Wiredframe Radar logo

Category intelligence

AI News Briefing — June 26, 2026

37 current items analyzed and ranked.

Executive synthesis

AI News Summary

Hardware and infrastructure dominated the day. IBM claimed the world's first sub-1nm chip, packing nearly 100 billion transistors on a fingernail-sized die for AI data centers. Amazon committed a fresh $13B to AI infrastructure in India, while Europe pushed back on US chip export controls tied to the proposed MATCH Act. Databricks' former AI chief unveiled Un-0, pitching up to 1,000x lower power for generation.

Open-source releases challenged proprietary leaders:

  • DeepReinforce released Ornith-1.0, an MIT-licensed agentic coding family (9B dense to 397B MoE) that learns its own RL scaffolds
  • Baidu shipped Unlimited OCR, a 3B mixture-of-experts model keeping the KV cache flat for long-document parsing

Funding, agents, and competition rounded out coverage:

Key Themes

AI Geopolitics and Model Theft · 5Government Oversight of Model Releases · 3Open Source Model Releases · 2Agentic AI and Harnesses · 9AI Infrastructure, Hardware and Energy · 8Funding, M&A and Market Dynamics · 6AI and Society · 6

Primary evidence

Top Ranked Signals

News Ars Technica - All content Jun 25

IBM claims world’s first sub-1 nanometer chip technology

By Jeremy Hsu

70 score
AI Analysis

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.

A new chip architecture from IBM can integrate nearly 100 billion transistors on a chip the size of a human fingernail—nearly twice the transistor density of the company’s previous generation of chip technology. The resulting improvement in chip compute performance and energy efficiency comes from what IBM describes as the “world’s first sub-1 nanometer chip technology” for AI data centers. “It's not just an incremental step, it's a meaningful leap forward,” said Jay Gambetta, director of IBM Re
AI hardwareChip technologyEnergy efficiencyAI infrastructure
News MarkTechPost Jun 25

DeepReinforce Releases Ornith-1.0: An Open-Source Coding Model Family That Learns Its Own RL Scaffolds

By Asif Razzaq

68 score
AI Analysis

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.

DeepReinforce has released Ornith-1.0, an open-source model family built for agentic coding. The lineup spans four sizes, from a 9B dense model to a 397B mixture-of-experts flagship. Every checkpoint ships under the MIT license on Hugging Face. The models are post-trained on top of pretrained Gemma 4 and Qwen 3.5. Most coding agents pair a model with a fixed, human-designed harness. Ornith-1.0 instead learns to write its own. The DeepReinforce research team reports state-of-the-art results am
Open sourceAgentic codingReinforcement learningModel release
News AI News & Artificial Intelligence | TechCrunch Jun 25

General Intuition’s $2.3B bet that video games can train AI agents for the real world

By Rebecca Bellan

64 score
AI Analysis

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.

General Intuition has raised $320 million to scale AI trained on millions of hours of gameplay, betting action data can help AI develop something closer to human intuition.
World modelsAgentic AIRoboticsFundingAI training data
News MarkTechPost Jun 25

Baidu Releases Unlimited OCR, a 3B Model That Keeps the KV Cache Flat for Long-Document Parsing

By Asif Razzaq

60 score
AI Analysis

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.

Most end-to-end OCR models slow down as output grows. Each generated token adds to the KV cache. Memory rises and generation drags. Parsing dozens of pages becomes impractical. Baidu’s Unlimited OCR addresses this directly. It swaps the decoder’s attention for a design that keeps memory constant. TL;DR Unlimited OCR is a 3B-parameter Mixture-of-Experts model, with only 500M parameters active. It replaces decoder attention with Reference Sliding Window Attention (R-SWA), ke
Open sourceOCREfficient architecturesModel release
News AI News & Artificial Intelligence | TechCrunch Jun 25

Amazon ups India bet with fresh $13B AI infrastructure investment

By Jagmeet Singh

56 score
AI Analysis

Amazon announced a fresh $13B investment in AI infrastructure in India as global tech firms race to expand data center capacity in the country. The move deepens AWS's regional footprint.

Amazon’s latest India investment comes as global tech companies race to expand AI infrastructure in the country.
AI infrastructureData centersInvestmentGlobal expansion
News aibusiness Jun 25 Old anchor

Anthropic Alleges That Alibaba Pilfered Claude Capabilities

By Esther Shittu

55 score
AI Analysis

The IP-hypocrisy hot takes were already flying on Twitter yesterday, AI Business covers Anthropic's allegation that Alibaba distilled Claude's capabilities, noting the case implies enterprises should scrutinize how AI labs handle their data. It frames the dispute around due diligence and vendor data practices.

The alleged distillation suggests that enterprises should conduct due diligence into how AI labs handle their data.
Model theft and distillationAI geopoliticsEnterprise AIAI policy
News Ars Technica - All content Jun 25 Old anchor

Anthropic says Alibaba must be punished for largest Claude cloning attack

By Ashley Belanger

55 score
AI Analysis

Anthropic told US senators it measured the largest campaign yet to illicitly extract Claude's capabilities, alleging operators tied to Alibaba and its Qwen lab ran nearly 25,000 fraudulent accounts generating 28.8 million exchanges between April and June. The accusation ties into China racing to match Anthropic's restricted Mythos model and is framed as a national-security and trade-policy issue.

Anthropic has accused the Chinese firm Alibaba of launching the largest attack yet attempting to clone Claude, as China races to match the capabilities of Anthropic's leading model following Mythos' release and subsequent restriction from foreign markets. Ars obtained a June 10 letter sent to Senators Tim Scott (R-SC) and Elizabeth Warren (D-Mass.) one day ahead of a Senate committee hearing on “AI and the American Dream.” In the letter, Anthropic shared “new, confidential evidence of the larges
AI geopoliticsModel theft and distillationAI policyUS-China competition
News AI News & Artificial Intelligence | TechCrunch Jun 25 Old anchor

The White House is asking OpenAI to slow roll the release of its new model over safety concerns

By Lucas Ropek

55 score
AI Analysis

OpenAI reportedly plans a limited-preview release of its next model, GPT-5.6, sharing it only with select partners after the Trump administration requested a staggered rollout over safety concerns. The government would reportedly approve customer access case by case.

OpenAI reportedly plans to share its newest model, GPT 5.6, with a select group of partners instead of with the broader public. The reason: the Trump administration told it to.
AI policyAI safetyGovernment oversightModel release
News AI | The Verge Jun 25 Old anchor

OpenAI will delay GPT-5.6 after Trump administration request

By Hayden Field

55 score
AI Analysis

The Verge reports that OpenAI will release GPT-5.6 in limited preview to a small set of enterprise customers in compliance with a Trump administration request over security concerns, with the government approving access case by case. It is described as a more favorable arrangement than rival Anthropic received.

The Trump administration, apprehensive of potential security issues, has reportedly asked OpenAI to stagger the release of its next big-ticket model, GPT-5.6. The Information reported that OpenAI CEO Sam Altman told employees Wednesday in a company Q&A that it would release GPT-5.6 in limited preview form - granting access only to a small group of enterprise customers - in compliance with a request from the federal government. During that preview period, the Trump administration it
AI policyAI safetyGovernment oversightModel release
News The Decoder Jun 25 Old anchor

Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen

By Matthias Bastian

55 score
AI Analysis

Google has integrated Computer Use directly into the existing Gemini 3.5 Flash model, enabling it to operate computers, browsers, and mobile devices autonomously, scoring 78.4 on the OSWorld benchmark and roughly matching GPT-5.5. Developers can build automation agents via the Gemini API.

Google has integrated "Computer Use" directly into Gemini 3.5 Flash, letting the model operate computers, browsers, and mobile devices on its own. On the OSWorld benchmark, it scores 78.4, putting it on par with GPT-5.5. Developers can use the Gemini API to build agents for software testing or office automation. The article Google bakes computer control directly into Gemini 3.5 Flash, letting the model see and operate your screen appeared first on The Decoder.
Agentic AIComputer useBenchmarksGemini
News AI News & Artificial Intelligence | TechCrunch Jun 25

Europe is pushing back on Washington’s chip war

By Connie Loizos

55 score
AI Analysis

TechCrunch reports European pushback against US export controls, noting that the proposed MATCH Act would bar China from buying older deep-ultraviolet lithography tools that have shipped for roughly a decade. ASML's CEO is cited on the implications for the chip supply chain.

As ASML CEO Christophe Fouquet told TechCrunch in May, what China can currently buy are older-generation deep ultraviolet tools — gear first shipped about a decade ago — the same machines the MATCH Act would now put off-limits.
Chip warAI geopoliticsExport controlsAI policy
News AI News & Artificial Intelligence | TechCrunch Jun 25

Databricks’ former AI chief thinks he can cut AI’s power bill by 1,000x

By Russell Brandom

53 score
AI Analysis

A startup led by Databricks' former AI chief unveiled Un-0, an image-generation system pitched as a demonstration that its technology can replicate conventional AI systems while drastically cutting energy use, claiming up to a 1,000x reduction in power. The work targets AI's escalating power costs.

Un-0 is an image-generation system tool that shows for the first time how the company's technology can replicate conventional AI systems.
Energy efficiencyAI hardwareImage generationStartups