seen. i've seen a company both say "AI max. it's the future! don't be late" and then go on to increase limits to compensation across the board.

Sure, but what you're not taking into account is that $1500/month is subsidised by VC money.

If Uber were asked to pay full price, they'd either get the same value for significantly more outlay, or keep their outlay capped but get significantly fewer tokens.

This would, I'm sure you'd agree, change the ROI outcome.

Zitron is arguing that if we all pay the true cost once the subsidies go, we'll either stop paying for the value we get today, or we'll find AI to be less useful than we do today, and those businesses won't be sustainable, and then <pop> goes the bubble.

I enjoy keeping an eye on https://isaiprofitable.com to see just where we are overall. Interesting reading if you like to think about such things.

I think the prices Uber et al are paying - the list prices for the API - pay out a healthy margin to the AI labs.

They're not profitable against the R&D spend, but they could be profitable if the AI companies attract more paying customers while keeping their model training costs static.

The growth in Anthropic's revenue over the past 12 months has been astonishing. There's real market demand for what they are selling.

Reply on news.ycombinator.com