Yeah, there's lots of stuff in Steam that I never use and don't even understand. Like the Steam Points and the Trading Cards and Steam Level and so on.
But the purchasing experience is top notch and they even have a generous refund policy. It's just lightyears ahead of the competition.
Yeah, they could have spent the money on founding new studios. With that sheer amount of money you could pretty much poach any talent you wanted too, but it wouldn't be as obviously anticompetitive.
Instead, they paid way over the odds for IPs that seem past their prime.
I lived there for around 6 months like 15 years ago so perhaps it's changed a lot since then.
But even as an Englishman, it was very different to home. I remember the supermarket was shut all Sunday and was only open until 12 on the Saturday, and it shut early in the week too (at like 5pm or 6pm or something?) so by the time I'd got the train back home from work it was already closed. I had to get up early every Saturday just to make sure I could get the shopping done.
I remember once I waved at my neighbours who were sitting eating in a common garden area and they acted super confused that I would wave to them.
It didn't seem like an especially friendly place and there were so many rules about everything too, like just being able to take the rubbish or recycling out you had specific days and times.
My company has a Claude Code and Codex one and I use Claude Code because I am more familiar with it. That said, I just use Opus for planning and Sonnet for implementation and it's pretty cheap. Codex seems decent too so I should try it out some more.
But you can get an awful lot done even with just like $200 a month at API pricing if you are careful not to waste a powerful model on an easy task, or carry around a bloated context window etc.
I think a lot of the 'tokenmaxxing' people spending thousands every month are simply using the tools ineffectively (like having loads of Opus agents doing tasks that Sonnet or even Haiku could do). I suspect this will only get worse now with the release of Fable, but Anthropic must love it.
When you say the cheaper models do you mean like Deepseek or GLM? I haven't tried those but they look interesting. It'd be nice to shift to open weights and not be tied to one company.
Thanks! I don't tend to use the GitHub discussion very much (just commenting on a few Issues) so I didn't know!
The Talkyard example is interesting - how does it differ to the algorithms Reddit uses to sort by "Best" or "Hot" when looking at comments (I think Best is the default now too, not Top, so presumably it takes recency etc. into account somehow).
The troubles over copyright infringement in AI training data remind me a bit of Eli Whitney and the cotton gin.
There he suffered massive patent infringement, that basically stopped being enforced due to the sheer economic importance of the cotton gin.
In a similar manner, I think there is a reasonably strong argument that it was wrong to use copyrighted material for AI training without paying royalties nor even asking for permission. But equally, every country wants to have the most powerful models and enforcing such royalties would make it effectively impossible to train them as the amount of material required would cost an insane amount in royalty fees.
So I expect the law will continue to turn a blind eye (perhaps enforcing some token payments like that $1.5B mentioned in the article) because "if we don't make these models, the Chinese will" etc.
Ignoring the bizarre inclusion of training compute for the AI company estimates, the other comparisons are still valid.
> The rest of the software market trails. The top 1% of companies spend $89k per engineer per year on AI, 40% of a fully-loaded $224k senior engineer salary. The median spends $137. That is the gap : ... 0.4x at the top of the market, near zero at the median.
So it's not more expensive than an engineer it's 40% as expensive, and for many companies use-cases the cost is virtually negligible.
Even here in Europe where developers are much cheaper than in the US, it still makes sense to pay for the LLM Enterprise subscriptions.
I wonder if a hybrid might work well - a Reddit/HN style system for comments, but a simple forum style method of post ranking by last activity. So if you make a comment on a post, the post goes to the top of the page.
This could work for comment threads too - where the comment threads on the post are also ranked by last activity.
It keeps the nice branching comment threads we've grown used to, but avoids having upvotes and downvotes and the opaque algorithm deciding what gets shown first (or at all).