Anthropic launches Claude Fable 5.1 and says it’s up to 45 percent cheaper for agentic work
Anthropic says Fable 5.1 cuts agentic task costs up to 45% while delivering stronger performance than Fable 5.
Daily edition
Google adds agentic video to Gemini 3.7 Flash as Anthropic debuts cheaper Claude Fable 5.1
Daily edition
80 of 81 feeds responded
372 distinct stories
34 previously issued, held back
20 worth your time
Collection and selection
372 candidate events; 57 shortlisted stories; 44 with article text available; 20 selected. Counts describe different stages; clustering, validation and refills can change the shortlist.
Active filters: none. Sort: Ranked.
All 20 stories, in ranked orderMore filters
Path to Astra: critical capabilities and frontier safeguards
OpenAI reports Astra meets Critical cybersecurity capability threshold, prompting stronger safeguards before release.
Introducing agentic video understanding with Gemini
Google says Gemini 3.7 Flash’s agentic video cuts token usage and costs while improving accuracy for tasks like anomaly detection.
Also covered byGoogle (Gemini)Atlas: A World Model for Spatial Intelligence
World Labs says Atlas is an omni‑modal diffusion transformer that natively handles text, images, video and 3D, scaling performance with more compute.
Claude Fable 5.1 made me a really nice animated pelican
Willison notes Fable 5.1 scores 52.6% on Terminal‑Bench‑Science 0.1, far above previous models, highlighting coding and knowledge gains.
BenchMIRT: What are LLM benchmarks actually measuring?
Hugging Face says BenchMIRT audits individual prompts, revealing that benchmarks often mix multiple abilities like safety and reasoning.
Also covered byAi2OpenAI delayed its new model’s development after the Hugging Face hack
OpenAI says a hack on an unreleased model forced it to postpone Astra’s development to reinforce safety measures.
Frontier models can recover up to 65% of facts they can't directly recall — just by thinking longer
Google Research says frontier models encode 95‑98% of facts, with recall—not encoding—being the main hallucination bottleneck.
How Agentic AI Turns IC Engineers into Architects
SemiWiki reports agentic AI lets IC engineers generate RTL, run tests and debug autonomously, shifting them to architectural roles.
Anthropic opens Claude AI text detection to regulators, media, fact-checkers, and others
Anthropic says its Claude watermark API lets regulators and media verify AI‑generated text, using a pattern that may survive editing.
Healthcare organizations can now connect EHR and additional industry data to ChatGPT
OpenAI says ChatGPT now integrates Epic EHR and public health data, letting clinicians retrieve patient context securely.
Developing Enterprise Frontier Safeguards with our customers
Anthropic says it is co‑creating frontier safeguards with enterprise customers to manage high‑risk AI deployments.
PII-TRACE: Detecting Personal Data Before It Leaves a Device
Perplexity says PII‑TRACE detects personal data on devices before it leaves, helping prevent accidental leaks.
Manus resumes solo operations after collapse of US$2 billion Meta deal
SCMP reports Manus resumes independent operations after Beijing blocked its $2 billion acquisition by Meta.
Also covered byDigiTimesLaunch HN: Nori Robotics (YC S26) – A low-cost humanoid robot for development
Community post describes Nori Robotics as a low‑cost humanoid platform for home tasks and skill sharing.
Google needs Hollywood more than the studios need AI
The Verge reports Google is courting studios for AI training licenses, offering cash in exchange for copyrighted content.
Dell Raises Sales Forecast on AI Server Demand
The Information says Dell raised its AI server sales forecast after a 58% revenue jump, signaling strong demand.
How AI-native companies turn workflows into operating capability
OpenAI reports AI‑native firms like Basis and Clay use agents to automate onboarding and account management, boosting token output per user.
PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors
Latent Space notes top AI open‑source projects now block external PRs, using internal agent‑driven software factories instead.
Mapping global methane emissions from space with deep learning
Google Research says its EMIT model turns satellite data into global methane plume detection, aiding rapid climate mitigation.
No stories match. Active filters.