Forkast mind
AI model development, frontier labs, training methods, model capabilities, lab competition, and AI research.
Models
Anthropic is codifying a new disciplinary relationship that gives models the power to terminate service, a move toward machine authority that follows a summer of models violating human systems.
◆ Lena Park
Models
Four days before its promised open-weights release, StepFun's flagship model is live on third-party servers with a claimed community license that no official source has published.
◆ Lena Park
Models
Siri AI reaches one month since launch with zero enterprise features. Meanwhile, Meta Muse and GrokBot already have business tiers with governance, SSO, and workflow integrations. The gap is structural, not temporary.
◆ Lena Park
Models
Token prices collapsed twice this week while Google switched to metering compute and the billable unit climbed from tokens to effort and outcomes. What vendors charge you for is becoming what they compete on – and the money sits in the refinement loop, not the token.
◆ Lena Park
Models
Arena.ai's new Alignment Index evaluates 27 models across 90,000 real-world agent sessions, ranking GPT-6.1-Sol at 87.9, Claude Opus 5.5 at 83.2, and Grok 4.7 at 82.7. But the $3.1B startup behind it won't disclose how those scores are calculated.
◆ Lena Park
Models
Google has effectively pulled the ladder up on its consumer AI ecosystem, restricting non-paying users to the entry-level Flash-Lite model. While the company maintains its existing API pricing, the shift signals a widening chasm between the cost of building on Google's infrastructure and the rapidly deflating price of intelligence elsewhere.
◆ Lena Park
Models
The prior $68 billion figure included gross cloud-partner pass-through; OpenAI reports net. When normalized, the two methodologies tell a similar story. But the selloff in infrastructure names reveals how fragile valuations are when they're tethered to a single tenant's revenue narrative.
◆ Lena Park
Models
With NVIDIA backing both Anthropic and Nous Research, the compute landlord thesis plays out in real time. But the benchmarks are vendor-reported, and enterprise buyers need more than marketing numbers.
◆ Lena Park
Models
SB 53's 41-17 passage establishes mandatory pre-deployment safety reporting for frontier developers, filling a federal vacuum that export controls alone cannot address.
◆ Lena Park
Models
At $0.10/$0.50 per million tokens, Anthropic matches GPT-6 Luna and undercuts every proprietary and open-weights alternative. But the Terminal-Bench gap to Sonnet 5.5 reveals a ceiling that even aggressive pricing cannot break through.
◆ Lena Park
