I'll think about it if/when it happens - the space changes so quickly that whatever your plan is now, by the time you need the plan, everything will have changed so much that your current plan will be completely irrelevant.
And having to pay more and/or using slightly less capable models isn't a disaster, it's not like you need a frontier model for everything.
This was a more relevant conversation six months, since then OpenAI have been much more competitive on price for top models, Anthropic much less so but they're showing some signs of improvement in the 5.5 releases, and we've had an abundance of cheaper open source models.
Generally, I'm not and never was reliant on any of them to begin with. LLMs are just my rubber duck at best. That said, I never trusted OAI or Anthropic, let alone the absolute security hellhole those agents and harnesses are. With that in mind: everything local. Qwen 3.8 27B is phenomenal, and it runs on pretty much consumer hardware. There is a catch: my mistrust for the above mentioned and knowledge that the rug pull will inevitably happen made me buy an Asus GX10 a little after they came out - they were the cheapest by a country mile and overall 4000 bucks well spent there. For all I care, let OpenAI and Anthropic burn.
All these and many Chinese models are so close in capability anyway that I imagine the price war is gonna continue for a while, good for consumers to get cheap subscription tokens. Will probably be years before the dust settles, and then just switch over to the best value. Something like DeepSeek is more than good enough for most of my work today, just no reason to switch yet while subscriptions tokens are dirt cheap. There is no lock in anyway so no point in worrying about it now. I could literally switch tomorrow and wouldn't miss Anthropic/OpenAI much.
Yup! I've been working with Ollama for a while now and it really made me realize how OS models are quite good! (maybe not cutting edge but it's only a matter of time until they get incredibly good as well)
What about when it decides to ban specific things, like say writing gambling-related applications? Or partner with, say, an auth provider and decides that all Claude-generated user management now goes through their affiliate? Replace auth provider with any backend feature.
There is enough competition, so it is not likely that we will see a rug pull. Just switch to deepseek/qwen/glm. Chinese models are good, maybe 3mo behind Claude/GPT, and IMO at the same level as Gemini.
as someone who's been paying api prices for tokens to gcloud for the past few years (im a masochist), i'm sure access to ai is going to be metered once these companies stop their subsidised war for market share.
we should be prepared for that. a bunch of enterprises were shocked after github switched to usage based pricing for copilot lol, and now devs stuck w just $29 of spend for a month are just coping.
i think as long as the economic benefit to a person is higher than their inference cost, it's fine, but a lot of the wasteful burn of tokens on silly things will most likely end
will never happen, the labs already make a hefty profit on inference, and algorithmic improvements along with competition have driven prices down month after month
My best guess is that there will be a round of consolidation first --- possibly combined with direct market manipulation from government.
I don't believe AI only vendors can survive on their own over the long term under current conditions. They have no moat.
I look for the Mag7 to absorb individual AI vendors at fire sale prices once they start having trouble raising new capital to meet their already huge debt obligations.
And at some point, these same players will push to ban Chinese models (for "national security" reasons) and start to collude on pricing.
Their real long term objective is an effective monopoly on AI compute. Any plans you make now are unlikely to prove effective at that point. You won't be able to afford the hardware or power to run your own competitive model.
But don't worry, be happy --- and let the cheap tokens roll --- for now. And global warming is just fake news.
i really hope we find a new technology to run the models locally with less expenses
I'll think about it if/when it happens - the space changes so quickly that whatever your plan is now, by the time you need the plan, everything will have changed so much that your current plan will be completely irrelevant.
And having to pay more and/or using slightly less capable models isn't a disaster, it's not like you need a frontier model for everything.
This was a more relevant conversation six months, since then OpenAI have been much more competitive on price for top models, Anthropic much less so but they're showing some signs of improvement in the 5.5 releases, and we've had an abundance of cheaper open source models.
Generally, I'm not and never was reliant on any of them to begin with. LLMs are just my rubber duck at best. That said, I never trusted OAI or Anthropic, let alone the absolute security hellhole those agents and harnesses are. With that in mind: everything local. Qwen 3.8 27B is phenomenal, and it runs on pretty much consumer hardware. There is a catch: my mistrust for the above mentioned and knowledge that the rug pull will inevitably happen made me buy an Asus GX10 a little after they came out - they were the cheapest by a country mile and overall 4000 bucks well spent there. For all I care, let OpenAI and Anthropic burn.
All these and many Chinese models are so close in capability anyway that I imagine the price war is gonna continue for a while, good for consumers to get cheap subscription tokens. Will probably be years before the dust settles, and then just switch over to the best value. Something like DeepSeek is more than good enough for most of my work today, just no reason to switch yet while subscriptions tokens are dirt cheap. There is no lock in anyway so no point in worrying about it now. I could literally switch tomorrow and wouldn't miss Anthropic/OpenAI much.
Yup! I've been working with Ollama for a while now and it really made me realize how OS models are quite good! (maybe not cutting edge but it's only a matter of time until they get incredibly good as well)
What about when it decides to ban specific things, like say writing gambling-related applications? Or partner with, say, an auth provider and decides that all Claude-generated user management now goes through their affiliate? Replace auth provider with any backend feature.
There is enough competition, so it is not likely that we will see a rug pull. Just switch to deepseek/qwen/glm. Chinese models are good, maybe 3mo behind Claude/GPT, and IMO at the same level as Gemini.
as someone who's been paying api prices for tokens to gcloud for the past few years (im a masochist), i'm sure access to ai is going to be metered once these companies stop their subsidised war for market share.
we should be prepared for that. a bunch of enterprises were shocked after github switched to usage based pricing for copilot lol, and now devs stuck w just $29 of spend for a month are just coping.
i think as long as the economic benefit to a person is higher than their inference cost, it's fine, but a lot of the wasteful burn of tokens on silly things will most likely end
Celebrate!, because finally web can rest of all those "end of the world" and "end of SW/programming" statements made every week.
My plan would be to open a bottle of Champagne and celebrate the boost in job security.
will never happen, the labs already make a hefty profit on inference, and algorithmic improvements along with competition have driven prices down month after month
I'll be thankful that I already built the groundwork for various projects and the rest can be done on lower quality AIs.
My best guess is that there will be a round of consolidation first --- possibly combined with direct market manipulation from government.
I don't believe AI only vendors can survive on their own over the long term under current conditions. They have no moat.
I look for the Mag7 to absorb individual AI vendors at fire sale prices once they start having trouble raising new capital to meet their already huge debt obligations.
And at some point, these same players will push to ban Chinese models (for "national security" reasons) and start to collude on pricing.
Their real long term objective is an effective monopoly on AI compute. Any plans you make now are unlikely to prove effective at that point. You won't be able to afford the hardware or power to run your own competitive model.
But don't worry, be happy --- and let the cheap tokens roll --- for now. And global warming is just fake news.
not trying to be glib, but, this is an excellent question for a lab model to churn on for a bit.
it’s not just about what’s changing, but also what you specifically would need or want, and what you can afford to do or buy.
put differently, you’ll get anecdotes here, but what you need and what responders need is likely to differ.
“Best agentic coding harness,” for example. For what end?