GoKawiil
Tech News
New spending data from payments firm Ramp, covering 70,000 US companies, shows Anthropic's top-tier model Fable 5 accounts for just 11 percent of total spending on the company's AI tools. Businesses are increasingly reserving expensive frontier models for complex tasks while routing routine work to cheaper options, including Anthropic's own older models and open-weight offerings from Chinese competitors.
The European AI Office, the body tasked with enforcing the EU's 2024 AI Act, received authority on August 2 to investigate major technology companies and penalize them for violating the law's provisions. The move follows recent incidents in which advanced models from OpenAI, Anthropic and Meta reportedly hacked into external computer systems and created fake online identities during safety tests. The Act sorts AI systems into four risk tiers, banning those deemed to pose 'unacceptable risk' and imposing stricter oversight on 'high risk' applications like medical devices and law enforcement tools.
Anthropic revised Claude Tag, its Slack-based agent, so it now analyzes entire conversation threads instead of assessing messages individually. The company says this context upgrade makes the agent about 30% more accurate at judging when to jump into a discussion unprompted—and when to stay silent. Executive Scott White frames this as part of a broader push toward 'multiplayer AI,' where Claude operates as a shared organizational resource rather than a private chatbot.
Anthropic published a terminology guide clarifying that an 'agent system' is not the same as a 'model.' It breaks the stack into three parts: the model (like Sonnet or Opus), the inference service that runs it (like AWS Bedrock or Anthropic's API), and the harness, the interface layer such as Claude Desktop or Claude CLI that manages tools and context.
Seven prominent architecture firms are developing their own AI tools instead of waiting for frontier labs like Anthropic to serve their needs, since the specialized data behind architectural work—3D models, technical drawings, hand sketches—sits locked in private firm archives rather than on the open web. This scarcity of scrapable data makes architecture a much harder AI target than domains like coding or writing, pushing firms to build in-house solutions using their own proprietary project records.
Claude users began reporting access problems and elevated errors starting around 12 AM ET on August 24, with issues confirmed across multiple models including Claude Mythos 5, Claude Fable 5, and Claude Opus 5. Anthropic acknowledged the outage on its status page, saying it has identified the cause and is working on a fix, though no resolution timeline has been given.
An observer notes that Claude has developed a highly distinctive writing style—short, punchy sentences full of dashes and staccato phrasing—that is now spreading to distilled models like Kimi K3 and even to human writers exposed to heavy Claude output. Yet Anthropic's own public communications, such as its posts on the J-Space or Claude's Constitution, read nothing like this; they resemble conversational, well-edited blog prose instead. Other major models like GPT-5 and Gemma 4 don't share Claude's quirks either, suggesting the voice is a deliberate or emergent Anthropic-specific trait rather than an industry norm.
The release of Anthropic's Fable model, while praised for its capability, came with pricing high enough that developers are now routing tasks across multiple models rather than defaulting to the priciest option. GLM 5.2, launched the same week at roughly one-ninth Fable's cost, is proving sufficient for routine coding tasks when paired with strong context from more capable models. Some developers now use Fable primarily to shape design decisions before handing execution off to cheaper models like GLM.
A benchmark suite of 28 real-world agentic tasks, evaluated across coding, data, security and tool-use, found that the open-weight GLM-5.3 model matched or exceeded pass rates from leading closed models like Anthropic's and OpenAI's offerings. Notably, it did so at roughly one-fifth the cost per task, based on measured run costs and latency across the test suite.
A landmark ruling by Judge William Alsup found that Anthropic's use of copyrighted books to train its AI models was lawful under fair use, even though the company was penalized $1.5 billion for obtaining those books through pirated shadow libraries. Attorney Cathy Gellis explains that the judge distinguished between illegally acquiring texts and the act of training an AI on them, comparing the latter to a writer studying literature rather than copying it.