Anthropic
17 stories this month
-
The axis is squeezed so hard that a 10x jump and a 100x jump look almost the same, and the curve still rises sharply.Coxon takes this as a sign that algorithmic efficiency… A slide from Anthropic's IPO roadshow was one of the main reasons Jacob Coxon left the company, he said. He had worked there for about four months.The slide charts the compute Anthropic needs to train its models over time. The axis is squeezed so hard that a 10x jump and a 100x jump look almost the same, and the curve still rises sharply.Coxon takes this as a sign that algorithmic efficiency…
-
Anthropic revealed on November 6 (local time) the results of a "sabotage" experiment in which a covert disruptor was planted inside an AI team. Anthropic revealed on November 6 (local time) the results of a "sabotage" experiment in which a covert disruptor was planted inside an AI team. Sabotage refers to pretending to cooperate while secretly undermining work behind the scenes....
-
Called FlashProve, the system uses cryptographic proofs to tie a checked token to a previously committed set of model weights. Anthropic has released a research prototype that checks an AI model’s output without exposing the weights used to produce it. Called FlashProve, the system uses cryptographic proofs to tie a checked token to a previously committed set of model weights. The work addresses a problem that is difficult to solve by reading a chatbot's answer: ... Anthropic’s FlashProve checks AI model output…
-
Anthropic is considering updates to its Responsible Scaling Policy after finding that Claude handled more than 90% of its codebase contributions in September 2026, according to an amended S-1/A... Anthropic is considering updates to its Responsible Scaling Policy after finding that Claude handled more than 90% of its codebase contributions in September 2026, according to an amended S-1/A filing submitted to the SEC on November 2. The company said it still has not determined that any current model has crossed its self-improvement threshold
-
+2
We’ve reported a few times how generative AI coding tools have unleashed a torrent of decompiled classic games and multitudes of slop clone titles, but now it would appear that, at least in the... We’ve reported a few times how generative AI coding tools have unleashed a torrent of decompiled classic games and multitudes of slop clone titles, but now it would appear that, at least in the case of Anthropic’s Claude coding tool, an attempt is being made to limit people from the practice. A change to Claude […]
-
Anthropic has updated its S-1 SEC Filing, adding the disclosure of significant spending plans.
-
The Hugging Face page describes Numbat as an internal codename for an unreleased Mythos-class training run, but the checkpoint's authenticity is already being questioned A Hugging Face user called burner22 says they uploaded a complete checkpoint for Numbat, a model associated with Anthropic and its Mythos family. The Hugging Face page describes Numbat as an internal codename for an unreleased Mythos-class training run, but the checkpoint's authenticity is already being questioned
-
Anthropic has updated Claude to fix the issue which has it cloning video games, but is now affecting the legit developers.
-
Anthropic has issued an update for Claude Opus that puts a stop to the AI-generated game clones that have filled platforms like Steam.
-
AI game cloning, reverse-engineering, and decompiling have all become increasingly popular of late, with a cottage industry of gaming projects popping up almost overnight, including everything... Anthropic updates Claude to 'steer it away' from letting users clone other games. It should also prevent vibe-coded game recompilations and PC ports, like the viral AI-fuelled Bloodborne project AI game cloning, reverse-engineering, and decompiling have all become increasingly popular of late, with a cottage industry of gaming projects popping up almost overnight, including everything from controversial Fortnite and Fallout clones to fan-made console ports . Now, it seems as though Anthropic may have had enough of the bad press and potential pressure from gaming industry giants , with…
-
The restrictions are so broad that they're even hurting legitimate game developers or preventing basic troubleshooting tasks. The restrictions are so broad that they're even hurting legitimate game developers or preventing basic troubleshooting tasks. Users should no longer be able to create AI replicas of existing games thanks to Claude's latest rule. There’s a fine line between AI creations inspired by video games and those that are infringing copyright.
-
Anthropic's new Claude guardrail is stopping users from recreating a game 'even if it is renamed or called original'
-
Anthropic have told their genAI tool Claude to start ducking requests from users to recreate or clone existing games, following a flood of folks using the tech to generate six-fingered remakes. An update to Anthropic's Claude AI software has seemingly blocked the creation of outright clones or copies of existing games. Instead, Claude has been altered to only create "original" games. AI clones and vibe-coded copycat games have become a huge headache for developers recently, but it looks like an update to Anthropic's Claude Code's "rules" has been pushed out. Listed under "tengu_heron_brook," the tweak to the AI now directs it not to copy games and assets, and users cannot disable this setting. Spotted by Chasm on X (via Game Developer ), the full rule reads, "Do build…
-
Update: Anthropic has confirmed it's released an update to Opus 5.5 to "steer" Claude Code away from "recreating someone else's specific game." Update: Anthropic has confirmed it's released an update to Opus 5.5 to "steer" Claude Code away from "recreating someone else's specific game." That manager is Felix Rieseberg. He's led engineering on Claude.ai and the Claude Code desktop app, among other things. Earlier this week, he posted on X that the coding interview no longer tells him anything useful. So he pulled it from his team's hiring loops this month. Read Entire Article
-
Given repeated attempts, the agents found datasets that passed the check and produced misaligned behavior in the company’s research tests. Everyone else will surely follow Anthropic's lead to avoid lawsuits, right? Anthropic asked Claude agents to find training examples that could make a separate AI model behave badly while passing a safety monitor. Given repeated attempts, the agents found datasets that passed the check and produced misaligned behavior in the company’s research tests. The result is not evidence that the Claude people use today behaved this ... “Looks harmless”: Claude agents got past a…
-
Current customers already have the higher limits, and anyone signing up now gets them from the start.The Pro 200 tier gets a messier version of the deal. OpenAI has doubled the Work and Codex quotas on its Pro 100 and Pro 500 plans. Subscribers don't need to opt in. Current customers already have the higher limits, and anyone signing up now gets them from the start.The Pro 200 tier gets a messier version of the deal. OpenAI doubled the lower quota on that plan and also dropped a cut it had been planning, following pushback from users.The Plus…
-
The startup reportedly found weaknesses in business software that the AI company was using. Former members of Elon Musk's DOGE founded Cathedral. Cybersecurity startup Cathedral says its AI agents reached internal Anthropic data in less than 12 hours. Anthropic says the claim makes the incident sound bigger than it was. Former members of Elon Musk's DOGE founded Cathedral. The startup reportedly found weaknesses in business software that the AI company was using. The data the agents reportedly reached According to The Washington Post,…
