Jev Explained: The AI Model That Makes Decisions Instead of Writing Answers
Learn how TypeSafe AI’s Jev makes fast, structured decisions using calibrated probabilities, and why it could become a new judgment layer for software.
Latest
Conjectures Miners Solve Two More Erdős Problems With Lean-Verified Proofs
Miners settled Problems 108 and 653 days after announcing a separate Erdős Problem 96 result.
Nous Research Says Hermes Agent Used 1,393 Subagents to Cut Its Own Codebase by 34%
The open-source agent’s self-refactor reduced non-test Python lines by 34.4%, though Nous says review caught real regressions before merge.
Odyssey Unveils Odyssey-3 World Model for Robots, Vehicles, Drones, and Games
Odyssey is betting one world model can become a shared control layer for robots, humanoids, cars, drones, and video-game agents.
Nous Research Launches Plugin Catalog for Hermes Agent
The catalog gives Hermes Agent users a reviewed directory of official and community plugins, with one-command installs and commit-pinned provenance.
What Is Yuma Consensus? How Bittensor Turns Scores Into Rewards
Learn what Yuma Consensus is, how Bittensor turns validator scores into miner rewards, and why stake-weighted clipping matters for TAO emissions.
TypeSafe AI Launches Jev, a System One Model Built for Software Decisions
The launch pairs TypeSafe’s first machine-native model with a $40 million seed round led by DCVC.
Figure Index Crosses 86,000 Weekly Active Users Uploading Robotics Data
The company says Index is now drawing 86,000 weekly contributors as it expands a real-world training pipeline for general-purpose robots.
Reward AI Introduces OM-1 Robot Policy for Human-Learned Manipulation
Reward AI’s Omnibody stack uses a wearable hand interface to collect multimodal human demonstration data and run one robot policy across arms, mobile robots, and humanoids.
NIOME Plans Capped GENOME-to-SN 55 Alpha Conviction Swap Ahead of Q4 Migration
NIOME is piloting a capped GENOME-to-NIOME “conviction swap” to move Base holders into Bittensor Subnet 55 without shocking the alpha pool. The first tranche is limited to 5% of GENOME’s snapshot market cap, and delivers NIOME as locked alpha with decaying conviction.
Agility Robotics Unveils Digit 5 Humanoid for Safer Factory Work
The fifth-generation Digit adds a 50-pound payload, faster charging, and a safety architecture designed for work closer to people.
404—GEN Launches First Game Jam for Games Built With AI-Generated Assets
404-GEN has announced a two-week contest that asks creators to build playable Three.js games with 404’s game generator repo, with 10 TAO in rewards.
Grok Bot vs Hermes Agent (2026): Which AI Bot Builder Is Better?
Grok Bot vs Hermes Agent compared on pricing, models, automation, memory, hosting, security and ease of use. Find the best AI agent platform for your workflow.
Nous Research Launches Hermes Business for Team-Based AI Agent Workflows
The new Nous Portal tier gives organizations shared balances, member-level spending controls, and team-wide Hermes skills, while Enterprise adds private deployment options.
Conjectures Miners Disprove Erdős Problem 96 With Lean-Verified Construction
The Bittensor subnet says its miners found a Lean-verified counterexample to a long-standing conjecture about unit distances in convex polygons.
Refinery Turns Bittensor Subnet 125 Into an Optimizer Research Market
The subnet rewards miners for Python optimizer code that improves a standardized 360M-parameter training run, with confirmation required before emissions are paid.