Good morning {{first_name| Artificial intelligence lovers}}. Three weeks of high-profile testimony, sensational leaked emails, and $100 billion worth of claims have been made for a massive experiment – but Elon Musk’s conclusion against OpenAI hasn’t lived up to the hype.
With jurors rejecting the case in favor of OpenAI, and Musk calling the ruling a “technical orthodox” and vowing to appeal, the most dramatic outcome of the massive tech trial may have to wait for another day in court.
-
Elon Musk loses lawsuit against OpenAI and Microsoft
-
Cursor’s Composer 2.5 approaches the limits of programming
-
3D model anything using Claude and Blender
-
Universal multimedia and multiplayer models of Odyssey
-
4 new tools for AI, community workflow, and more
Elon Musk vs. Obinay
Image source: Images 2.0 / The Rundown / @ElonMusk on
Rundown: Elon Musk’s $100B+ lawsuit against OpenAI, Sam Altman, Greg Brockman, and Microsoft was just a $100B lawsuit. It was rejected After a high-profile three-week trial, the jury unanimously found that the case had been brought years too late.
-
The lawsuit claimed that Altman and Brockman “ripped off a charity” by turning OAI into a for-profit organization, but the jury felt that Musk had known about this for years before filing the lawsuit.
-
OAI’s lawyers argued that Musk supported a for-profit structure himself, pushed for control, and only filed a lawsuit after he launched his own AI competitor xAI in 2023.
-
Musk’s claim from Microsoft was also dismissed, after he accused the company of helping OpenAI by supporting it with billions of dollars.
-
Musk to publish X argued that the jury’s verdict was not based on “the merits of the case, but merely on the technical evaluation” and that he would appeal the decision.
Why it matters: The massive trial, filled with private transcripts and testimony from the billionaire, ended not with much fanfare, but with a hasty dismissal due to the legal clock. It’s a big win for OAI, but an unsatisfying one for anyone hoping this case will resolve the question of who controls the nonprofit AI giant once billions of dollars enter the picture.
Together with YOU.COM
Rundown: Choosing an API by scanning a benchmark table and calling it “done” is a shortcut that can obscure what actually matters in production – like accuracy. This guide is from Ant.com It explains why raw response time is a deceptive signal and why accuracy, along with other real-world metrics, is what you should be measuring instead.
-
Why does p50 latency mask the failures users are actually experiencing
-
The “time to useful result” framework captures what the standards ignore
-
Four hidden cost drivers that show up in your records, not vendors’ spreadsheets
-
How to evaluate APIs on your actual concurrency levels, not on experimental conditions
Indicator
Rundown: Cursor only Released Composer 2.5, the company’s internally developed coding model built on Moonshot’s Kimi K2.5, shows near-boundary benchmark performance at much lower nominal prices.
-
Composer 2.5 approaches Anthropic’s 4.7 and OAI’s GPT 5.5 across the highest development benchmarks, while showing an approximately 10% increase over its predecessor.
-
The average CursorBench task costs less than $1 with Composer 2.5, compared to up to $11 per task in Opus 4.7 and GPT-5.5 at similar score levels.
-
Composer 2.5 He was “Partially trained on Colossus 2,” also with the cursor Revealing It is currently training a larger SpaceXAI model with 10 times larger computations.
Why it matters: Months ago, Composer 2 was teetering on the border at 1/10 the cost of GPT-5.4. Composer 2.5 picks up on this theme, this time earning Opus category scores of 4.7 while remaining under $1 per assignment. With xAI’s Colossus computing power fully behind the benchmark, its next model could be one that takes on the frontier.
Artificial intelligence training
Rundown: In this guide, you’ll learn how to connect Blender to Claude Code using MCP, and then use plain English to create and edit a 3D scene.
-
Download and install Mixer And so is her MCP extension. Once that’s done, open Edit > Preferences in Blender, find MCP, and enable the extension
-
In the terminal, open Claude Code in your project folder and connect Blender using: “clude mcp add Blender — uvx Blender-mcp-server claude mcp get Blender”
-
Ask Claude to verify the setup with: “Connect Claude Code to Blender and make sure the mcp.json or Claude MCP configuration is set up correctly. Verify that Blender is running the MCP server, confirm the host and port, register Blender, and verify the connection before editing.”
-
Test with: “Use Blender MCP to design my name in 3D. Add disco balls, lighting, reflective materials, and camera angle to make it like an event poster”
Pro Tip: Once connected, download forms from sites such as BlenderKitopen it in Blender, and have Claude arrange the objects, adjust the lighting, and set up the rendering.
Introduction by Lightfield
Rundown: Lightfield is an AI CRM software with agents looking out for you. They find companies that are similar to your fastest-growing customers and engage with them with messaging proven effective through previously won deals.
-
Recording of accounts on suitability, signals and mutual communications
-
Run email and LinkedIn sequences in your customers’ words
-
Update your goal list based on what’s working
Odyssey
Rundown: Odyssey only marked Two back-to-back world debuts, with Starchild-1 being dropped, which the company calls the world’s first real-time multimedia model, and Agora-1which allows multiple players to interact within the same AI-generated world.
-
Starchild-1 can quickly generate synchronized audio and video while absorbing and adapting to user input, with no fixed generation length.
-
Agora-1 can host up to 4 players in a single AI-generated global broadcast, displayed via a GoldenEye video game simulation where every pixel is produced live.
-
Agora maintains the state of the game shared between participants, tracking agent details such as location and health as actions change the world.
-
The company is framing Agora as an early preview of multiplayer games, bots and agents training together inside simulations.
Why it matters: Many of the smartest minds in technology believe that universal models are the future of the industry, and these previews look like a big step up the capabilities ladder. Moving from clips to live streaming, editable co-streaming opens up entirely new horizons for both creativity (gaming and storytelling) and simulation (robotics and AI training).
-
🐳 Mobi 2 – Brands trust AI-driven e-commerce players to run meta ads, create Klaviyo campaigns, and create creatives that actually work*
-
⚙️ Composer 2.5 – An efficient, close-to-pointer internal encoding model
-
🎆 Kriya 2 – Krea’s first indoor photo model, now generally available
-
🧑💻 Devin – New long-term memory auto-encoding security feature
Anthropic acquired Stainless, the startup behind the official SDK and MCP server tools, is adding the team previously responsible for Claude’s developer libraries.
OpenAI a partner with Malta to offer free ChatGPT Plus to every citizen who completes a national AI literacy course, the first deal of its kind nationwide.
Amazon Rolled Alexa Podcasts, a new NotebookLM-style custom podcast builder in Alexa+ that creates a conversation between two AI-powered hosts about any topic.
dead He is Demobilization It cut up to 8,000 employees this week and no longer plans to hire another 6,000 open positions, as part of boosting the company’s artificial intelligence efficiency.
OpenAI Announce New partnership with Dell to run Codex within the company’s data centers, connecting the Codex agent to the organization’s internal systems.
In each newsletter we show how the reader is using AI to work smarter, save time or make life easier.
Today’s workflow comes from the reader Michael D. In Littleton, Colorado:
“I teach mindfulness and meditation. I’m not a professor – just someone halfway through a long road with several years of serious practice and a pile of small wins, missteps, and rehash that tend to be associated with new students.
Over those years, my notes piled up everywhere – Gmail, OneNote, Word, Paper, the Notes app – and none of them were organized, and most of them were “somewhere else” when I really wanted them to be. A few weeks ago, I started moving them all over to Obsidian as markdown files, tagging each with metadata: title, date, topics, teaching capabilities, and free-form notes.
Now that Claude has been directed to my obsidian vault in Claude’s Cowork, he can pull up my own musings, stories I’ve already lived, and quotes I’ve already memorized. Students receive teaching based on real practice rather than general synthesis.
The unexpected bonus is for my own practice. Sometimes Claude uses my notes in ways I wouldn’t have thought of, and I end up with a new perspective on my own experience that I didn’t even plan for.
How do you use artificial intelligence? Tell us here.
That’s all for today!
Before you go, we’d love to know what you thought of today’s newsletter to help us improve The Rundown experience for you.
Rowan, Joey, Zack, Shubham and Jennifer – the humans behind The Rundown
