Anthropic, too, Is Very Confused About What Its Business Model & Strategy Are: WAFFLE OF THE DAY
From $5 Ubers to metered intelligence: Anthropic’s pricing plan panic unfolds in real time. When your GPUs are maxxxing out and you can’t quite decide what to charge for any of it, it’s not just your users who are confused. The datacenters are running hot, the subsidies are being cut back, and the leading frontier lab is fumbling for direction in this world where “chat” is a toy, “reasoning “ is a semi-tool, and there is a hope that “agents” might be able to do some real work…
In May, we sent you an email announcing that starting today, the Claude Agent SDK, claude -p, and third-party apps built on the Agent SDK would stop drawing from subscription rate limits and move to a dedicated monthly credit. We’re writing to let you know that we’re not making this change today. We’re working to update the plan to better support how users build with Claude subscriptions.
What this means for you
Nothing changes for now. Agent SDK, claude -p, and third-party app usage continues to work with your subscription exactly as it did before today, and there’s no credit to claim. Your subscription limits are unchanged. When we have an update, we’ll share it with advance notice before it takes effect…
“Chat” LLMs are a party trick, useful for summarization, brainstorming, and “talk to the rubber duck”, but not for much else; “reasoning” LLMs are good as computer-programming assistants and make fewer really dumb or really self-contradictory mistakes; “agentic” LLMs can actually do some work—but if there is not a definite right answer and an error flag if it misses and needs to try again, you get yourself into big trouble.
A standard summarization task that I might assign to an LLM would, right now, if I sent it out to anthropic/claude-sonnet-4-7, cost me $0.50 in token charges in chat mode, $5 in reasoning mode, and $50 in agentic mode.
Nevertheless, there are enough people who are finding “agentic” workloads useful, or who are betting that they need to use frontier models intensively right now to build capabilities to manage what is coming for them, that data centers are working flat-out.
Frontier Labs, therefore, no longer have any incentive to charge less than marginal cost to their users: rather than creating sticky future customers, such teaser discounts right now simply lead to data center overload, congestion, and a very lousy user experience.
Hence Anthropic has decided to try to make money right now so that it can get a good IPO price.
And the hyperscalar platform monopolists, feeling that they have more running room as Anthropic is no longer trying to take all of their LLM customers right now, or following suit in an attempt to light less money on fire during the present gold rush.
No, I do not know what OpenAI thinks it is doing.
Now comes this interesting waffle. Anthropic, it seems, is no longer in the business of raising prices for “agentic” workflows in order to make total demand for its tokens low enough that its data centers have enough headroom to run smoothly and rapidly. Anthropic is not sure what business it is in right now.
If reading this gets you Value Above Replacement, then become a free subscriber to this newsletter. And forward it! And if your VAR from this newsletter is in the three digits or more each year, please become a paid subscriber! I am trying to make you readers—and myself—smarter. Please tell me if I succeed, or how I fail…
