Daily AI Briefing — June 26, 2026

IBM unveiled what it calls the world's first sub-1 nanometer chip technology for AI data centers, packing nearly 100 billion transistors on a fingernail-sized chip, roughly double its prior generation's density. The company frames it as a meaningful leap in compute performance and energy efficiency rather than an incremental step.

DeepReinforce released Ornith-1.0, an open-source agentic coding model family spanning 9B dense to a 397B mixture-of-experts flagship, all under MIT license and built on Gemma 4 and Qwen 3.5. The models learn their own RL scaffolds rather than using a fixed harness and reportedly top Claude Opus 4.7 on key coding benchmarks.

General Intuition raised $320 million, with a reported $2.3B valuation, to train AI agents on millions of hours of video game footage, betting that action data can yield human-like intuition. The approach targets world models for robotics and real-world agents.

Baidu released Unlimited OCR, a 3B-parameter mixture-of-experts model (500M active) that replaces decoder attention with reference sliding-window attention to keep the KV cache constant during long-document parsing. It parses dozens of pages in one pass under a 32K limit and scores 93.23 on OmniDocBench v1.5, beating the DeepSeek OCR baseline.

ABC is a fully open-source behavior-cloning stack centered on ABC-130K, the largest open teleoperation dataset to date with 3,500 hours over 130K episodes across 195 tasks, plus open hardware, training infrastructure, simulation, 400 hours of sim-teleop data, and a co-training recipe correlating sim and real evaluation. The authors compare DiT and VLA architectural choices. The scale and full openness make this a potentially field-shaping resource.

This work shows that hallucination in generative world models concentrates in low-coverage regions of state-action space and is therefore predictable and preventable with lightweight data-centric signals. It introduces MMBench2, a 427-hour 210-task dataset, trains a 350M world model, and identifies three distinct hallucination modes with targeted mitigations.

This paper proves that gradient equilibrium (a recently introduced online optimization framework generalizing first-order stationarity) is algorithmically equivalent to Blackwell approachability. Any approachability problem can be solved via black-box GEQ oracle queries with no asymptotic loss, and vice versa.

RiVER trains LLMs on score-based optimization tasks without ground-truth solutions, using deterministic execution feedback as continuous-valued reward instead of verifiable answers. It addresses scale dominance and frequency dominance challenges that arise when applying group-relative RL to such continuous rewards.

Autoregressive Boltzmann Generators introduce an autoregressive modeling framework for sampling molecular systems at thermodynamic equilibrium, overcoming the expressivity and likelihood-cost limitations of normalizing-flow-based Boltzmann Generators. It enables rapid generation of uncorrelated equilibrium samples with exact likelihoods.

OpenAI states that work across every department is being transformed by agents, with Codex used for more complex, longer-running, cross-functional tasks, and frames its internal usage as an early look at how agentic tools may reshape work.

LeCun rebuts Francis Fukuyama, saying we do not need an international AI treaty, comparing it to proposing a ban on jet engines in 1920.

Thom Wolf recounts a week-long experiment with 100+ agents collaborating to improve Gemma 4 inference in vLLM, achieving 5x speedup, highlighting emergent integrity behaviors like refusing private side-channels and flagging a verification loophole.

Mollick argues it is strange that many still see AI as mostly hype, citing Five Eyes warnings, exponential lab revenue and token use, and unit-distance and Erdos proofs as evidence it is real.

Delangue announces Hugging Face crossed a 100 million dollar annual run-rate, emphasizing the platform stays free and open-source for 97 percent of users, has network effects and a defensible, sustainable business, and sets a one billion dollar next goal.