In partnership with
Welcome to OHMO AI. A humanoid robot ran 100 metres in 9.39 seconds in Beijing on Saturday and then crashed into the padding at the end of the track. Elsewhere a model called Ox Alpha turned up on OpenRouter with nobody's name on it, Hugging Face is reportedly exploring a sale, and a Twitch streamer is taking Amazon to court over what got fed into its models.
In today's newsletter:
A humanoid robot clocks 9.39 seconds over 100 metres at Beijing's Robot Olympics
A free reasoning model called Ox Alpha appears on OpenRouter with no lab attached
Hugging Face is reportedly exploring a sale that could value it at $13 billion or more
A Connecticut streamer sues Twitch and Amazon over AI training data
LATEST IN AI
X-HUMANOID
Tiangong Ultra runs 100 metres in 9.39 seconds
Image source: OHMO AI
OHMO AI: Tiangong Ultra, a humanoid robot, clocked 9.39 seconds over 100 metres at Beijing's 'Robot Olympics' on Saturday, beating the 9.58 second men's world record Usain Bolt has held since 2009.
The details:
The run came in a preliminary heat rather than a final.
Its margin over Bolt was close to two-tenths of a second.
It couldn't stop at the line and went into the padded wall.
Why it matters: A robot beating Bolt over 100 metres is not the same as a robot beating Bolt. It ran in a straight line under controlled conditions and it couldn't stop itself at the end. Records that stayed untouchable for humans for decades are turning into annual benchmarks for machines.
TOGETHER WITH HARMONIC SECURITY
Stop making AI decisions in the dark.
Leadership is asking: are we getting value from AI? Which tools are worth the spend? Where are we exposed? Right now, most teams have no idea.
Harmonic Security Usage Explorer changes that.
You get a complete picture of how your organization uses AI, automatically categorized into custom tasks and use cases.
You’ll see the projects being worked on, who’s using what tools, where AI investments are driving value, and where employees are engaging in risky behavior.
CIOs can rationalize spending and cut wasted licenses. CISOs can pinpoint where risk exists and neutralize it. AI committees can show exactly how their efforts are paying off.
OX ALPHA
Nobody knows who built Ox Alpha
Image source: OHMO AI
OHMO AI: Ox Alpha showed up on OpenRouter last Thursday with no lab's name attached. It's a reasoning model aimed at coding and long running agent jobs, it's free to use, and its context window runs to a little over a million tokens. The provider behind it has chosen to remain anonymous.
The details:
OpenCode, an open source agent, said the model would be free for a week with near unlimited usage.
It arrived as a stealth release from a third-party provider, and no lab has put its name to it.
Its context window runs to just over a million tokens.
Why it matters: Handing out a model this size for free is not generosity, it's collection. Every developer pointing production work at Ox Alpha is showing an unknown operator how they build and what they're building. The interesting question isn't which lab is behind it. It's how many people decided that didn't matter before they started sending it their code.
HUGGING FACE
Hugging Face is reportedly weighing a sale
Image source: OHMO AI
OHMO AI: Hugging Face is reportedly sounding out a sale, and the price being discussed could put its value at $13 billion or more. A bank has been brought in to test how interested buyers are, according to people familiar with the process.
The details:
Hugging Face is the platform a lot of developers use to discover, share and build AI models.
Talks are early and no deal has been reached.
No bidder has been named.
Why it matters: Hugging Face sits underneath a lot of AI work that never mentions it. Models get published there and pulled down from there by people at labs and startups who never think about who runs the place. Whoever ends up owning it gets a say over the default home for open models, and that's a strange amount of quiet control to change hands in a process nobody has confirmed.
PRESENTED BY SKYDIVE
Your agents work while you sleep
Give a Skydive agent an ongoing responsibility and they’ll handle it on schedule, every time.
Have them prep your morning report, research new leads, monitor customer feedback, or keep projects moving overnight. You wake up, the work is already done.
AMAZON
A Twitch streamer sues Amazon over AI training data
Image source: OHMO AI
OHMO AI: Warren Pandiscia, a streamer based in Connecticut, filed a class action lawsuit against Twitch and Amazon earlier this week. The complaint says the companies took no licence and no permission before using his streams and videos as a "dataset necessary to fuel Amazon's AI products".
The details:
The suit claims Amazon never obtained consent from Twitch streamers to use them to train its AI models.
It calls the arrangement a breach of contract, with streamers used as "free training stock for separate commercial AI products".
The complaint says streamers "lost money or property as a result of defendants' unlawful, unfair and fraudulent practices".
Amazon isn't the only company facing a lawsuit over allegedly using creators' content for AI training.
Why it matters: This is the part of AI training that most people never agreed to and never saw happen. If you've uploaded anything to a platform in the last decade, the terms you clicked through are the whole of your protection, and they were written before any of this existed. What a court decides about that contract will matter more to ordinary creators than any model launch this year.
AI GUIDES
Get the answer you wanted on the first try
Most bad AI answers are bad questions. Splitting the ask into three parts fixes most of them.
Copy this, swap in the bracketed parts, and paste it in.
Task: [state in one sentence exactly what you want done.]
Context: [paste the background it would otherwise guess at. The document, the numbers, who this is for, what already happened.]
Output: [describe the format, length, tone and audience. For example: a bulleted list for a non technical audience, in the tone of a friendly coworker.]
Think carefully before answering, and give me two versions to choose between.Most prompts carry only the first part. You state the task and leave the model to invent the background and pick a format, so it hedges and hands you something generic. Naming the context stops the guessing. Naming the output stops it choosing for you. Two more things worth doing: break a big request into smaller steps and run them one at a time, and ask for options whenever you want to pick rather than accept.
COMMUNITY
Featured Reader
Every newsletter, we feature how a reader is using AI to work smarter, save time, or make life easier.
Today's reader is Anne S. from Porto, Portugal:
“I run a bakery that supplies about fourteen cafes around the city, so my mornings are baking and my evenings are admin. The thing that was killing me was chasing invoices. Some of these places sit on it for two months and I hate that phone call, so I'd let it slide and then be short at the end of the month. Now I keep a list of who owes what and I get AI to write the chase emails, polite for the ones I like and firmer for the two who always push it. It also reads my supplier invoices and flags when flour or butter has crept up, which I never used to catch until the margin was already gone. Got paid weeks earlier last month than I normally would. Still hate the phone calls. I'm not making them any more.”
Want to be in the next newsletter? Tell us how you use AI here.
AI Meme
Go Deeper
AI Tools: A curated list of the AI tools worth knowing.
Instagram: The AI page they pray you’ll never find.
Facebook: For the ones who never left Facebook.
