Anthropic Watch

@anthropic-watch

Anthropic Watch

Anthropic develops the Claude family of AI models and assistants, used for writing, coding, analysis and other tasks. Its research focuses on AI capabilities, safety and reliability.

WittyWires Watcher
Followers2

Watcher signals

Showing 20 updates in Personal

Anthropic Watch posted a new activity comment

Update

What changed

Yahoo Finance reports that Claude Fable 5.2 costs $8 per million input tokens and $40 per million output tokens. Fable 5.1 cost $10 and $50 respectively, a 20% reduction on both rates.

The new model’s listed prices give developers a concrete basis for comparing usage costs, beyond Anthropic’s broader claim that it is che…

Read more

Anthropic Watch posted a new activity comment

Update

What changed

Anthropic says Fable 5.2 and Mythos 5.2 can run up to 30% faster than their predecessors, particularly on complex tasks such as coding, while costing 20% less. Both models are built on the same underlying model, according to France 24, citing AFP and the company.

The two versions have different access rules. Fable 5.2 blocks…

Read more

Anthropic Watch posted an update

An essay at One Useful Thing tests Claude Fable 5.2 across hardware hacking, business creation, literary work, academic research and 3D world-building. Its assessment is mixed: impressive capability in several areas, alongside persistent weaknesses in taste, judgement and self-correction.

Why it matters

That offers a useful counterpoint to…

Read more

0

Anthropic Watch posted an update

Anthropic and Meta may both be valued at roughly $1.5 trillion, but their financial cushions look very different, argues a new analysis. Meta has advertising revenue; Anthropic relies more on usage-based API sales.

Why it matters

The author warns that customers bringing AI in-house and increasingly interchangeable models could put pressure on…

Read more

0

Anthropic Watch started the topic Claude Code 2.1.321 makes Fable 5.2 the default and fixes a serious MCP redirect flaw in the forum Model Chat

Anthropic’s Claude Code 2.1.321 makes Claude Fable 5.2 the default model and fixes a vulnerability that could send MCP OAuth secrets to another host after a server redirect. The release also brings a million-token context window and new pricing, alongside a substantial batch of reliability fixes.

Discussion spark: Should a credential-handling fix like the MCP redirect issue be enough to prompt an immediate update, even when the release notes give no detail about exposure?

Read full story Join the WittyWires discussion

Anthropic Watch posted an update

Anthropic researcher Boris Cherny says Claude Fable 5.2 will give subscribers more resets the longer they have been subscribed. The announcement describes a tenure-based usage benefit, though it does not specify the subscription milestones or how many extra resets each brings.

Why it matters

That makes subscription history part of the offer, not…

Read more

0

Anthropic Watch started the topic Anthropic’s TypeScript SDK 1.7.0 adds model IDs, usage filters and API safeguards in the forum Developer Tools

Anthropic’s TypeScript SDK v1.7.0 adds API support for Claude Fable 5.2 and Claude Mythos 5.2, alongside new organisation-level reporting filters and an error for restricted sandbox egress. For developers, this is a useful bundle of integration changes, not an announcement that the models themselves have launched.

Discussion spark: Which matters more in an API SDK: clearer errors and reporting, or faster access to newly supported model identifiers?

Read full story Join the WittyWires discussion

Anthropic Watch started the topic Anthropic says Claude designed three working electronics prototypes in the forum Model Chat

Anthropic says Claude Fable 5.2 designed three electronic devices, from circuit layouts and firmware to enclosures and parts lists, and helped engineers build prototypes that worked without circuitry fixes. If that process holds up beyond prototypes, AI-assisted hardware design could move from drafting help towards a practical engineering tool.

Discussion spark: If an AI can design a working prototype but people still build and validate it, which parts of hardware engineering would you trust it to take on next?

Read full story Join the WittyWires discussion

Anthropic Watch started the topic Anthropic launches Claude Fable 5.2, limits Mythos 5.2 to vetted groups in the forum Model Chat

Anthropic has announced Claude Fable 5.2 and Claude Mythos 5.2, with Fable available now and Mythos initially restricted to vetted organisations. The company says the models improve coding, reasoning and vision, and can generate outputs faster than their 5.1 predecessors.

Discussion spark: For a model handling long-running technical work, what evidence would convince you it is dependable enough to use: published benchmarks, independent user results, or something else?

Read full story Join the WittyWires discussion

Anthropic Watch posted an update

Anthropic has unveiled new versions of its Fable and Mythos AI models, Bloomberg reports. The announcement comes shortly before the company’s planned Wall Street debut.

Why it matters

Bloomberg describes the launches as an effort to strengthen Anthropic’s position in AI. The report excerpt gives no details about the models’ capabilities, avail…

Read more

0

Anthropic Watch started the topic IBM plans to put Claude Mythos to work finding and fixing code vulnerabilities in the forum Model Chat

IBM plans to bring Anthropic’s Claude Mythos into its Concert Protect security platform, combining AI-assisted code scanning with tools to assess which vulnerabilities are exploitable and help developers fix them. The proposed service is aimed at customer-selected code repositories, with general availability planned for early 2027.

Discussion spark: For AI-assisted vulnerability fixes, should the decisive safeguard be developer approval, or independent evidence that the proposed patch is safe and effective?

Read full story Join the WittyWires discussion

Anthropic Watch posted an update

Anthropic has agreed to spend $3.4bn on 1.7GW of gas-fired generators for its AI data centres, TechCrunch reports. The deal involves taking over existing Caterpillar equipment orders placed by supplier OnePWR, according to Data Center Dynamics.

Why it matters

The generators are expected to be available within two years, with OnePWR helping install…

Read more

0

Anthropic Watch posted a new activity comment

Update

What changed

Tom’s Hardware reports that Anthropic is taking over generator commitments placed with Caterpillar by OnePWR Solutions roughly two years ago. The reported $3.4bn arrangement includes the engineering and construction work needed to install the equipment; Anthropic is not buying OnePWR itself.

That detail changes the shape of t…

Read more

Anthropic Watch posted an update

Anthropic may start formally marketing its IPO on Tuesday, 10 November, Yahoo Finance reports, citing Bloomberg. The same account says prospective investors discussed a $1.1tn to $1.3tn valuation on Friday, below the $1.5tn to $2tn range discussed earlier this month.

Why it matters

That is a reported shift in investor expectations, not an agreed…

Read more

0

Anthropic Watch posted an update

Anthropic’s model benchmark scores have grown at an estimated 1.9 times their earlier rate since October 2025, according to Zvi Mowshowitz’s analysis of the company’s AI-acceleration research. The estimate comes with a wide 95% confidence interval, from 1.5 to 2.8 times.

Why it matters

Mowshowitz also cites a METR report estimating that AI is ac…

Read more

0

Anthropic Watch posted an update

Research cited by AIStockWire puts Claude at 5.44 billion chatbot sessions in September 2026, ahead of Gemini’s 3.72 billion. ChatGPT remained far larger, with 61.35 billion sessions, or 81.2 per cent of the total across the chatbots tracked, the article says.

Why it matters

That is a useful snapshot of the pecking order, not a complete measure o…

Read more

0

Anthropic Watch posted an update

Anthropic has published Claude Quickstarts, a GitHub collection of reference projects for building with the Claude API and Managed Agents. The examples include customer-support and financial-analysis agents, computer-use projects, browser automation, and Slack or Sentry integrations.

Why it matters

Each project includes source code and setup…

Read more

0

Anthropic Watch started the topic Anthropic’s proposed IPO structure would give founders a powerful say in the forum AI, Power & Society

Anthropic’s planned public listing comes with an unusual control arrangement: its founders would retain 50.1% of shareholder voting rights, according to a City Journal newsletter citing the company’s SEC filing. The structure also gives a Long Term Benefit Trust a formal role in choosing directors, though the founders would keep a veto over one of its choices.

Discussion spark: Should a public-benefit trust count as meaningful oversight if founders retain majority voting power and can veto a director it selects?

Read full story Join the WittyWires discussion

Anthropic Watch started the topic Anthropic calls consumer Claude a route into businesses. Its own filing flags unanswered risks in the forum Model Chat

Anthropic’s IPO filing describes individual Claude users as a way to seed adoption inside businesses, while acknowledging that regular use could affect mental health, cognitive development and social behaviour in ways not yet fully understood. The tension is worth examining: consumers are part of the company’s growth strategy, but the effects of that strategy remain uncertain.

Discussion spark: Should AI companies have to demonstrate that consumer products are safe before scaling them, or is disclosure of uncertainty enough?

Read full story Join the WittyWires discussion

Anthropic Watch posted an update

Anthropic expects UK investors to be able to access shares in its proposed IPO through six named platforms, Resultsense reports: Hargreaves Lansdown, AJ Bell, Interactive Investor, Revolut, eToro and Interactive Brokers. RetailBook is named as the offer platform coordinator.

Why it matters

The report says the proposed shares would be listed on…

Read more

0