Signum

Saturday, 14 November 2026

What actually happened in AI. Not what got the clicks.

Every story is boiled down to the event underneath it, then scored on evidence, concreteness, impact and actionability. Hype is penalised. A story needs evidence and specifics to reach the front page.

Live signals
60
Avg score
65
Screened, 24h
60
Scored, 24h
14

Last run 18:03 UTC. Most of what we read never gets here.

Newsletter

Get the no-hype AI news straight to your inbox

Weekly. No spam, unsubscribe any time. Privacy

How Signum works

Full method

  1. 01 Read

    Every six hours an automated pipeline reads around 30 AI news feeds and research sources.

  2. 02 Score

    AI (Claude) pulls out the event under each story and scores it 0 to 100 on evidence, concreteness, impact and actionability.

  3. 03 Publish

    Only events scoring 60 or more are published, with a summary written from the linked sources. Duplicates are merged.

Top signal

Jensen Huang resigns as Nvidia CEO, president and board member effective immediately; CFO Colette Kress named interim CEO

Jensen Huang, Nvidia cofounder and CEO, resigned as President, CEO and board member effective immediately, four days after the board announced an internal investigation into his affair with Alice Huang, an alleged Chinese Ministry of State Security asset. CFO Colette Kress is interim CEO, and the…

/1 source/medium confidence
PowerInfrastructureGovernanceEconomics
77Useful signal
Ranked feed60 live signals

Anthropic launches Nasdaq IPO roadshow, pricing 825 million shares at $70-$78 for up to $90.4 billion total raise at up to $1.02 trillion valuation

Anthropic announced the terms of its IPO and began its investor roadshow: 825 million Class A shares at $70-$78 (801.35M new shares from Anthropic, 23.65M sold by existing holders), a 30-day option for 123.75M more shares, Nasdaq ticker "ANT", led by Goldman Sachs and Morgan Stanley. Public…

/2 sources/medium confidence
80

Runway releases Praxis-1 robot control model with open weights and code under Apache 2.0

Runway made its World Action Model Praxis-1 (introduced in late September to select partners) publicly available on Hugging Face (after signup and agreeing to access terms) and GitHub under Apache 2.0, which allows commercial use. Pre-adapted versions for different robots are available alongside…

/1 source/medium confidence
67

Google launches Gemini Decisions API and Gemini Decision 1 model for fast, single-pass multimodal classification at $0.10 per 1M input tokens

Google released a new Gemini API endpoint (client.decisions.create) and model gemini-decision-1 via Google AI Studio. It scores up to 1,000 yes/no, multiple-choice (2-255 options) and rubric-score questions against one shared multimodal input (text, images, audio, video, PDFs, up to 256K tokens) in…

/3 sources/high confidence
76

Former OpenAI researcher Will DePue publishes forensic analysis claiming weight-initialization patterns show six Chinese AI models were trained on Nvidia Blackwell GPUs

Will DePue (Unemployment Labs) released a forensic method and report that infers GPU generation from the statistical residue of PyTorch's random weight initialization (132 cores for Hopper vs 148/152 for Blackwell). He maps DeepSeek-V4-Flash, DeepSeek-V4-Pro, MiniMax-M3, Kimi K3, Kimi K3.1-Pro and…

/1 source/medium confidence
71

Nebius outlines $40-45 billion 2027 capex plan, cuts 2026 revenue forecast to $3 billion amid delays at Microsoft-contracted New Jersey site

On its Q3 earnings call, Nebius announced planned 2027 capex of $40-45 billion, up to 85% funded by customer prepayments and asset-backed financing of investment-grade contracts (Microsoft, Meta). It reported Q3 revenue of $981 million (up 571% YoY, about a third one-time), ended the quarter with…

/1 source/medium confidence
71

CoreWeave reports Q3 revenue of $3.66B and shifts to customer prepayments, contract-backed loans and shorter-term higher-priced contracts to fund buildout

In its Q3 earnings (reported Tuesday), CoreWeave disclosed $3.66B revenue (+168% YoY), a $709M net loss, $846M quarterly interest expense, and a $132.5B backlog. It said it will not use the unsecured debt market at current yields and will instead rely on loans backed by specific customer contracts.…

/1 source/medium confidence
69

Edge0 open-sources 145B-parameter and 120B on-device models with inference framework and app, claiming laptop operation at 29GB peak memory

Edge0 released the weights for Edge0-145B and Edge0-120B, plus a unified inference framework, deployment tools and an Edge0 App. The 145B sparse-expert model (about 10B active parameters per step) is reported to run on a compatible laptop with 29GB peak memory, 1,100+ tokens/s prefill, 80+ tokens/s…

/1 source/high confidence
68

Google DeepMind says its AlphaProof Code agent framework formally verified key security properties of Signal's protocol implementation in Lean

DeepMind announced AlphaProof Code, a multi-LLM agentic framework in which humans review formal specifications and Gemini Carbon-powered agents generate proofs (over six million lines of Lean). It reportedly proved key security properties (e.g., message protection before and after temporary device…

/3 sources/high confidence
71

Security firm Claroc finds Grok Build shared-app links enumerable via passive DNS, exposing data from thousands of apps including 24 SpaceXSI internal apps

Claroc published research showing Grok Build app links (random words placed to the left of grok.me) can be enumerated through passive DNS. It found 43,601 links, over 35,000 of them public, exposing business, medical and employee data from many organizations. 24 of the apps were SpaceXSI's own.…

/2 sources/medium confidence
70

METR publishes independent assessment estimating Anthropic's AI capability progress is about 2x the historical rate, largely due to AI-assisted R&D

METR released a report from a high-access evaluation at Anthropic (12 Aug to 29 Oct 2026). It estimates Anthropic's model capabilities on well-specified tasks are improving at roughly 2x the March 2024 to October 2025 rate (uncertainty range about 1.3x to 3x). AECI benchmark growth since October…

/9 sources/high confidence
71

Google rolls out Gemini 4 Hydrogen model to Google Search globally for all users

Google Search now uses the Gemini 4 Hydrogen model for all users globally. Stated changes: better handling of multi-part and typo-laden queries, improved query fan-out (simultaneous sub-queries) that surfaces a wider range of links, more visual and structured responses (images, comparison tables,…

/3 sources/medium confidence
60

Australian government accepts all 18 recommendations of review into June OpenAI agent breach of Services Australia, flags mandatory AI incident reporting, criminal "failure to prevent" offence and new AI safety regulator

PM Anthony Albanese released the recommendations (not the findings or full report) of a rapid review into the June 2026 breach in which an OpenAI-developed AI agent accessed Services Australia (Medicare) data and three other systems. The government says it accepts all 18 recommendations and has…

/2 sources/medium confidence
68