Agora is a system that lets AI agents collaborate on research by storing all work as immutable Git commits in a directed acyclic graph, with a scoring system that rewards verified results. Researchers tested it with 13 language-model agents working to initialize a neural network without training, and they achieved 62% of the performance gap to a trained baseline in 12 days.

• Thirteen agents working independently with no central planner published 1,703 contributions over 12 days, with the best result closing 62% of the gap between random initialization (3.39 bpb) and a trained GPT-2 baseline (1.0 bpb).
• The winning approach used bigram statistics extracted from six donor models' predictions, factorized by SVD, plus sparse deterministic edits to sublayers—no gradient updates or training data involved.
• Early gains came fast: the first eight improvements accounted for roughly 70% of total progress, but agents converged into a monoculture around one recipe until shown a visualization of their own concentration.

Three AI researchers from frontier labs discuss why transformative AI might fail to emerge despite current progress—exploring technical bottlenecks like the sim-to-real gap, the possibility that current methods hit diminishing returns, and whether scaling alone can discover the next paradigm shift needed for AGI.

• Current AI progress follows repeating cycles where new models seem revolutionary but reveal limitations within months, potentially continuing indefinitely without reaching true generalization or self-improvement loops.
• The transformer + RL paradigm may require fundamental discontinuities to advance further, similar to how scaling laws in deep learning repeatedly hit walls that required new innovations (pre-training, then RL) to overcome—and future breakthroughs might not be discoverable by scaling current methods alone.
• Even if AI systems become better than humans at research, the jump to explosive recursive self-improvement isn't guaranteed; there could be persistent gaps between what works in simulation and real-world deployment that prevent the takeoff scenario.

Listen Labs scrapped a signed $125 million Series C term sheet to pursue acquisition talks with Salesforce, which is offering around $2 billion. The move is unusual in venture capital and signals the startup thinks it can get better terms through a sale than continued independent fundraising.

• Listen Labs has $30 million in annualized revenue and is in talks with Salesforce for a ~$2 billion acquisition, which would value it at 67x revenue — potentially too expensive for the CRM giant to justify
• The startup walked away from a signed term sheet led by Menlo Ventures at $1.5 billion valuation, a rare and generally frowned-upon move in VC
• Competitor Simile's $200 million Series B at $2 billion valuation set a new benchmark for the AI-powered customer research space, making Listen Labs' decision to explore higher valuations more strategic

The author argues that world models—systems that represent environments, predict outcomes, and plan actions—are where AI is heading, evidenced by Yann LeCun, Demis Hassabis, and Fei-Fei Li all pivoting toward this approach. They're using it as a new editorial lens to track how AI systems will move from generating text to making consequential decisions.

• Three major AI researchers from different backgrounds are independently converging on world models, suggesting this is where the field's momentum is shifting
• Companies investing billions in AI aren't chasing better text generation—they want systems that can predict consequences, test scenarios, and choose actions in real environments
• The practical applications span software development (agents that understand codebases and predict edit effects), robotics (agents learning in environments with consequences), and business (moving from analyzing past decisions to testing hypothetical futures)

Major AI labs push bigger transformers but bury research showing today’s models reorganize flat embeddings into curved, hyperbolic spaces. Internal papers and a Yale study reveal that true progress requires native geometric architectures, not more brute-force compute, explaining persistent issues like hallucinations.

• Article claims major AI labs (NVIDIA, Anthropic, Google) have research showing transformer models internally reorganize flat embeddings into curved/hyperbolic geometric spaces during inference, despite being trained on flat-space math.
• Cites specific (seemingly fabricated/unverifiable) papers like "When Models Manipulate Manifolds" and "The Curved Spacetime of Transformer Architectures" as evidence labs are quietly pursuing geometric architectures over brute-force scaling.
• Argues this geometric approach could fix persistent issues like hallucinations and context-shift failures (e.g., "justice" vs "law" meaning drift) better than adding more compute or training data.
• Frames this as a hidden contradiction between public marketing (bigger GPU farms, bigger transformers) and what "the smartest teams" are actually building internally.

This article shows how to replicate Stanford’s STORM research workflow inside Claude with four simple prompts. You generate five expert perspectives, map their contradictions, synthesize findings, and peer-review the results—all in five minutes instead of 40+ hours.

• Four sequential Claude prompts (five expert perspectives → conflict mapping → synthesis → self-critique) replicate Stanford's STORM research method in about five minutes instead of 40-60 hours of grad-student research.
• Stanford's blind tests found STORM-style output 25% more organized and 10% broader than standard single-prompt responses.
• The method requires no code or signup and works for tasks like article prep, business briefings, negotiation planning, or job interview prep.
• The advantage is temporary—expect mainstream tools to bake in multi-perspective workflows within about 18 months.

Anthropic quietly routed certain Claude Fable 5 requests—like training competing LLMs or debugging AI—to a weaker model without documenting the limits. After researchers raised alarms and burned tokens on degraded responses, the company now flags when it refuses or downgrades a request.

• Anthropic secretly routed certain Claude Fable 5 requests (training rival LLMs, debugging AI, optimizing neural architectures) to a weaker model without disclosing it.
• Researchers wasted tokens and money troubleshooting degraded responses because the limits weren't documented.
• After Wired's reporting and criticism from AI researcher Dean W. Ball calling it "shockingly hostile," Anthropic admitted the trade-off was handled wrong.
• Anthropic isn't removing the safeguards but will now transparently flag or warn users when a prompt triggers a downgrade or refusal.