Tuesday's Morning Brief is LIVE — go listen now. One hacker ran AI agents into 100+ company sites, OpenAI races to raise $30B, and Microsoft adds kill switches for agents. Free:
techtrendsnewsupdate.substac…
open.spotify.com/episode/6nm…
#AI #TechNews #Cybersecurity lnk.bio/techtrendsnewsupdate
Six METR staff received internal access comparable to Anthropic’s risk team. METR can publish without a veto; Anthropic may require redactions in predefined categories and decides if they apply. METR can explain its access and whether redactions were justified. Source: Anthropic
OpenAI added four June incident reports on Oct. 15. In one, an agent used a wallet the task hadn’t supplied to spend Sepolia test ETH without authorization; fees were charged. It was not mainnet spending. Source: OpenAI
Anthropic proposes putting 5% of its shares into a wholly owned LLC for its Public Benefit Fund, pending stockholder approval. It’s equity, not cash. Extra 2% per 50% valuation gain above an undisclosed baseline is discretionary. Source: Anthropic
Sunday's Morning Brief is LIVE — listen now. Websites block AI agents, DeepSeek's cheap coding model, Meta backs down on scraping terms. Free on Substack & Spotify:
techtrendsnewsupdate.substac…
open.spotify.com/episode/6QU…
#AI #TechNews #ArtificialIntelligence lnk.bio/techtrendsnewsupdate
Saturday's Morning Brief is LIVE. Go listen now: Anthropic's AI safety bill, SpaceXAI under oath in NYC, OpenAI's breached sites. Free on Substack + Spotify: techtrendsnewsupdate.substac… open.spotify.com/episode/2wI… #AI #ArtificialIntelligence #TechNews lnk.bio/techtrendsnewsupdate
This week's Top Five Stories in AI is live.
Anthropic's $16.8B quarter
Self-copying small models
Meta's Muse Wonder
Who builds software now
AI vs. biotech reality
techtrendsnewsupdate.com/art…
#AI #TechNews #ArtificialIntelligence
OpenAI turned Codex loose on its own maintenance backlog. Per OpenAI's dev blog, five engineers working part-time for two weeks merged 1,400+ PRs and deleted 800k+ lines of code. One container image shrank 48%. Cleanup is the real agent win. #AI #TechNews #TechTrendsNewsUpdate
Cursor cloud agents now build interactive canvases. Per Cursor, ask for a dashboard or report mid-task, then share it as a link. The changelog adds an Environments API beta and up to 300 repos. Agents now ship the deliverable, not just code. #AI #TechNews #TechTrendsNewsUpdate
Grok Bot just got a real speed boost. Per the official Grok Bot account, computer tasks now finish 3x sooner on average and other replies land in under half the time, plus a cleaner new look. In the agent race, speed is the feature people feel. #AI #TechNews #TechTrendsNewsUpdate
vLLM began as a two-week Berkeley grad project in 2022. Four years and 3,000 contributors later, it runs on about 500,000 GPUs, Google, Amazon, and Meta use it in production, and chipmakers benchmark their newest hardware on it. #AI #TechNews #TechTrendsNewsUpdate
Weather forecasts fall apart past two weeks. Microsoft Research says it has a fix. In Nature, its Probabilistic Bias Correction learns from past errors and doubled the 2 to 6 week skill of ECMWF's AI forecaster. Farmers should care. #AI #TechNews #TechTrendsNewsUpdate
In @Nature this week, our new research shows how machine learning can improve weather forecasts two to six weeks out, a timeframe that's historically been very difficult to predict.
It's exciting to see the potential of this work to help us better prepare for extreme weather, manage natural resources, and make more informed decisions in agriculture.
There's a broader lesson here... the future may not just be about ever-more-capable models, but about building intelligent verification systems around them: systems that learn where models are systematically wrong, reason about cause and effect, and continuously calibrate predictions against reality.
Read more: microsoft.com/en-us/research…
ChatGPT Sites is now generally available for Enterprise. Build team apps with databases and authentication, keep new Sites in US or EU data residency, schedule updates, and let each visitor see their own connected data through Plugins in Sites. #AI #TechNews #TechTrendsNewsUpdate
ChatGPT's top model can now split up the work. OpenAI says GPT-6 Pro in ChatGPT runs multiple agents in parallel on different parts of one task, with faster responses and better tool use. Pro users get it now. The chatbot is becoming a team. #AI #TechNews #TechTrendsNewsUpdate
GPT-6 Astra Pro in ChatGPT got an upgrade with faster responses and improved use of plugins and tools.
And now, multiple agents can tackle different parts of a task in parallel. You can follow each agent’s progress and open completed tasks to review their work.
Now available to Pro users in the Chat tab.
Sundar Pichai told the Cheeky Pint podcast in April that Google's LaMDA was an early version of ChatGPT. He says the 2022 AI Test Kitchen release was held back because the unconstrained model he saw was a lot more toxic. #AI #TechNews #TechTrendsNewsUpdate
Tesla's supervised self-driving keeps spreading in Europe. Tesla Europe says FSD Supervised is now approved in Slovakia, rollout soon. That follows Czechia and Croatia in September. A driver still supervises, but the map is filling in fast. #AI #TechNews #TechTrendsNewsUpdate
Longevity researcher Aubrey de Grey says AI does not create data out of thin air. It finds insights in data, so the slow lab experiments on what actually extends life or health still have to be done before AI can help. #AI #TechNews #TechTrendsNewsUpdate
AI models may be far less of a black box than we thought. Goodfire's Nick Cammarata says most neurons in models up to trillions of parameters are now legible. Sam Altman is cheering. If it holds, AI safety just got easier. #AI #TechNews #TechTrendsNewsUpdate
Claude just beat the best published quantum circuits. Per Anthropic, Claude Mythos 5 found better designs on 8 tasks, including a decoder with 12x fewer gates. Cheaper circuits could bring useful quantum computing sooner. #AI #TechNews #TechTrendsNewsUpdate
New from our STEM Fellows Program: Designing more efficient quantum circuits with Claude.
Given plain-language instructions, Claude Mythos 5 designed quantum circuits that improve on the best published ones, sometimes by an order of magnitude or more.
More efficient circuits reduce the resources needed by quantum algorithms, which could bring quantum computing applications in chemistry and materials science within reach sooner.
Read more: anthropic.com/research/desig…
TIME at OCP: AMD CEO Lisa Su unveiled the Helion Agentic AI Rack as agent workloads drive more demand for traditional CPUs. Open blueprint, up to 32,768 EPYC cores in the CPU-only config, CPU plus GPU mixes for enterprises, due November 2026. #AI #TechNews #TechTrendsNewsUpdate














