3 ms·

This gives me hope that when the subsidization circus ends and everyone is on pure usage then it won't be entirely exclusionary to mere mortals who don't have $

by Havoc 6mo ago

This gives me hope that when the subsidization circus ends and everyone is on pure usage then it won't be entirely exclusionary to mere mortals who don't have $200pm budgets.

542458 6mo ago
IMO there are two things that make me optimistic that we won’t see a big rug pull where price-to-capability ratio skyrockets relative to today: * As you’ve noted, people keep finding ways of slamming more intelligence into smaller models, meaning that a given hardware spec delivers more model capability over time. * Hardware will continue to improve and supply will catch up to demand, meaning that a dollar will deliver more hardware spec over time. I hope that one day we’ll look back on the current model of “accessing AI through provider APIs” the same way we now look back on “everyone connecting to the company mainframe.”
- spacebanana7 6mo ago
  I also hope that we’ll find effective ways to distribute load between small local models and heavyweight remote models. Sort of like what Apple tried to do in iOS. So much of what I ask codex to do doesn’t require full GPT 5 intelligence, and if 75% of the tokens were generated locally that’d save a massive amount of cost.
100ms 6mo ago
By the time the dust settles I wouldn't be surprised if personal interactive usage couldn't even be had for under $200. I can't fit my modelling of the serving costs of these things to any public reporting, even the more bearish examples
- Havoc 6mo ago
  Comes down to what you mean by interactive usage. Most of chat & say openclaw usage is already within self-host range so no need to spend 200 a month on that. High end SOTA coding is harder, but even there I suspect a mix of usage based strong models and selfhost small is viable if necessary.
- pimeys 6mo ago
  We pay per token in our company. It is not hard to spend $100 for one morning coding session. So thousands per month per programmer. The company finds it valuable enough to pay for, but if I ever paid these from my own pocket I'd look into DeepSeek et.al.
- jerojero 6mo ago
  Not a lot of people have this budget, and I'm not sure how many people with that type of cash are also interested in paying it for AI. Of course, this is fine for people in the bay area earning hundreds of thousands of dollars a year. But then your client base becomes so reduced its hard to justify the valuation these companies have. These AI companies are not hyped so much because they will offer a luxury product, they're valued because they're supposed to "change the world" which luxury does not do.
itemize123 6mo ago
it will be more expensive when subsidization ends? how is it going to be more inclusive?

by Havoc 6mo ago

This gives me hope that when the subsidization circus ends and everyone is on pure usage then it won't be entirely exclusionary to mere mortals who don't have $200pm budgets.

542458 6mo ago

IMO there are two things that make me optimistic that we won’t see a big rug pull where price-to-capability ratio skyrockets relative to today: * As you’ve noted, people keep finding ways of slamming more intelligence into smaller models, meaning that a given hardware spec delivers more model capability over time. * Hardware will continue to improve and supply will catch up to demand, meaning that a dollar will deliver more hardware spec over time. I hope that one day we’ll look back on the current model of “accessing AI through provider APIs” the same way we now look back on “everyone connecting to the company mainframe.”

spacebanana7 6mo ago
I also hope that we’ll find effective ways to distribute load between small local models and heavyweight remote models. Sort of like what Apple tried to do in iOS. So much of what I ask codex to do doesn’t require full GPT 5 intelligence, and if 75% of the tokens were generated locally that’d save a massive amount of cost.

100ms 6mo ago

By the time the dust settles I wouldn't be surprised if personal interactive usage couldn't even be had for under $200. I can't fit my modelling of the serving costs of these things to any public reporting, even the more bearish examples

Havoc 6mo ago
Comes down to what you mean by interactive usage. Most of chat & say openclaw usage is already within self-host range so no need to spend 200 a month on that. High end SOTA coding is harder, but even there I suspect a mix of usage based strong models and selfhost small is viable if necessary.
pimeys 6mo ago
We pay per token in our company. It is not hard to spend $100 for one morning coding session. So thousands per month per programmer. The company finds it valuable enough to pay for, but if I ever paid these from my own pocket I'd look into DeepSeek et.al.
jerojero 6mo ago
Not a lot of people have this budget, and I'm not sure how many people with that type of cash are also interested in paying it for AI. Of course, this is fine for people in the bay area earning hundreds of thousands of dollars a year. But then your client base becomes so reduced its hard to justify the valuation these companies have. These AI companies are not hyped so much because they will offer a luxury product, they're valued because they're supposed to "change the world" which luxury does not do.

itemize123 6mo ago

it will be more expensive when subsidization ends? how is it going to be more inclusive?