Good morning, { AI enthusiasts }. Sam Altman wasn’t just able to welcome back Considering Machines talent — he’d been talking to them for months.
Recent reports reveal last week’s messy split at Mira Murati’s startup had been brewing behind the scenes, with co-founders unhappy with the corporate’s direction and OpenAI positioned to scoop up the pieces when things fell apart.
In today’s AI rundown:
-
Contained in the Considering Machines meltdown
-
Google brings on Hume’s CEO, engineers
-
Learn any subject with NotebookLM podcasts
-
Runway: 90% cannot spot AI-generated videos
-
4 recent AI tools, community workflows, and more
LATEST DEVELOPMENTS
OPENAI & THINKING MACHINES

Image source: The Rundown
The Rundown: Recent details are emerging about last week’s split at Mira Murati’s Considering Machines Lab, in response to the NYT and The Information, including a tense meeting, secret talks with Sam Altman, and a failed acquisition discussion with Meta.
The small print:
-
CTO Barrett Zoph and two other co-founders reportedly confronted Murati just days before his firing, pushing for him to take control of technical decisions.
-
Murati allegedly told Zoph to “just do his job” as CTO and fired him later that week, with nine other TML employees also leaving for OAI or receiving offers.
-
Zoph was reportedly talking to Altman behind Murati’s back for months, with the co-founders unhappy with TML’s direction and pushing for a sale to Meta.
-
Zoph will now lead OAI’s enterprise AI sales plans as a part of a reorg to assist close the gap between research and product teams.
Why it matters: What began as a messy departure now looks more like (one other) power struggle that had been brewing for months. With TML struggling to boost at its $50B goal valuation and co-founders pushing for a sale Murati didn’t want, the cracks were already showing — and OAI and Altman were able to catch the fallout.
TOGETHER WITH AUTH0
The Rundown: AI agents are reshaping digital experiences, but securing them requires rethinking identity and access controls built for a human-first world. Auth0 for AI Agents is now generally available, giving developers the whole auth solution for constructing AI agents securely.
Key features include:
-
Token Vault to attach agents to users’ apps securely
-
Human-in-the-loop approvals for critical actions
-
Nice-grained authorization for RAG to forestall data leaks
-
Native support for LangChain, LlamaIndex, Cloudflare Agents, and more
GOOGLE & HUME

Image source: Google
The Rundown: Google DeepMind hired Hume AI’s CEO and roughly seven engineers as a part of a brand new licensing agreement, marking the most recent acqui-hire move by the tech giant — with Hume’s emotionally intelligent voice tech set to be integrated into Gemini.
The small print:
-
CEO Alan Cowen and the engineering hires will help integrate voice and emotional intelligence into Google’s models, including Gemini’s voice features.
-
Hume AI will keep operating under recent CEO Andrew Ettinger, continuing to provide voice training data and tools to other labs constructing conversational AI.
-
The deal follows Google’s $3B Character AI licensing play last 12 months and mirrors similar moves by Microsoft (Inflection) and Meta (Scale AI’s CEO).
Why it matters: Voice is quickly becoming one in every of the most important ways people interact with AI — and understanding tone, emotion, and nuance is what separates a useful assistant from a clumsy, clunky one. Plus, with AI wearables and devices getting closer than ever to mainstream, upgraded voice capabilities will change into much more vital.
AI TRAINING

The Rundown: On this tutorial, you’ll learn how one can generate a podcast on any topic using Google’s NotebookLM, a system you should utilize to make a substantial dent within the firehose of data you receive every day.
Step-by-step:
-
Go to notebooklm.google.com and create a brand new notebook. Add sources corresponding to YouTube links, PDFs, or website links
-
You’ll be able to manage what sources go into the AI’s brain by checking and unchecking them within the left sidebar
-
Under “Studio” on the precise, click “Audio Overview” to start out generating a podcast. It can take 5-10 minutes to generate and can show up on the precise
-
Once generated, click the three dots to download the mp3 to your phone or listen directly via the browser or NotebookLM mobile app
Pro tip: “Call in” to ask Qs to the AI podcast hosts by clicking the “interactive” button.
PRESENTED BY SCROLL AI
The Rundown: Scroll.ai turns your knowledge base into AI-powered experts that automate the workflows draining your team’s time, with accuracy and depth far beyond generic models.
With Scroll, you’ll be able to:
-
Deploy custom AI experts that deeply understand your domain
-
Automate compliance questionnaires, RFPs, and documentation in minutes
-
Equip sales and marketing teams with reliable, on-demand expertise
Try Scroll today and use code THE-RUNDOWN-2026 for 2 free months of the Starter plan.
AI RESEARCH

Image source: Runway
The Rundown: Runway published research showing that 90%+ of participants couldn’t tell between real footage and image-to-video clips generated by its Gen-4.5 model, with the corporate calling the capabilities a societal “tipping point” for AI media.
The small print:
-
Over a thousand participants watched 20 five-second clips and tried to discover which were AI-generated, with just 99 scoring 75% or higher.
-
Nature scenes and buildings were graded the toughest, with generated versions in those categories proving more convincing than the actual videos.
-
Gen-4.5 currently sits atop Artificial Evaluation’ text-to-video rankings, with Runway rolling out the brand new image-to-video features this week.
Why it matters: Low-quality AI videos are already tricking people, and Runway’s Gen-4.5 and other models take things to a brand new level of realism that’s virtually unattainable to detect. Despite a push for verification standards, no fix can solve for a world where “seeing is believing” now not stands.
QUICK HITS
Google DeepMind co-founder Shane Legg posted a brand new ‘Chief AGI Economist’ job opening, saying that “AGI is now on the horizon”.
OAI CFO Sarah Friar floated “value sharing” at Davos, where the corporate would take a licensing stake in discoveries made using its AI, starting with drug development.
Amazon One Medical launched a brand new agentic health assistant in a position to answer questions on personal medical insights, test results, medications, and booking care.
xAI released a brand new update to its Grok Imagine creative suite, now allowing for generations as much as 10 seconds long with improved audio and video capabilities.
Google added its ‘Personal Intelligence’ feature into AI Mode for search, allowing the experience to attach and reference info from users’ Gmail and Photos.
Yelp agreed to acquire AI startup Hatch for $270M, adding tools that automate customer communication for service businesses to its platform.
COMMUNITY
Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.
Today’s workflow comes from reader Nicole S. within the UAE:
“One small way AI has upgraded my each day routine: turning reading time into movement time. I exploit NotebookLM to create audio summaries from newsletters, PDFs, and URLs. I turn them into an audio file, then listen through short yoga stretch, home tasks, or lunch—giving my eyes a rest while the educational continues.”
How do you employ AI? Tell us here.
-
Read our last AI newsletter: Anthropic writes Claude’s Structure
-
Read our last Tech newsletter: Biotech’s big 3 (and so they’re wild)
-
Read our last Robotics newsletter: A crawling, detachable robot hand
-
Today’s AI tool guide: Learn any subject w/ custom NotebookLM podcasts
That is it for today!
- ⭐️⭐️⭐️⭐️⭐️ Nailed it
- ⭐️⭐️⭐️ Average
- ⭐️ Fail
Login or Subscribe to participate
See you soon,



