EARLY RELEASEGemini 4 launch timeline set by new DeepMind chief
Google DeepMind's new leader says the next flagship model is in refinement and aims for an early release.
EVERY PUBLISHED BRIEF
AI briefs.
Browse Crunchy Tech’s latest ai briefs, grouped by date.
158 briefs · Page 2 of 7
PUBLISHED
Thursday, 24 September 2026
PUBLISHED
Wednesday, 23 September 2026
TASK AUTOMATIONChatGPT Voice gains calendar, email and Slack access
OpenAI's GPT-6-powered voice assistant can now handle tasks like booking meetings and cancelling duplicate charges.
FEARS PERSISTAI users still fear regulation despite daily use
A new report suggests that familiarity with artificial intelligence does not calm public concerns about the technology.
COST CUTRingg’s AI agents resolve up to 65% of calls
Ringg uses OpenAI’s GPT-5.6 to build multilingual agents that handle voice, chat, WhatsApp and web for 90% less cost than GPT-4.1.
GENE EDITINGAnthropic’s AI finds new enzyme system in DNA database
Claude autonomously discovered a new enzyme system similar to Crispr, marking the first result from Anthropic’s wet lab and a test of its usefulness for science ahead of its IPO.
SKILLS ACCESSOpenAI Academy marks two years of expansion
OpenAI has marked two years of its Academy initiative, aiming to broaden access to AI skills.
SECURE MEMORYGoogle brings private, persistent AI memory to the cloud
A new secure memory layer lets AI assistants remember you across devices without exposing your data.
TEST STRATEGIESYouTube adds AI tools to help creators optimise performance
YouTube is rolling out new creator tools that use generative AI to test video strategies, generate dynamic thumbnails and provide live dubbing.
COST SLASHAlibaba slashes AI audio prices with Qwen Audio 3.1
Alibaba's Qwen team introduces five new audio models and cuts costs by up to 95 percent.
GENETIC DATABasecamp Research raises $140m to train AI on biology
London startup uses genetic data to design drugs and therapies.
REAL TIME AUDIONVIDIA releases Nemotron 3 for real-time speaker diarization
NVIDIA has released a new model designed to identify who is speaking in real-time conversations.
HUMAN IN LOOPKDE Drafts AI Guidelines: 'Don't Be a Meat Proxy'
KDE Plasma outlines strict rules for AI-assisted contributions, demanding human oversight and banning AI-generated text.
PUBLISHED
Tuesday, 22 September 2026
SAFETY STANDARDSOpenAI sets rules for third-party AI safety checks
OpenAI outlines priorities and principles for rigorous, secure, and independent third-party AI safety assessments of frontier models and safeguards.
AUTONOMOUS CONTROLAstroForge puts AI in charge of spacecraft
AstroForge plans to use a transformer-based AI model to control its next space probe, aiming for autonomous operations.
100 PROBLEMSOpenAI claims internal model solved 100 math problems in a month
OpenAI says a new internal model solved over 100 long-standing math problems after just a month of training, prompting the company to set up an independent advisory group.
TSAR APPOINTEDTrump launches AI Force, rejects slowdown calls
The US president appoints a new AI tsar and creates an AI Force while dismissing safety concerns as a hoax, despite rising public anxiety over AI risks.
PUBLISHED
Monday, 21 September 2026
EXPERTSOpenAI forms math advisory group
OpenAI is working with an independent group to guide the review of emerging AI results as its models solve more than 100 open problems.
AD CREATION SPEEDHiggsfield ships GPT-6 Astra video tools
Higgsfield AI releases new video features powered by GPT-6 Astra, aiming to speed up ad creation for small businesses.
PIPELINE COMPLETEByteDance launches Dramagic for AI short dramas
ByteDance introduces Dramagic, an end-to-end AI platform for creating short dramas, sold through BytePlus.
LIFECYCLE PROOFNVIDIA launches Halos safety stack for physical AI
A new framework aims to prove that AI-driven vehicles and robots behave safely across hardware, software and the operating environment.
TRANSPARENCYUS and China agree on AI dialogue ahead of Trump-Xi summit
Washington and Beijing will set up a notification mechanism for AI incidents as part of a dialogue ahead of the leaders' meeting.
CONTEXT SOURCEV7 gives AI agents institutional memory
OpenAI's V7 model converts company files into context for agents to complete complex, source-linked work.
PUBLISHED