What moves in AI, fact by fact, with its sources. Subscribe to the RSS feed
The feed summaries are written by an AI. Statuses come from the registry; the “not reviewed” label applies to the text.
Models and labs
Meta introduces Muse Wonder and Muse Wonder Full, its most capable models, available via the Meta Model API and Muse Code (Wonder) and to Muse Code subscribers (Full).
Available
Source : about.fb.com +2
Written by an AI · not reviewed
Science with AI
a contributor proposed on 11 October 2026 (integer-mult-bounds PR #514) a conditional bound κ = 0.001391523459591829 for multiplying large integers. The certificate can be replayed; the request is open, not merged. Status: not yet evaluated.
Not evaluated
Source : github.com +2
Written by an AI · not reviewed
Science with AI
Physicists post on arXiv 50 unpublished theoretical physics problems with numerical answers and test five AI agents on them. The best, GPT-6-Astra with Codex, gets 49 of 50 problems right with web access, 46 without. Preprint not peer-reviewed; one run per agent, no timed humans.
Not evaluated
Source : arxiv.org +1
Written by an AI · not reviewed
Science with AI
Science publishes a formal-proof-search agent (Lean) that resolved nine of 353 open Erdos problems and 44 of 492 OEIS conjectures, per the abstract.
Peer reviewedbatch
Source : api.crossref.org +1
Written by an AI · not reviewed
Apps and SaaS
Claude can build a live dashboard connected to a data platform and explainer animations exportable as MP4.
Available
Source : claude.com
Written by an AI · not reviewed
Apps and SaaS
Midjourney is testing on its Alpha site a mode that reruns an image with additional reasoning.
Demo only
Source : updates.midjourney.com
Written by an AI · not reviewed
Apps and SaaS
With GPT-6, ChatGPT combines text, visuals and interactive elements (comparisons, calculators, games) and picks the format itself.
Available
Source : openai.com +1
Written by an AI · not reviewed
Models and labs
Anthropic releases Claude Haiku 5.5, its small model, the first Haiku with an adjustable effort setting and an average cost about 75% lower than Haiku 4.5.
Available
Source : anthropic.com +2
Written by an AI · not reviewed
Models and labs
Liquid AI releases two open decision models that answer in a single forward pass without generating tokens, aimed at NVIDIA edge hardware.
Available
Source : liquid.ai +1
Written by an AI · not reviewed
Science with AI
On 7 October 2026, OpenAI withdrew three manuscripts from its mathematics release, on Weil classes, Kuga-Satake K3 and the rational Hodge conjecture, after a sign error. Status: withdrawn.
Withdrawnbatch
Source : raw.githubusercontent.com +1
Written by an AI · not reviewed
Apps and SaaS
Cursor's iOS app lets users see and reply to agents running on their own computer.
Available
Source : cursor.com
Written by an AI · not reviewed
Science with AI
Google Research posts a preprint and blog where place embeddings improve epidemiological models (MMR vaccination, dengue in Mexico, cholera in DRC, etc.).
Not evaluated
Source : arxiv.org +1
Written by an AI · not reviewed
Models and labs
Mistral opens a public preview of Mistral Large 4, a natively multimodal 1-trillion-parameter model with 52 billion active parameters, weights promised by end of month.
Available
Source : mistral.ai +2
Written by an AI · not reviewed
Science with AI
On 6 October 2026, OpenAI published the openai/math repository: 719 manuscripts (722 at launch) in 372 families across 17 disciplines, from an unreleased internal model. Status: under review; almost all its entries are not evaluated, one is contested, one withdrawn.
Under reviewbatch
Source : raw.githubusercontent.com
Written by an AI · not reviewed
Science with AI
Nature Biotechnology publishes a retinal vascular graph model that predicts pre-eclampsia before clinical onset, with external validation without retraining.
Peer reviewed
Source : nature.com
Written by an AI · not reviewed
Science with AI
Arc Institute presents PIE, which predicts which genes change after a perturbation, with AUPRC 1.2 to 3.2 times that of the strongest competitor on the Replogle-Nadig dataset.
Not evaluated
Source : arcinstitute.org +1
Written by an AI · not reviewed
Models and labs
Aleph Alpha releases Kolibri 1, a mixture-of-experts reasoning model (78B parameters, 3.46B active) for German and English, under Apache 2.0.
Available
Source : huggingface.co +1
Written by an AI · not reviewed
Science with AI
Matthew Schwartz (Harvard) posts on arXiv a toolkit for exact computation (Feynman integrals, completeness proofs) operable by AI agents, applied to genomics, ecology and other fields.
Not evaluated
Source : arxiv.org +1
Written by an AI · not reviewed
Devices and robots
Boston Dynamics presents Atlas's new hand: 13 degrees of freedom, directly actuated, built for high-fidelity simulation and mass manufacturing (7 degrees on the previous generation).
Demo only
Source : bostondynamics.com +1
Written by an AI · not reviewed
Apps and SaaS
A "Try on" button on ChatGPT product listings generates a virtual try-on from a selfie using ChatGPT Images.
Available
Source : help.openai.com
Written by an AI · not reviewed
Apps and SaaS
With the camera shared, Gemini Live describes surroundings aloud and guides framing with spoken cues.
Available
Source : blog.google
Written by an AI · not reviewed
Apps and SaaS
Copilot can operate desktop applications (click, type, scroll) on macOS and Windows, including software without an API.
Available
Source : github.blog
Written by an AI · not reviewed
Apps and SaaS
Suno opens a beta to everyone that generates spoken audio with original background music as a single track.
Available
Source : suno.com
Written by an AI · not reviewed
Models and labs
Google announces Gemini 4 Argon, a new frontier model rolling out first to trusted cyber defenders, with output limit raised to 1 million tokens.
Announced
Source : blog.google +2
Written by an AI · reviewed by a human
Apps and SaaS
Gemini can build skills from chats, run them automatically on matching prompts, and stack several at once.
Available
Source : gemini.google
Written by an AI · not reviewed
Science with AI
Nature Machine Intelligence publishes a pretrained sparse Vision Transformer that, on simulated FASERCAL events, matches scratch training with an order of magnitude more data using about 10^3 labelled events.
Peer reviewed
Source : nature.com
Written by an AI · not reviewed
Science with AI
A Microsoft Research post describes a system estimating space-weather risk per US substation and detecting nearly 80% of major events in the evaluation period.
Not evaluated
Source : microsoft.com
Written by an AI · not reviewed
Science with AI
Google DeepMind publishes in Nature watermarking methods for AI-generated protein sequences and structures; watermarked binders tested in the lab matched unwatermarked ones.
Peer reviewed
Source : nature.com +1
Written by an AI · not reviewed
Models and labs
OpenAI releases GPT-6.1 Sol, an upgrade of GPT-6 Sol that it says nearly matches Astra on agentic coding, computer use and professional work at one fifth of the price.
Available
Source : openai.com +2
Written by an AI · not reviewed
Apps and SaaS
ChatGPT now offers always-on agents that keep working between conversations, each with its own cloud computer and connected apps.
Tested by a third party
Source : help.openai.com +1
Written by an AI · reviewed by a human
Science with AI
Microsoft Research introduces Quine, a biology world model with a tool harness, and says its top-ranked compounds produced the largest shifts in tumour cell states.
Not evaluated
Source : microsoft.com +1
Written by an AI · not reviewed
Science with AI
Nature Biotechnology publishes a Bayesian-optimisation framework that found solid-state mRNA-LNP formulations in six iterations that kept bioactivity at 37 °C.
Peer reviewed
Source : nature.com
Written by an AI · not reviewed
Models and labs
Anthropic releases Claude Sonnet 5.5, second model of the 5.5 family, over 30% faster than Sonnet 5 at the same per-token price.
Available
Source : anthropic.com +2
Written by an AI · not reviewed
Science with AI
Nature Machine Intelligence publishes GenFocal, a generative framework producing fine-scale heatwaves and cyclones from coarse climate projections without temporally aligned training pairs.
Peer reviewed
Source : nature.com
Written by an AI · not reviewed
Apps and SaaS
Per specialist press, Google Assistant is gone from the Android devices checked in the US and Europe, replaced by Gemini.
Discontinued
Source : 9to5google.com
Written by an AI · not reviewed
Apps and SaaS
A Team Bot is given files, plugins and skills, then shared so a whole team works from the same context.
Available
Source : x.ai
Written by an AI · not reviewed
Science with AI
Ben Antieau and colleagues, working with a language model, posted an algebraic geometry text on Hexagon on 26 September 2026. The platform labelled it "Primarily AI-generated". Status: not yet evaluated.
Not evaluated
Source : hexagonmath.org
Written by an AI · not reviewed
Apps and SaaS
Microsoft announces a unified Copilot with a persistent agent (Autopilot), a natural-language app builder (Code) and a Home entry point.
Announced
Source : blogs.microsoft.com
Written by an AI · not reviewed
Science with AI
EMBL-EBI, Google DeepMind and NVIDIA add predicted viral protein complexes to the AlphaFold Database through a pandemic preparedness portal.
Not evaluated
Source : embl.org +1
Written by an AI · not reviewed
Devices and robots
Skydio announces the F10 Lightrunner, its first fixed-wing drone, designed to launch in seconds and fly up to 100 mph, with a MegaDock.
Announced
Source : skydio.com
Written by an AI · not reviewed
Science with AI
Anthropic reports that Claude agents spotted, in DNA data, a system (ART) pairing a reverse transcriptase, a partner gene and a repeat array, whose function is unknown.
Not evaluated
Source : anthropic.com +3
Written by an AI · not reviewed
Apps and SaaS
Rollouts watches each change through deployment and Security Review flags exploitable bugs on every pull request.
Available
Source : cursor.com
Written by an AI · not reviewed
Devices and robots
The FDA lists a Substantially Equivalent decision on 2026-09-23 for VIRTUE QA (ProNova, K260597), AI quality assurance for proton therapy, with an authorized change control plan.
Announced
Source : accessdata.fda.gov +2
Written by an AI · not reviewed
Devices and robots
Meta announces for Ray-Ban Display glasses navigation that cites visible landmarks, a voice-driven face hologram for WhatsApp calls, and expansion to five countries.
Announced
Source : about.fb.com +1
Written by an AI · not reviewed
Devices and robots
Meta presents 43 g audio glasses with an AI assistant (pre-order from USD 349, shipping October 13) and announces its Muse agent coming to its AI glasses in the coming months.
Announced
Source : about.fb.com +2
Written by an AI · not reviewed
Science with AI
A Stanford/Arc preprint introduces Minerva, which predicts interactions from sequence and describes prophage reverse transcriptase systems encoding arrays of non-coding RNAs.
Not evaluated
Source : biorxiv.org +1
Written by an AI · not reviewed
Devices and robots
Skild AI announces a self-play post-training result: a policy learned in simulation (Isaac Sim) is transferred to a humanoid for a soccer match.
Demo only
Source : skild.ai +1
Written by an AI · not reviewed
Models and labs
Anthropic introduces Claude Opus 5.5, first model of the Claude 5.5 family, which it says matches Fable 5.1 on most work at 40% lower cost than Opus 5.
Available
Source : anthropic.com +2
Written by an AI · not reviewed
Science with AI
Nature Medicine publishes an AI model for oesophageal cancer screening on non-contrast chest CT, validated on 80,612 patients in 12 centres across three countries.
Peer reviewed
Source : nature.com
Written by an AI · not reviewed
Models and labs
OpenAI adds two faster, cheaper models to the GPT-6 family, Sol and Luna, trained like Astra, with API prices cut 50% versus GPT-5.6 promotional pricing.
Available
Source : openai.com +3
Written by an AI · not reviewed
Models and labs
xAI releases Grok 4.7, described as its most capable model for coding and knowledge work, at the same price as Grok 4.6, on a new, larger base model.
Available
Source : x.ai +2
Written by an AI · not reviewed
Models and labs
Xiaomi publishes on Hugging Face, under MIT licence, MiMo-V2.6-Pro-RL, an omnimodal model of 1.02T total and 42B active parameters with 1M-token context.
Available
Source : huggingface.co +2
Written by an AI · not reviewed
Science with AI
Microsoft Research publishes in Nature a retrosynthesis model that ensembles two complementary models; chemists accepted about 90% of its complete routes.
Peer reviewed
Source : nature.com +1
Written by an AI · not reviewed
Devices and robots
The FDA 510(k) database lists a Substantially Equivalent decision on 2026-09-18 for Deterioration Index Version 2 (Epic Systems, K260008), an AI clinical-deterioration risk score.
Announced
Source : accessdata.fda.gov +1
Written by an AI · not reviewed
Apps and SaaS
CC gets its own Google account, produces a shared daily brief, updates calendar and tasks, and can fill in forms with permission.
Available
Source : blog.google
Written by an AI · not reviewed
Devices and robots
Figure presents Helix 2.5, a model pretrained on its Index dataset that tidies living rooms, folds towels and makes beds in 30 homes with no data collected in them.
Demo only
Source : figure.ai +1
Written by an AI · not reviewed
Apps and SaaS
OpenAI pairs GPT-6 Astra with a US legal search index and legal-analysis instructions, offered to selected law firms.
Available
Source : openai.com
Written by an AI · not reviewed
Science with AI
Stanford researchers describe a system of 37,000 AI agents that analysed clinical trials and proposed a drug strategy (an antibody-drug conjugate targeting B7-H3) for lung cancer.
Peer reviewed
Source : api.crossref.org +3
Written by an AI · not reviewed
Apps and SaaS
Anthropic makes Claude Cowork capabilities available from any chat (reports, spreadsheets, presentations) without picking a mode first.
Available
Source : support.claude.com
Written by an AI · not reviewed
Devices and robots
Agility unveils Digit 5, adding AI-based person detection, safety cues and an independent safety controller to work near people without barriers.
Announced
Source : agilityrobotics.com +1
Written by an AI · not reviewed
Apps and SaaS
New DeepL Voice models keep each speaker's voice and expressiveness across languages, and a desktop app covers Zoom, Teams and Meet.
Available
Source : deepl.com
Written by an AI · not reviewed
Models and labs
Google launches two real-time voice models, one of which reasons and speaks simultaneously, available through the Gemini API, Workspace and the Gemini app.
Available
Source : blog.google +2
Written by an AI · not reviewed
Apps and SaaS
Gemini Notebook lets users talk to their notebooks in real time and generates interactive study overviews (quizzes, flashcards, short videos).
Available
Source : blog.google
Written by an AI · not reviewed
Apps and SaaS
Kakao is ending the Kanana app and PC service about 17 months after launch, shifting the work to KakaoTalk.
Discontinued
Source : digitaltoday.co.kr
Written by an AI · not reviewed
Apps and SaaS
Apple rolls out a rebuilt Siri with personal context, onscreen awareness and systemwide app actions.
Available
Source : apple.com
Written by an AI · not reviewed
Apps and SaaS
The Perplexity Computer agent runs locally on Windows PCs with a 24 GB+ RTX GPU, going to the cloud only on request.
Available
Source : blogs.nvidia.com
Written by an AI · not reviewed