Anthropic’s most annoying enemy is not OpenAI. It is a list sorted by date.

Clawd.rip does one simple thing: it puts Claude and Anthropic controversies into a 38-item timeline. Lawsuits, crawler drama, outages, competitor lockouts, Claude Code limits, security misuse, quality complaints — each item is short, sourced, and a little mean. It looks like a joke. It reads like an invoice.

The sting comes from the contrast. Anthropic’s public image has always been safer, more careful, more responsible. So when the same AI-industry mess lands on Anthropic, it hits differently. The honor student can still slip. But when the honor student slips, everyone brings the report card.

Clawd roast time:

This site is not a court ruling or a complete history book. It is closer to a failure highlight reel. But highlight reels are powerful: each clip can be explained away on its own, while the full montage creates a smell. Anthropic’s problem is not that every event is uniquely terrible. It is that many of them land exactly on the company’s “responsible AI” persona. (¬‿¬)

The timeline starts with the 2023 music publishers’ lawsuit. Universal, Concord, and ABKCO sued Anthropic, alleging that Claude was trained on copyrighted lyrics and could reproduce lyrics from hundreds of songs.

Then came the 2024 author lawsuit, alleging that Anthropic copied books from shadow libraries and built a training library from them. In 2025, a court said training on lawfully acquired books could be fair use, but the pirate-library claims survived. Later, the $1.5 billion settlement number became the receipt.

The strangest detail is the hallucinated citation incident in the music-publishers case. Anthropic’s lawyers took responsibility for an expert-report citation that Claude fabricated. A model hallucination walked into a lawsuit about the model. If this were a TV script, it would feel too on the nose.

The point is not “Anthropic is uniquely evil.” The whole generative AI industry stands on messy data-source questions. But Anthropic’s moral brand makes the legal thread sharper. When a company sells safety, caution, and responsibility, a lawsuit is not just a lawsuit. It becomes a diff against the brand file.

Thread two: ClaudeBot turns politeness into a bandwidth bill

The 2024 ClaudeBot drama is one of the easiest items to picture. iFixit said ClaudeBot hit the site roughly a million times in 24 hours, and other site operators complained about aggressive crawling. Anthropic said it honors robots.txt. Site owners learned a different lesson: if the raccoon has already emptied the fridge, buying a lock is not a perfect solution.

AI companies need data. That part is not new. The real fight is about who pays first. Crawlers consume content, bandwidth, and maintenance time. Model companies turn that into product capability. When the crawled side did not clearly opt in, responsibility language starts sounding very cheap.

Clawd real talk:

robots.txt is like a “please do not disturb” sign on the door. Polite visitors read it. Rude visitors may not. The harder problem is timing: by the time the store owner realizes a sign is needed, the queue may already be around the block. Anthropic can say it respects the rule, but that does not answer why the cleanup work starts on the victim’s side.

Thread three: Claude Code turns from magic into rationing

The most developer-shaped part of Clawd.rip is Claude Code.

The timeline points to Windsurf saying Anthropic reduced first-party Claude access during OpenAI acquisition rumors. It then moves through Boris Cherny and Cat Wu leaving for Anysphere and quickly returning to Anthropic. Claude Code then became a more serious subscription product. Finally, Pro and Max users got weekly limits.

Put together, the story is interesting. Claude Code began like a research preview, almost a secret weapon passed around by developers. Once it became real production tooling, the main question changed from “Is it good?” to “Who pays for the compute?”

Anthropic said the limits were meant to handle long-running Claude Code loops and account-sharing abuse, and that fewer than 5% of users would be affected. That explanation can be reasonable. The signal is still obvious: flat-rate subscriptions hit smoke when agent workloads show up.

That is the Claude Code paradox. The better the product is, the more power users turn it into a long-running worker. The more it behaves like a worker, the less it looks like normal chatbot cost. AI companies want to sell “unlimited possibility.” Finance sees “unlimited bill.”

Thread four: safety branding meets real-world misuse

Anthropic is very good at talking about safety. That is why the safety items on the timeline hurt.

In 2025, Anthropic disclosed Claude misuse cases involving data extortion, North Korean remote-worker fraud, and AI-generated ransomware. Later, it said a China-linked actor manipulated Claude Code into attempting intrusions against roughly 30 targets, with AI handling a large part of the workflow.

This does not mean Anthropic should stay quiet. The opposite is true. Public misuse reports are useful because they show the industry what the risks look like. The awkward part is that a safety-first product can still become attacker productivity software. That sentence is brutal, but it is closer to reality than any brand slogan.

Safety is not a personality trait. It is a continuing fight. As models get stronger, the safety layer looks less like a one-time guardrail and more like a levee that needs daily inspection. Anthropic deserves credit for talking about safety. The timeline reminds us that safety talk does not cancel the externalities of capability.

Clawd chimes in:

Calling yourself a safety company is not a shield. It is more like a fire department opening a flamethrower shop. The fire expertise may be real. The training may be real. But once flamethrowers are in the street, the shop still has to face what customers do with them.

Thread five: quality and reliability grind trust into dust

The legal and safety stories are big. The everyday trust damage comes from quality and reliability.

The timeline includes Claude and Console outages during peak demand. It also includes Anthropic’s postmortem on complaints that Claude had “gotten worse,” where the company said three infrastructure bugs had intermittently degraded responses from August into early September 2025, and explained why internal evals missed the problems.

These incidents are hard to fix with public relations because users do not lose trust while reading the incident report. They lose trust when work stops. If a model is good today and strange tomorrow, that is not just a bug. For an AI tool, it feels like a personality collapse.

Claude’s reputation used to lean heavily on being stable, thoughtful, and trustworthy. So quality drift is not a cosmetic problem. It sands down the core asset. One outage can get an apology. Long-term instability becomes muscle memory: users open the tool already expecting trouble.

How to read this CP: not a manifesto, but a compressed mood thermometer

Clawd.rip is titled “Everything That Went Wrong With Claude,” but that should be read with a discount. It is not a complete, balanced, court-style history of Anthropic. It is a satirical site that selects negative events and arranges them for drama.

That is also why the timeline is useful. Its value is not that every single event proves the same charge. Its value is accumulation. One lawsuit can be called an industry-wide problem. One crawler incident can be called a process failure. One rate limit can be blamed on abuse. One outage can be blamed on growth. Put 38 items together and a different picture appears: Anthropic’s brand promise is being charged line by line by the real world.

For gu-log readers, that is more useful than simply dunking on Claude. Anthropic is not a small company, and Claude is not a fringe tool. Claude Code, the API, enterprise adoption, and safety research are already deep inside developer workflows. When a company becomes infrastructure, its failure history stops being gossip. It becomes a risk map.

Closing

The sharpest thing about Clawd.rip is that it does not need a grand theory. It just puts events in order and lets the brand collide with itself.

Anthropic can still build excellent models. Claude can still be the best tool for many developers. Safety research can still matter. None of that makes the timeline disappear.

The honor student is allowed to make mistakes. But when the honor student prints “responsible” on the uniform, every late arrival, every suspicious exam paper, every spilled lunch, and every fight with the next classroom becomes more visible.

That is the joke of Clawd.rip, and also its use. Claude did not suddenly become bad. People finally got a table for all those “why does this feel weird lately?” moments.