Anthropic thinks they're Apple. They're actually hypocrites.
TL;DR
Theo says Anthropic is enforcing the exact lock-in behavior they’re now celebrating bypassing at Apple — he points to Anthropic promoting an iMessage plugin that appears to violate Apple’s terms just days after pressuring Open Code to remove a Claude Code subscription plugin.
The real fight is over the Claude Code subscription endpoint, not the CLI itself — Theo argues users paying $200/month are often paying for model access and compute, not loyalty to the Claude Code harness, which is why they want to use that subscription through OpenClaw, VS Code, or T3 Code.
Anthropic’s policy is vague in public and strict in practice — he cites the Claude Code docs saying OAuth tokens from Claude Code and Claude.ai are only permitted inside Anthropic’s own products, while developers like Matt PCO have spent weeks trying and failing to get a clear answer on what wrappers are allowed.
Theo draws a surprisingly pro-Apple comparison: Apple’s wall is at least consistent and legible — his point is not that Apple is good, but that Apple clearly says iMessage is for its own app and devices, while Anthropic is simultaneously policing others for API misuse and shipping a product that does the same kind of workaround.
Anthropic is more closed than peers like OpenAI, GitHub, and Kilo on subscription portability — Theo contrasts Anthropic’s lock-in with OpenAI letting people use subscriptions and models across harnesses like OpenCode or OpenClaw, and with GitHub Copilot and other tools being more permissive.
The bigger accusation is cultural, not just legal — Theo says Anthropic’s communication feels intentionally ambiguous, anti-user, and condescending, creating a system where developers are scared of arbitrary bans while Anthropic reserves the freedom to draw the line whenever it wants.
The Breakdown
The iMessage plugin that set Theo off
Theo opens in full “Anthropic’s number one hater” mode, reacting to an Anthropic employee promoting an iMessage plugin for Claude Code. His core issue: Anthropic is advertising something that appears to break Apple’s terms of service while it has been aggressively policing third-party use of its own systems. He frames this less as a legal gotcha than as a pure hypocrisy story.
A quick detour: sponsor read, then an even funnier Anthropic leak detail
After a WorkOS sponsor segment, Theo jumps back in with an update from “Theo from the future.” He notes that in the recent Claude Code source leaks, people found references to Open Code in Anthropic’s codebase for things like autoscroll behavior and window sizing. He treats it as one more deliciously petty example of Anthropic borrowing from the same ecosystem it keeps threatening.
What people are actually paying for with Claude Code
Theo then slows down and explains the mechanics: Claude Code is just the harness, while the real economic value is the special subscription-backed endpoint behind it. He argues many users pay the $200/month Max tier for access to Anthropic’s coding models and compute, not because they’re married to the Claude Code interface itself. That’s why developers want to use the same subscription through other tools like OpenClaw, VS Code, or T3 Code.
Why Open Code got pressured and why developers are scared
He recounts how Open Code had to remove a plugin that let users route Claude Code subscription access through a different interface. Theo says Anthropic’s docs explicitly forbid using OAuth tokens from Claude Code or Claude.ai in any other product, but the company’s policy language keeps shifting and still feels murky. The result, in his telling, is a climate where developers are forced to tiptoe around arbitrary enforcement because getting banned is a real possibility.
Theo’s Apple rant, which turns into an Apple defense
This is where the video gets fun: Theo spends several minutes both praising and attacking Apple. He swats away common anti-Apple talking points like planned obsolescence and “overpriced hardware,” then pivots hard to say Apple’s App Store behavior can be borderline extortion. The point of the detour is to establish that he’ll defend Apple when criticism is lazy and attack it when criticism is deserved.
The Beeper fight and the principle Theo thinks matters
Using Beeper and iMessage as the bridge, Theo makes his actual principle clear: companies should not be legally forced to expose internal APIs for products they don’t want to build. Apple wants iMessage to work through Apple’s own trusted devices and apps; Anthropic wants Claude Code subscriptions to work through Anthropic’s own interfaces. On that narrow question, he says both companies are within their rights.
So why Apple gets a pass here and Anthropic doesn’t
Because, in Theo’s view, Apple is at least coherent. Apple clearly says you can’t reverse-engineer or automate iMessage access, while Anthropic is still dodging straightforward questions from people like Matt PCO about what wrappers and local tools are allowed. Theo’s big punchline is that Anthropic is now doing to Apple exactly what it just punished others for doing to Anthropic.
The closing argument: Anthropic is more locked down than Apple
Theo ends by saying Anthropic has created a harsher walled garden than the company everyone loves to call a walled garden. OpenAI, GitHub, Kilo, and others, he says, are more willing to let users take subscriptions and inference into different harnesses, while Anthropic seems focused on lock-in and control. His final note is pure disbelief: if your policy comparison makes Apple look transparent and user-friendly, you’ve really messed up.