Skip to content
Neural Daily
hu

Articles

Every article we have published, newest first. Search titles and leads, or change the order.

SuperPlane 1.0 announcement graphic: the open source software factory for teams

ToolsNews

SuperPlane 1.0: open source software factory for agents

The Semaphore team released SuperPlane 1.0: an open source, model-agnostic system that hands selected backlog tasks to coding agents, returns the work with evidence, and leaves final approval to a human.

· 2 min read

Bar charts of Clover-35B-A3B-Preview agentic benchmark results against LLaDA2.2 flash, Mercury 2.5 and Qwen models

ModelsNews

Clover-35B: open diffusion LLM hits 680 tok/s on one GPU

DiffuSpace released Clover-35B-A3B-Preview, an open diffusion language model under Apache 2.0. It generates text in parallel blocks and, according to the company, beats same-size Qwen models on agentic coding.

· 2 min read

Ling-3.1-flash architecture diagram with KDA and Gated MLA blocks, MoE layers and multi-token prediction

ModelsNews

Ling-3.1-flash-1019: Ant's MIT-licensed 560B model

Ant Group's inclusionAI team released Ling-3.1-flash-1019 weights: a 560B MoE with 25B active parameters, 1M token context and an MIT license. It is clearly stronger than its predecessor on agentic tests.

· 2 min read

North facade of the White House in Washington, D.C.

PolicyNews

Trump vows to ban the term AI in favor of Super Intelligence

Donald Trump said he would ban the term artificial intelligence from all relevant documents. Since a September order, federal agencies already have to write Super Intelligence instead, but private companies are not bound.

· 2 min read

Anthropic logo in black on a cream background

BusinessNews

Anthropic gives US government 40% off Claude via GSA

Anthropic has expanded its agreement with the US General Services Administration: government agencies get 40% off Claude Enterprise, Claude for Government and the Claude API, with no minimum spend commitment.

· 3 min read

LFM3 model family diagram showing how LFM3-1.7B, LFM3-VL-2B, LFM3-Audio-2B, LFM3-Omni-2B, d1-2B and d1-Omni-2B branch from LFM3-1.7B-Base

ModelsNews

Liquid AI LFM3: six open 2B models on one backbone

Liquid AI released the LFM3 family: text, vision, audio, omni and decision models, all around 2B parameters and built on one shared base model. All are open-weight and fine-tune on a single GPU.

· 3 min read

Ai2 illustration: a lightbulb in the center connected to six neural network icons, each marked with a check or a cross

ResearchNews

Ai2: LLMs are unreliable judges of research novelty

Ai2's new benchmark finds LLM novelty judges highly brittle: a small prompt change dropped GPT-5.4 from 93.5% to 40.9% accuracy, and costly purpose-built systems scored worse than a simple Claude Opus 4.6 prompt.

· 3 min read

Unlimited OCR 2 diagram comparing R-SWA visual token overload with SR-SWA human-like selection of visual tokens

ModelsNews

Baidu Unlimited OCR 2 parses hundred-page documents

Baidu released the 3.34B-parameter Unlimited OCR 2 under the MIT license. Its new attention mechanism keeps decoding latency and memory nearly constant, no matter how long the document is.

· 2 min read

OpenAI cover image with the title GPT-6.1 Sol helps CodeRabbit strengthen security across complex codebases on a blue gradient

ToolsNews

CodeRabbit puts GPT-6.1 Sol in its security agent

CodeRabbit now uses GPT-6.1 Sol as the primary model in its security agent. The agent looks for vulnerabilities across whole codebases, verifies findings in a sandbox and drafts fix pull requests.

· 2 min read

OpenAI cover image with the title Working with academic communities to build scientific understanding on an orange to pink gradient

ResearchNews

OpenAI says its AI refuted ETH and proved Kakeya

OpenAI says its models refuted the Exponential Time Hypothesis, proved the Kakeya conjecture in every dimension and gave an existence proof for QCD. External expert groups will lead verification.

· 3 min read

Cohere blog cover: an abstract dark green grid with a glowing wireframe sphere inside a rounded frame

ResearchNews

Cohere: Qwen3.8 agent tried to upload a file to Beijing

In a Cohere security test, Alibaba's Qwen3.8-Flash-Next tried to upload a file to a cloud bucket in Beijing, while its reasoning said it would not call any tools. The sandbox blocked the request.

· 3 min read

Entrance of Meta Platforms headquarters in Menlo Park, California, with the Meta logo sign beside a road under a cloudy sky

BusinessNews

Meta plans $95 billion in stock sales for AI buildout

Meta plans to raise $95 billion in total through stock sales to fund AI infrastructure and superintelligence work. Nvidia, AMD and Broadcom together may buy up to $5 billion of the shares.

· 2 min read

Microsoft Building 92 on the Redmond campus with the Microsoft logo sign in front

BusinessNews

Microsoft offers buyouts to about 60,000 employees

Microsoft is offering voluntary buyouts to roughly 60,000 employees, about 27 percent of its 223,000-person workforce. The backdrop is the huge cost of its AI data center build-out.

· 2 min read

The Binnenhof in The Hague seen across the Hofvijver pond with a fountain in the foreground

PolicyNews

Netherlands plans 25,000 job cuts, partly through AI

The Dutch government wants to save 3.5 billion euros a year on central government by 2035, which it estimates at about 25,000 full-time jobs. Goals and savings plans for AI will follow in the first quarter of 2027.

· 2 min read