They'll all do this eventually.

We're in the part of the market cycle where everyone fights for marketshare by selling dollar bills for 50 cents.

When a winner emerges they'll pull the rug out from under you and try to wall off their garden.

Anthropic just forgot that we're still in the "functioning market competition" phase of AI and not yet in the "unstoppable monopoly" phase.

"Naveen Rao, the Gen AI VP of Databricks, phrased it quite well:

all closed AI model providers will stop selling APIs in the next 2-3 years. Only open models will be available via APIs (…) Closed model providers are trying to build non-commodity capabilities and they need great UIs to deliver those. It's not just a model anymore, but an app with a UI for a purpose."

~ https://vintagedata.org/blog/posts/model-is-the-product A. Doria

> new Amp Free (10$) access is also closed up since of last night

Not going to happen.

Unstoppable monopoly will be extremely hard to pull off given the number of quality open (weights) alternatives.

I only use LLMs through OpenRouter and switch somewhat randomly between frontier models; they each have some amount of personality but I wouldn't mind much if half of them disappeared overnight, as long as the other half remained available.

I'm old, so I remember saying the same thing about Google and search.

I hope you're right!

I think the big difference is that Google is free: everyone is using Google because it doesn’t cost anything and for a long time was the best search engine out there. I am sure that if Google would suddenly charge a few dollars per month for access, Bing market share would explode overnight, because it would become “good enough but cheaper”.

With the AI models, using a model that is “good enough but cheaper” is already an option.

There's no reason that a sizeable portion of LLM usage can't and won't end up free/ad-sponsored. Cutting edge stuff for professional use will probably be monetized via subscription or API credits for a long time to come. But running an older and less resource intensive model works just fine for tasks like summarization. These models will just become another feature in a "free" product that people pay for by watching or clicking ads.

I imagine the split will look a lot like b2b vs b2c in other technologies, b2b customers tend to be willing to pay for tech when it offers a competitive advantage, reduces their operating costs etc. b2c customers mostly just guzzle free slop.

It's actually pretty bonkers when you think about how basically every cutting edge professional you deal with is getting ads for all of their top search results for all of their work.

(Not quite "every", but outside of tech, most professional workplaces don't support ad blocking or Kagi.)

I too am old. Google search is free, hard to replicate, and while there used to be lots of search engines, Google was (and arguably still is) miles ahead of all the others in terms of quality and performance.

A model is hard to train but it doesn't need to be hyper up to date / have a new version come out every day. Inference is cheap (it seems?) and quality is comparable. So it's unclear how expensive offerings could win over free alternatives.

I could be wrong of course. I don't have a crystal ball. I just don't think this is the same as Google.

Of course I could be entirely mistaken and there could emerge a single winner

I would say Google's monopoly mainly comes from its name recognition, definitely not because its still ahead in core search as I have been using DuckDuckGo for 2 years once I noticed search results are the same or better than Google.

In the first years, I remember no other search engine was close to Google quality. We all ditched AltaVista because Google was incredibly better. It would have been awful to switch back to any other options. We can already switch between the 3 big proprietary models without feeling too much differences, so it’s quite a different landscape.

Yes, my point exactly.

There was never a high quality alternative to Google search, until Kagi - and even that isn't free!

This is saying we have hundreds of open source OSes and Windows will never be a monopoly.

Software always gets monopoly simply by usage. Every time a model gets used by esoteric use cases, it gets more training data (that a decentralized open weight model doesn't get) and it starts developing its moat.

I think windows has historical monopoly.

They bundled it with PC hw and the vast majority of apps only ever got published for windows, and this over decades (one would argue it’s still true).

The starting point for LLMs is very different. Who would publish today a software that only integrates with chatGPT? Only a small minority.

Thus I agree, I struggle to see how a monopoly can exist here. A GPU monopoly or duopoly though, perhaps.

It’ll be a bunch of tiny moats in that scenario. LLMs are way too generic, adaptable, flexible in how you use it to make a big most out of it.

> Software always gets monopoly simply by usage

Most software isn't made by monopolies. More directly, enterprise-software stocks are getting hammered because AI offers them competition.

It's more like saying AWS has a monopoly on virtual machine hosting.

(For those unaware, AWS doesn't have a VM monopoly, and the market dynamics seem similar)

>This is saying we have hundreds of open source OSes

we don't, we have about 3 operating systems that have the decades of hardware and software compatibility that makes them widely usable. They're the most complex and complicated things we've built. LLMs are a few thousand lines of python hooked up to a power plant and graphics cards. This is the least defensible piece of software there ever has been.

OpenRouter falls in the acceptable use category. They targeting users that are misusing their Claude OAuth token on non-Anthropic products.

They will [try to] ban open weights for ethics / security reasons: to stop spammers, to protect children, to stop fascism, to defend minorities. Take your pick; it won't matter why, it will only matter which media case can they thrust in the spotlight first.

Yes of course they will; the CEO of Anthropic makes that argument, very openly, all the time. But it will be hard to do, I think.

Hopefully, but they'll lobby hard, supported by all the money they raised.

> They'll all do this eventually

And if the frontier continues favouring centralised solutions, they'll get it. If, on the other hand, scaling asymptotes, the competition will be running locally. Just looking at how much Claude complains about me not paying for SSO-tier subscriptions to data tools when they work perfectly fine in a browser is starting to make running a slower, less-capable model locally competitive with it in some research contexts.

Imagine having a finite pool of GPUs worth more than their weight in gold, and an infinite pool of users obsessed with running as many queries against those GPUs in parallel as possible, mostly to review and generate copious amounts of spam content primarily for the purposes of feeling modern, and all in return for which they offer you $20 per month. If you let them, you must incur as much credit liability as OpenAI. If you don't, you get destroyed online.

It almost makes me feel sorry for Dario despite fundamentally disliking him as a person.

Hello old friend, I've been expecting you.

First of all, custom harness parallel agent people are so far from the norm, and certainly not on the $20 plan, which doesn't even make sense because you'd hit token limit in about 90 seconds.

Second, token limits. Does Anthropic secretly have over-subscription issues? Don't know, don't care. If I'm paying a blistering monthly fee, I should be able to use up to the limit.

Now I know you've got a clear view of the typical user, but FWIW, I'm just an aging hacker using CC to build some personal projects (feeling modern ofc) but still driving, no yolo or gas town style. I've reached the point where I have a nice workflow, and CC is pretty decent, but it feels like it's putting on weight and adding things I don't want or need.

I think LLMs are an exciting new interface to computers, but I don't want to be tied to someone else's idea of a client, especially not one that's changing so rapidly. I'd like to roll my own client to interface with the model, or maybe try out some other alternatives, but that's against the TOS, because: reasons.

And no, I'm not interested in paying metered corporate rates for API access. I pay for a Max account, it's expensive, but predictable.

The issue is Anthropic is trying for force users into using their tool, but that's not going to work for something so generic as interfacing with an LLM. Some folks want emacs while others want vim, and there will never be a consensus on the best editor (it's nvim btw), because developers are opinionated and have strong preferences for how they interface with computers. I switched to CC maybe a year ago and haven't looked back, but this is a major disappointment. I don't give a shit about Anthropic's credit liability, I just want the freedom to hack on my own client.

You're not "rolling your own client." You're using a subscription that prices in a specific usage pattern, the one mediated by their client, and trying to route around it to extract more value than you're paying for. That's not hacking, it's arbitrage, and pretending it's about editor philosophy is cope.

Anthropic sells two products: a consumer subscription with a UI, and an API with metered pricing. You want the API product at the subscription price. That's not a principled stance about interface freedom, it's just wanting something for less than it costs.

The nvim analogy doesn't land either. Nobody's stopping you from writing your own client. You just have to pay API rates for it, because that's the product that matches what you're describing. The subscription subsidises the cost per token by constraining how you use it. Remove the constraint, the economics break. This isn't complicated.

"I don't give a shit about Anthropic's credit liability," right, but they do, because it's their business. You're not entitled to a flat-rate all-you-can-eat API just because you find metered pricing aesthetically displeasing.

It seems to me the consumer subscription still has metered token limits (it's not all you can eat), so why should it matter how we use those tokens?

As my sibling mentioned, it's not all-you-can-eat, it's metered and capped. Is it the client mediating usage patterns? I don't know, but why not do it server-side and let people use it however they want? Control of course, but that's not in the interest of users.

I'm not trying to arbitrage or route around anything, I just want predictable billing to access the model. Maybe the API would be cheaper for me, I don't know. I'm just a normal user, not scheduling an agent army to blast at maximum 24/7.

You don't need to explain what Anthropic is selling, I get it, but you're off-base claiming that I'm pretending about editor philosophy as cope. I think Anthropic is where they are today because they have a good model for coding, made popular by software folks who value things like editor choice and customization. Anthropic is free to do as they wish of course, as am I, but I'm displeased with their decision here, and voicing my opinion about it.

If usage is truly constrained by the client not the server, I guess I can't argue that, but it still feels bad as an end user. As a consumer, I just want a fair deal and freedom to use what I purchase however I see fit. But that seems harder to find these days, and most businesses seem intent on maximum extraction by any means possible. I might be wrong, but this feels like business move to build a consumer moat by controlling the interface, because consumers don't want the API. It's not in my best interests, which alienates me as a customer.

Why do you fundamentally dislike him as a person?

The only thing I've seen from him that I don't like is the "SWEs will be replaced" line (which is probably true and it's more that I don't like the factuality of it).

It’s kinda obvious he’s a well spoken shark. Personally not an issue for me, you have to be at the top of a unicorn, but it isn’t something people in general like.

Interesting, are there any sources for the shark claims ? I recently saw an interview with Hassabis and him and thought: <well at least those are two actual scientist leading AI labs/devisions>, so that gave me some hope that what they discussed regarding security and eventual equal distribution of "AI" benefits had some genuine intention.

Don’t be mad at it, be happy you were able to throw some of that sweet free vc money at your hobbies instead of paying the market rate.

Oh I'm not mad, it's more of a sad clown type of thing. I'm still stoked to use it for now. We can always go back to the old ways if things don't work out.

They offer an API for people who want to build their own clients. They didn't stop people from being able to use Claude.

at a significantly higher price... which of course is why they're doing this.

That's what the API is for.

Sure they are having a blast, they are paying 20$ instead of getting charged hundreds for forr tokens.

It's simple, follow the ToS

So basically you are saying Anthropic models are indispensable but you are too cheap to pay for it.

Nowhere did I say they're indispensable, and I explicitly said I'm still paying for it. If all AI companies disappear tomorrow that's fine. I'm just calling out what I think is tone-deaf move, by a company I pay a large monthly bill to.