They're subsidizing their tokens as long as you use their software. That's a fair exchange, I never understood why people took issue with it.
If you don't want to get locked in to Claude Code, you can pay more. Just like you can pay more for an unlocked non-carrier subsidized phone. (Which I personally do.)
Part of the point is that they've added all sorts of caveats - `claude -p` used to be "using their software", but isn't anymore, even though it's just invoking the same harness non-interactively
And their constant flipflopping is why they lost their Goodwill with me. It was too much work just to make sure I wasn't breaking their AUP week to week. They will say one thing and then directly contradict it in a social media post. I was constantly losing features that were the reasons I signed up with them.
The issue with this kind of subsidy is that it is designed to lock you in and not as a loss leader.
In other words, it is not like giving free nuts at bar to make you drink more beer but more like Nestle giving baby formula so that the babies get use to that instead of mothers milk
I think the subscription is also designed as a loss leader. The ai labs know the real money is in enterprise, and that enterprises mostly don’t want to use the most expensive option. How do they convince enterprises to sign up for their expensive api then? Give it to the employees for cheap so they tell bosses they want Claude code at work.
> What does it matter which tool I use when I hit the limit?
A third party harness may have a misconfiguration of prompt caching, leading to more load on Anthropic's servers, they could also have wildly different usage patterns (Hermes, openclaw etc), also making it hard to predict the load. If you constrain everyone to use your own harness, at least you're in control of that.
The first point is sort of funny though, because their own harness had a misconfigured prompt caching for several months during the times where they were crying about third party harnesses (/resume was busting cache all the time).
I want only their model, the rest is just garbage. I used Opus with pi.dev, worked perfectly fine. Fast and does exactly what it needs to do. Claude Code is slow, sluggish, buggy. Why are you forcing me to use it? Why should I pay an order of magnitude more to use pi.dev? That doesn't sound reasonable to me.
Well, they are wishing to lock you into their ecosystem, so that switching would be a high-friction decision. And then they can simply raise prices without you leaving.
You shouldn't have to pay your carrier to unlock your phone and in many countries it's actually illegal.
>if you don't want to get locked in to Claude Code
All of these companies are built around eventually locking you in then selling to some other company. "Just pay more" is not a real solution. Once the competition dies down and there are only a handful of companies (if we are even lucky enough for that) with viable LLM products, your ability to jump ship is going to diminish. It always does. There is no way they aren't working on better lock in tactics.
> and there are only a handful of companies (if we are even lucky enough for that) with viable LLM products, your ability to jump ship is going to diminish
The global LLM market seems trending towards fragmentation. And lock-in appears to be diminishing, not increasing, particularly for companies that set up a multi-model workflow.
> Wait until investors stop giving out money out of FOMO
OpenRouter isn’t subsidizing much. No subsidy would mean the subscriptions go away in favor of API only. That doesn’t really speak at all to either of your points regarding consolidation or lock-in.
> They're subsidizing their tokens as long as you use their software. That's a fair exchange, I never understood why people took issue with it.
Except they themselves cannot afford it and as the article also mentioned, their level of service and uptime is atrocious for the price they are offering.
Their unpredictable pricing model is designed to make it easy for you to lose more money without you knowing. Unannounced model switching, nerfing old/new models, classifiers detecting 'unsafe' instructions to downgrade the selected model and overcharging API prices and locking subscriptions to a broken vibe-coded harness.
You don't get anything by defending Anthropic, given their past behavior.
Yes, the subsidized loss leader product is evolving fast. If you want stable costs, the API plan is right there. You simultaneously say they cannot afford the API pricing but also claim they are overcharging, it sounds like you just want to get mad ad Anthropic.
> If you want stable costs, the API plan is right there.
Why don't you ask Microsoft [0], Tesla [1] and Alibaba [2] on why they believe it is not only expensive, but it also a security risk for them?
Like I said, You don't get anything for defending Anthropic and they are not going to hire you into their sales team.
> You simultaneously say they cannot afford the API pricing but also claim they are overcharging, it sounds like you just want to get mad ad Anthropic.
Read it again. "Except they themselves cannot afford it"
They is "Anthropic" given that I said "themselves". Who are the ones that cannot afford to subsidize tokens forever due to the even cheaper Chinese alternatives. They all know that the answer is open weight models and why Anthropic cannot raise prices and why the above companies are ditching them for open weight models.
I mean, it has definite Borg Bill Gates vibes, doesn't it?
It's totally reasonable for tech folks to be leery of a company using a strong market position from one system (Windows/Claude) to push mediocre complementary software (IE/CC).
That analogy only works partially, because when IE6 was released, it was the best browser by far. IE only became terrible once MSFT actively stopped developing it, and other browsers kept getting better.
On the other hand, Claude Code was the best coding agent when it was released, but there's no way Anthropic is going to let its cash cow stagnate. Like, I think pretty much all of Anthropic's revenue spike in the last few months was driven by the tokenmaxxing mania.
My take is most of Claude Code's problems originate from insufficient compute capacity and all kinds of workarounds they're doing to mitigate that fundamental limitation.
If that were the only issue, I’d say Anthropic were fine.
Dario took the tremendous goodwill his row with Hegseth gave him and blew it. The major problems are Anthropic’s back-end instability and front-end kludginess (Electron!) revealing the gap between Anthropic’s capability and marketing, and Dario bizarrely copying Altman’s 2023 fire-and-brimstone playbook that had already massively backfired.
Using Electron in this day and age shows a particular lack of effort, considering how easy it is to use an agent harness to build native versions of apps. I mean, we're a tiny little business and we somehow find the bandwidth to do it.
Anthropic being a big company means they could trivially hire an iOS team. The failure to make a good native app is a business decision. Claude or not, leadership could've snapped their fingers, allocated the budget, and fixed the frontend UX by now if it were important enough.
> Anthropic being a big company means they could trivially hire an iOS team
Anthropic is primarily limited by how fast they can identify and onboard people who will be a good fit. Allocating that to ios would trade off against allocating that staffing growth elsewhere (ex: reliability, making the next generation model even better).
It just seems like Anthropic has made a deliberate choice to make themselves a TypeScript dev shop.
From Claude Code being written in React (!!) to acquiring the Bun people so they they could have their very own bespoke JS/TS runtime (and then... rewriting the whole thing), to having the CC "native" app being Electron, etc.
It just seems like it comes down to a choice at the highest level to hire people from a "full stack" background, and build in and and prioritize TypeScript as their dev language, for better or for worse.
I personally think it's worse and that shows in the performance and quality of CC, but I'm biased.
Codex, FWIW, is written in Rust, and it is by far snappier and more reliable, though less featureful for now.
Interestingly I just skimmed through a video linked in TFA: https://www.youtube.com/watch?v=SlGRN8jh2RI and Boris Cherny explains that they chose TypeScript and React for Claude Code primarily because it was "on distribution" and the models back then just weren't good enough at other languages.
So it's interesting that Codex is written in Rust. Amongst other things it could mean OpenAI had more powerful models that could handle Rust, or their engineers had to handhold the agents a lot more up front, or Rust has structural advantages that could overcome being less represented in the training data.
Codex also I think put its first version out a bit later?
IMHO building the agent harness at least partially by hand is kind of key? I think going hogwild with "vibe coding" the actual harness is partially why CC is a bit of a rats nest. I think it's genuinely gotten out of control in terms of their ability to reason about it. Just my opinion, I'm sure Boris disagrees.
Also Codex itself distributed via NPM same as Claude. But it distributes a compiled Rust binary. Which is compiled from open source.
I used to think Claude Code was released much earlier too, and my initial theory was that OpenAI as a follower had the benefit of more powerful models... but when I looked it up, Codex was first released in April 2025 [1], whereas CC Beta was released in Feb 2025 [2] -- only a couple of months apart!
I'm sure each lab is keenly aware of what the other is doing (how else could they time so many of their releases so close to each other?) so it's highly probable they started developing each app about the same time, and likely even knew the technical details involved. Which is why it's additionally interesting that OpenAI started off with Rust while CC used Electron.
I don't understand why people care about Electron? Evaluate the quality of the software based on its own merits IMO. VS Code, Slack, Postman, Obsidian.. I use this software every day and my only complaint is Slack's RAM usage (which honestly has no negative effect on my machine, but just seems silly for a messaging app).
Claude is a buggy mess because it's slop, not because it's electron. Heck, it runs a full Linux VM under the covers without asking.. it's insane.
> don't understand why people care about Electron?
Most Electron apps are crap because they transparently communicate the developer prioritizing their own efficiency over the trade-offs to UX choosing a cross-platform platform entails.
In most cases, I don’t care that an app looks crappy as long as it works. For Anthropic, however, the irony undermines its argument that Claude is ready to replicate developers.
Native apps suck too. i cannot list 10 native mac apps that are truly great, including Apple’s own.
To me these days native on desktop means web, not Appkit. I can resize the whole interface with a common shortcut, I can select text trivially, native Mac apps constantly fail to do both.
Developing native apps for the Mac is a mess, with immature SwiftUI, the absolute garbage liquid glass is on the desktop, the outdated Appkit framework and iOS based Mac Catalyst.
Electron is a sane choice and good enough for Figma, Slack, Notion, leading apps in their respective categories.
Yet… VS code somehow became the most popular IDE that happens to be generally viewed favorably by most of its users.
I personally hate it due to various reasons but I don’t feel like the Electron part is the issue (Java based Jetbrains IDEs generally seem way more bloated)
I'll say their poor Chat product is the worst. Its been having this issue of losing connection will kill the chat session since day 1, and never bother fix it.
Their super app is oversized, but that's the MS playbook with Teams, overusing resource to justify upgrade budget from IT dept, typical enterprise sales trick.
>... Dario bizarrely copying Altman’s 2023 fire-and-brimstone playbook that had already massively backfired.
I've said this before, they always knew it was terrible marketing, but they just can't help themselves because they actually believe it.
From multiple accounts, the people working at these labs, who are most exposed to the latest models' capabilities and how they're being used out in the world, are simultaneously excited and terrified about what they're building.
In a way that's even scarier than the "Capitalist sociopaths marketing AI to other Capitalist sociopaths" rationale everybody assumes.
I'm not sure the security/safety stuff is entirely in their control at this point, though you can argue that they are indirectly responsible through encouraging regulation via their positions on safety/risk.
How does it stifle competition / innovation? You can use other harnesses, pay API for anthropic models or cheap chinese models.
You're just upset that you don't get to use their VC money the way you want to use it lol. It either is a good deal or it isn't, if it is just say, thanks free VC money, if it isn't say fuck you.
Because the US models haven’t had to compete on the actual price of providing the unsubsidized service, the Chinese models are probably ahead of us in terms of what can actually be delivered profitably (which is a pretty bad result given the level of investment).
>Any service or product delivered at a loss seems pretty plainly anti-competitive.
You have to get into the weeds though on what exactly counts as a "product delivered". Like, Apple doesn't charge for new versions of macOS. But are "Macs" separate stacks together or are they the fusion of hardware and software, and if so on what levels? There are all sorts of products surrounding us that are running software that we still treat as unified objects after all, right down to smart light bulbs or a "plain" lithium battery pack which still requires controllers to manage charging and USB negotiation etc. Chips and software are in tons and tons of "basic" hardware stuff yet we just buy the object as a singular entity.
I'm not saying that dumping can't be a thing but the lines aren't always clear cut either. You also have to get into bog standard business scaling issues and profit vs investment. If you take on debt to invest in capital that you believe will lower per unit costs if you build enough volume and then turn a profit, you're "selling at a loss" but that's how tons of business works, that's why there is risk right? Doesn't seem good to discourage that.
It seems more fruitful to approach things from the perspective of monopolies, competition, corporate governance etc in general, granted not that a lot of governments have been great about that either in recent history.
Well if their gross margins are positive (I really doubt they aren’t) there isn’t much of a case to be made. You can expect meaningful R&D and capital investments to pay off short term and you can’t make any meaningful projections either to determine if they are dumping or not.
The entire LLMaaS industry is priced below cost. Even if the marginal cost of electricity and bandwidth to produce one token is less than the price of that token, the amortized infrastructure and R&D costs make the entire venture unprofitable.
If pricing below cost were illegal, it would essentially make starting up a business—any business—impossible.
You can’t calculate what that cost is though regardless. If the gross margin is positive you can’t confidently challenge their longterm growth projections however optimistic or unrealistic they are.
the fact that there are 10+ harnesses means you are wrong.
Also, if other harnesses need to work without the literal SOTA then they need to be better than Claude Code. If their incentive is to reduce token usage then again, they will evolve to be better than CC
Do you also think all you can eat buffets should be illegal? What about early bird specials? Free soda with a large fry? Happy hour wings? Where does the “bundling” end?
1.) All you can eat buffets ARE the product, not a loss leader, so there's no bundle there.
2.) Early bird specials, again, aren't loss leaders. They're a way to get sales on the left side of the bell curve. The restaurant food is still the product, not a subsidy.
3.) Free soda with large fry doesn't actually exist (though it sounds like it should, right?)
4.) Happy hour wings (again, see #2)
None of these are bundling so I don't really get why you bothered to comment
I just searched and found free medium fries with a large soda at McD if you order through the app. Seems like bundling to me, and you have to use their particular app.
Sure, and perhaps it should be prohibited. But it's also a matter of impact.
But it's just hard to care too much about a commoditized product and an app with tons of competition. And McDonald's is not keeping any competitors away by locking their fries offer behind their app.
Agreed. The level of entitlement around how we all deserve to use a product in a category that didn't even exist 12 months ago with increasing performance at the same price forever is insane.
Consumers want to pay as little as possible and companies want to charge them as much as they possibly can. That’s how markets work I don’t understand what does “entitlement” have to do with anything. If loud consumers somehow manage to coerce companies into lowering prices/offering better products that’s a massive win for almost everyone (of course usually its just noise that doesn’t change anything, however it did work on a few occasions).
The the article was talking about how Anthropic costs too much and how there are cheaper options available then yes I would agree with you. But this article has a conspiratorial and moral crusade tone which is what I'm responding to. The article says "In my opinion (which may be in the minority), it’s unethical to:
lock in your customers to a closed system for maximum market gain
put down the competition when they pose a risk to your product
make hypocritical claims on how your product increases quality when your own software sucks
artificially restrict one’s own product as a fear-mongering marketing stunt
test dynamic pricing on your users to see how much more they’ll pay for less
change the terms of your product after the sale without notifying your user base"
anthropic is not just a business operating in a market, they are "unethical" and "fear-mongering" and "hypocritical".
I also think that the people who are complaining loudest about Anthropics pricing and changes are the customers Anthropic is least interested in retaining.
From my experience people who complain loudly about the price tend to be people who are either (1) working primarily on hobby projects so are unwilling to pay much for Claude Code or (2) using astronomical amount of tokens through elaborately orchestrated multi-agent setups that only make sense when someone else is paying the bill.
Short term yes. But there is a reason why IDEs, game engine and other development tools have very high cheap or free versions for this market segment. Since these users end up having a huge influence on what companies they work at / end up working at spend their money on.
Of course LLMs are a commodity at this point but if someone is using Codex, Pi etc. at home it becomes more likely they won’t be picking Claude Code at their day job either.
C'mon, get real. The monthly subscription costs on these services are several multiples what software engineers have formerly paid for development tools.
I have worked in shops that made it difficult for me just to get a paid for RustRover or CLion license, and had to out of pocket it. The monthly sub for Claude or Codex is equivalent to the total license cost for that.
It's not entitled to expect your $200/month tooling expense to have a level of reliability and consistency.
If your employer is paying for it, fine. I am my own employer. I don't fancy becoming dependent on tooling that costs me as much as my family's whole monthly mobile plan and then having it degrade in unpredictable ways or rug pull me on quota. Or have geo-political drama as part of its release cycle.
If you don't want to get locked in to Claude Code, you can pay more. Just like you can pay more for an unlocked non-carrier subsidized phone. (Which I personally do.)