Good morning, AI enthusiasts. OpenAI’s Sora AI platform has seemingly been breaking bad on celebrity likenesses since its launch — but Bryan Cranston just made it literal.
With a brand new joint statement from the AI leader and Hollywood agencies that vows to tighten guardrails, one in all the AI video app’s biggest viral strengths is quickly becoming a liability.
In today’s AI rundown:
-
OpenAI to tighten Sora guardrails after Hollywood complaints
-
Anthropic brings Claude Code to the browser
-
Automate performance reviews and PIPs
-
Napster pivots to AI with holograhphic companions
-
4 latest AI tools, community workflows, and more
LATEST DEVELOPMENTS
OPENAI
📺 OpenAI to tighten Sora guardrails after Hollywood complaints

Image source: Reve / The Rundown
The Rundown: OpenAI just released a joint statement alongside actor Bryan Cranston, major Hollywood agencies, and the SAG-AFTRA union, saying it would ‘strengthen guardrails’ on Sora 2 over replication of celebrity’s likenesses and voices.
The main points:
-
Cranston found AI videos of himself circulating on Sora 2, including one showing him taking selfies with Michael Jackson — despite never opting in.
-
OpenAI apologized for the “unintentional generations”, vowing to implement stronger guardrails in a “productive collaboration” with the union and agencies.
-
The statement also rallied support for the NO FAKES Act, calling for protections against unauthorized AI replicas of performers’ voices and likenesses.
-
SAG-AFTRA President Sean Astin warned that Cranston represents “countless performers” in danger, calling opt-ins “the one method to do business” with AI firms.
Why it matters: Sora has been breaking bad (🥁) since launch, and it’s no surprise to see motion from estates like Martin Luther King Jr. and now Cranston. While there have been opt-ins from celebrities like Mark Cuban, much of the virality of the platform has been built on what now appears like a really murky legal area for AI and digital likenesses.
TOGETHER WITH MBZUAI
🧠 The brand new standard in open-source reasoning

The Rundown: Meet K2 Think: the open-source reasoning system setting a brand new benchmark for AI. With only 32B parameters, it delivers performance on par with flagship models 20X larger and at unprecedented speed — combining efficiency, transparency, and power into one breakthrough model.
With K2 Think, you’ll experience:
-
The highest-ranked model on global math benchmarks
-
Compact, fast performance at 2,000 tokens/second
-
A system ready-to-use on Hugging Face today
Start with K2 Think now.
ANTHROPIC
⚙️ Anthropic brings Claude Code to the browser

Image source: Claude
The Rundown: Anthropic just introduced Claude Code on the net in research preview, allowing developers to code directly from browsers as an alternative of in a terminal — with the agentic assistant also in a position to handle multiple tasks directly within the cloud.
The main points:
-
Users can connect GitHub repositories and assign tasks, with the coding assistant in a position to run several instances directly with real-time progress tracking.
-
Each coding session happens in its own isolated workspace, stopping any single task from accessing files or networks outside its designated boundaries.
-
Pro and Max users can now access Claude Code via browser or iOS app, assigning tasks and getting automatic pull requests when work is complete.
Why it matters: Bringing Claude Code out of command-line tools appears like an enormous convenience move for today’s on-the-go era of labor, providing a streamlined experience for routine tasks. The net interface can also be likely a more accessible way for less experienced users to experiment with the tool in a less intimidating interface.
AI TRAINING
💯 Automate performance reviews and PIPs

The Rundown: On this tutorial, you’ll learn the best way to connect ChatGPT to your Google Drive knowledge base after which use agent mode to routinely transform messy worker notes into structured performance reviews and improvement plans.
Step-by-step:
-
Create a Google Doc called “Worker Review” with rough notes for every worker (name, role, KPIs, feedback)
-
In ChatGPT, click “+” → “Connected apps” and enable Google Drive access
-
Prompt with system instruction + templates: “Analyze my ‘Worker Review’ doc and create performance reviews and 30/60/90 PIPs using these formats: [paste templates with SMART goals structure]”
-
Enable Agent Mode and prompt: “Update the Worker Review doc with the structured sections for every person”
-
Schedule monthly automation: “Run third Saturday at 9 AM, update Worker Review doc from latest notes, regenerate reviews and PIPs”
Pro tip: Keep adding rough notes weekly — the scheduled agent consolidates and formats monthly. If the agent cannot edit initially, grant permissions and re-run.
PRESENTED BY TEXTJAM
✍️ Write faster along with your team and favorite AIs

The Rundown: TextJam is a multi-player AI editor with access to all leading AI models and novel tools that save time. Users can precisely control what AI keeps vs. changes using pen, pencil, and inline prompt tools – avoiding limitless prompting and copy-paste.
Do your finest work faster, with TextJam:
-
Edit with teammates and AI in realtime — creativity loves company
-
Suit your writing to any length, with intuitive gestures
-
Seamlessly mix your words and the AIs, so all of it feels like… you
Start writing faster.
NAPSTER
🪞 Napster pivots to AI with holographic companions

Image source: Midjourney
The Rundown: Former file-sharing and music streaming service Napster just launched Napster 26, pivoting to an AI platform with a $99 holographic display that projects 3D AI assistants above Mac screens without the necessity for glasses.
The main points:
-
The platform offers 15,000+ AI companions like coding specialists or wellness coaches, which appear as 3D holograms through the Napster View display.
-
Users may create AI “digital twins” of themselves that Napster claims can attend meetings, answer questions, and manage an internet presence.
-
The corporate was acquired for $207M this 12 months by 3D technology firm Infinite Reality, which transitioned the music platform to an AI companion product.
-
Napster 26 launches for Mac with subscriptions starting at $19 monthly and the View hardware priced at $99 or coming free with annual plans.
Why it matters: While Napster’s brand may evoke nostalgia for the 90’s kids, it is a pretty wild “rebrand”. AI corporations have been pushing AI that may view user’s actions, and this hologram takes that have very literal — though given xAI’s Grok companion moves, there could also be more of a marketplace for this experience than we expect.
QUICK HITS
🛠️ Trending AI Tools
-
⚙️ Claude Code – Anthropic’s agentic coding assistant, now available on web
-
🤖 Lindy AI CMO – A marketing team powered by AI agents
-
🐠 Fish Audio S1 – Expressive text-to-speech model with voice cloning
-
💨 Claude 4.5 Haiku – Anthropic’s latest small, cost-efficient model
📰 All the pieces else in AI today
Anthropic launched Claude for Life Sciences with latest scientific platform connectors, AI skills for lab protocols, and improved performance on biomedical tasks.
Researchers found that LLMs can develop “brain rot” when exposed to low-quality web content, losing reasoning, memory, and safety skills that stay even after retraining.
Krea open-sourced its realtime-video model, a 14B parameter system allowing users to create, modify, and restyle video outputs via a near-instant stream.
Elon Musk announced that xAI is postponing the launch of Grokipedia to the top of the week, saying more work still must be done to “purge out the propaganda.”
DeepSeek released DeepSeek OCR, a tool that compresses image-based docs 10x while retaining 97% of the data, enabling AI models to handle for much longer files.
Wikipedia revealed that its page views have dropped 8% within the last 12 months, attributing the decrease to AI models scraping content over directly visiting the location.
COMMUNITY
🤝 Community AI workflows
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 Carolyn B. in the US:
“I teach asynchronous courses and I take advantage of AI across all features of my work, from generated subject material content, instructional design of student learning experiences, storyboarding media and interactive assets, styling HTML/CSS to supply elegant inviting course pages, configuring and embedding chatbots to support student engagement, evaluating student submissions, quite a few admin tasks…, and analyzing the effectiveness of the entire above.”
How do you employ AI? Tell us here.
🎓 Highlights: News, Guides & Events
-
Read our last AI newsletter: Karpathy’s AI agent reality check
-
Read our last Tech newsletter: Apple may finally embrace touch
-
Read our last Robotics newsletter: Unitree’s ‘bionic face’ humanoid
-
Today’s AI tool guide: Automate performance reviews and PIPs
-
RSVP to our next workshop @ 4PM EST Friday: Learn context engineering
That is it for today!Before you go we’d like to know what you considered today’s newsletter to assist us improve The Rundown experience for you.
|
|
Login or Subscribe to take part in polls. |
See you soon,
Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown
