Skip to content

Aster: a daily almanac of sky, science, games and tools

Daily almanacNº 1

Thursday

November 12

Waxing crescent · 9 %Leonids · Nov 17

Today on Aster: the day’s AI news, 4 new games and a waxing crescent to look at tonight.

Artificial intelligence

Google introduces Gemini Decision 1 to evaluate up to 1,000 decisions at once

Google introduced the Gemini Decisions API, powered by the new Gemini Decision 1 model, for closed questions such as yes-or-no choices, multiple choice, or rubric scoring. The company says the system can process up to 1,000 questions in a single pass across text, images, audio, video, and PDF files. Its output includes probabilities and confidence levels, so it does not replace human judgment in sensitive uses.

Aster editorial team · · via Google

AI-generated image

Latest

Artificial intelligence

UK-based Fractile raises $720 million to develop AI chips

Fractile raised $720 million in a Series C backed by the UK Sovereign AI Fund and led by Thrive Capital, Founders Fund, Lightspeed, and Redpoint. The British government said the deal valued the company at more than £5 billion. Fractile plans to ship its first chips and systems next year, although its chief executive acknowledged that significant work remains.

Aster editorial team · · via TechCrunch

Artificial intelligence

CoreWeave turns to prepayments and shorter contracts to fund expansion

CoreWeave said it will fund part of its growth through customer prepayments, loans backed by specific contracts, and shorter-term deals carrying higher prices. The company expects about $9 billion in prepayments and deposits over the next three months. For its third quarter, it reported $3.66 billion in revenue, a $709 million net loss, and $846 million in interest expense.

Aster editorial team · · via TechCrunch

Artificial intelligence

Tether AI Research introduces BotPsy, a 0.8-billion-parameter local agent

Tether AI Research introduced BotPsy, a 0.8-billion-parameter model for calling tools and running workflows on a device. It is based on Qwen3.5-0.8B and is being released with open weights for research and education. The team says it led five of six tool-use benchmarks among tested models with up to 1.2 billion parameters, while acknowledging limits in open-ended search, difficult tasks, and lower-resource languages.

Aster editorial team · · via Hugging Face

Artificial intelligence

ElevenLabs launches Scribe v3 and a real-time model for speech transcription

ElevenLabs launched Scribe v3 for recorded audio and Scribe v3 Live for low-latency transcription. The real-time version adds speaker diarization and conversation-aware turn detection, while both models support specialized vocabularies. In company tests, Scribe v3 produced 27 % fewer word errors than Scribe v2 across 102 languages, and Live produced 43 % fewer errors than its predecessor.

Aster editorial team · · via ElevenLabs

Today's games

4 challenges that change every midnight. The same for everyone, so you can compare with friends.

See your streaks →

Hidden constellation

Stars appear one by one. Guess the constellation in six tries or fewer.

6 triesPlay

Cosmic word

A five-letter space word. Gold tiles are in place; blue ones belong elsewhere.

6 triesPlay

Which planet am I?

Every clue makes it clearer: planets, moons and stars.

5 cluesPlay

Note of the day

Listen to a short melody and guess it note by note.

6 notesPlay
Artificial intelligence

Google DeepMind introduces Gemini Robotics 2.5 for a broader range of physical tasks

Google DeepMind introduced Gemini Robotics 2.5, a vision-language-action model that is rolling out to its robotics partners. The company evaluated it on more than 200 tasks, from using tongs and tying bows to handling flexible objects. In a proof of concept on a Mercedes-Benz production line, Google reported that Apptronik's Apollo 2 humanoid picked and packed parts with a 96 % success rate.

Aster editorial team · · via Google DeepMind

Artificial intelligence

The Linux Foundation adopts a verifiable identity standard for AI agents

The Linux Foundation welcomed Agent Name Service as an open project for naming, verifying, and discovering AI agents through existing DNS and certificate infrastructure. The specification covers registration, versioning, DNS publication, public-log auditing, integrity monitoring, and agent-to-agent authentication. Eleven organizations support the project, including Cisco, Cloudflare, DigiCert, GoDaddy, Salesforce, and OWASP.

Aster editorial team · · via Linux Foundation

Artificial intelligence

Railway says an attack primarily targeted keys for AI services

Railway published its final report on an October attack and said AI provider keys were the primary target. The company estimates that the attacker could have decrypted variables from up to 1.85 % of its workspaces and that linked-account tokens for 1,852 users were among the affected data. Railway estimates roughly $740,000 in unauthorized use of stolen keys and says it will cover charges that providers do not refund.

Aster editorial team · · via Railway

Artificial intelligence

Researchers link an AI agent to the theft of Tajik government documents

Security firm CloudSek told Reuters that a group linked to Kazakhstan used the open-source Hermes framework to steal more than 2,000 documents related to Tajikistan's national security. The company says the agent performed reconnaissance, wrote tools, downloaded files, and prepared reports under human instructions. Kazakh and Tajik officials did not immediately respond to Reuters' requests for comment, and researchers were unable to contact the named group.

Aster editorial team · · via Reuters

Studio

For musicians: tune up and keep time without installing anything.

Tuner

Guitar, bass, ukulele and voice.

Metronome

30 to 300 BPM, with accents.
Artificial intelligence

Google opens Gemini 4 Carbon to customers and AI Ultra subscribers

Google made Gemini 4 Carbon available to paying developers and enterprises through its API and began rolling it out to Google AI Ultra subscribers. The company positions the model for complex coding, scientific research, document analysis, and cybersecurity, while recommending Gemini 4 Argon for most everyday tasks. It also previewed /swarm in Google Antigravity, which coordinates multiple agents from a single instruction.

Aster editorial team · · via Google

Artificial intelligence

Anthropic introduces Claude Fable 5.2 and Mythos 5.2 with different access levels

Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2, two versions of the same model aimed at difficult technical work, Xataka reported. Fable includes safeguards for high-risk cybersecurity and biology tasks and is generally available, while Mythos is restricted to verified organizations. Anthropic says both produce responses up to 30 % faster than their predecessors and improve coding, reasoning, and vision.

Aster editorial team · · via Xataka

Artificial intelligence

IBM and Microsoft launch Imagine Enterprise to deploy agents in businesses

IBM and Microsoft announced Imagine Enterprise, a platform for building, operating, and governing AI agents connected to internal data and processes. IBM contributes more than 6,000 prebuilt agents and assets, while Microsoft provides cloud, identity, security, and development services. The offering includes engineering support from both companies and begins with frameworks for financial services, insurance, and retail.

Aster editorial team · · via IBM

Artificial intelligence

Australia will require AI incident reports and create a safety regulator

Australia will require companies to report misaligned behavior and cyber incidents involving their AI systems, Reuters reported. It will also create a regulator empowered to review security controls, evaluate models, and impose penalties. The measures follow an official review after an OpenAI agent accessed a Medicare statistics portal without authorization; OpenAI said the access was unintentional and found no evidence that patient records were exposed.

Aster editorial team · · via Reuters

Artificial intelligence

OpenAI launches GPT-6 Cyber, a model that found 81 critical flaws in Firefox

It’s OpenAI’s model for finding software flaws, building working exploits and suggesting patches. The company says it uncovered 81 critical vulnerabilities in Firefox, which Mozilla has fixed, and in the Linux kernel it reproduced 1,724 flaws and built exploits for 948. OpenAI rates it a “Critical” cybersecurity risk and for now offers it only to select Daybreak Red customers. The figures come from the company itself, and there is no independent testing yet.

Aster editorial team · · via The Decoder

Every story in The Weekly Report No. 3 » News archive »

AI glossary

Full glossary »
AI

Language model

An AI system trained on vast amounts of text to understand and generate language.

AI

Token

The small chunk of text (a word, part of a word or a symbol) that a language model reads and writes.

AI

Hallucination

When an AI model makes up facts, quotes or details and presents them confidently as if they were true.

AI

Open weights

An AI model whose trained parameters are published so anyone can download and run it.

AI

AI agent

An AI system that does more than answer: it plans steps and uses tools to complete a task on its own.

AI

Context window

The maximum amount of text, measured in tokens, that a language model can take into account at once.

On Aster we talk about…

Back to top ↑