In partnership with
TL;DR
OpenAI's Jalapeño chip beats Nvidia on speed and power efficiency
Bill Gates says AI has already crossed several danger thresholds, and slowing down isn't politically possible
Gemini 3.5 Transcribe cleans up your speech before it hits the page
Meta pays $17bn and rolls out new teen safety limits nationwide
1 Million Pixels lets you claim internet real estate for your brand, forever
Claude Design solves its consistency problem by turning your best output into a reusable skill
👑 Too busy building your empire to keep up with can't-miss business news? That's where The CEO Report comes in, keeping you up to date with can't-miss (tech-leaning) business news that will help you be a better Founder & CEO, at the end of every day! Sign up for free here →
BILL GATES SAYS THE DANGEROUS PART ALREADY HAPPENED
Bill Gates just published a 6,000-word essay, and the headline isn't "AI is coming." It's "the thresholds are already crossed." Bio risk, cyber risk, psychosocial harm, job destruction, loss of control. He's not warning about a future tipping point. He's saying we're past several of them right now.
This is coming from the guy who spent the 1980s getting accused of the exact same things Silicon Valley gets accused of today: killing jobs, moving too fast, not caring what it breaks. Gates isn't an outside critic here. He built one of these companies. And he's telling you the default path from here has a "high chance of a net negative."
His proposed fixes are a mix of pragmatic and wild. "Human-reserved" jobs, basically nature reserves but for work humans still do on purpose. A payroll tax on robots and AI systems doing human jobs. A global coordination body, because he's convinced no single country slows down on its own. He also says privately, lab leaders sound a lot more nervous than their fundraising decks let on. Publicly optimistic, privately not so sure.
OUR TAKE
Gates isn't a safety researcher with something to prove. He's a billionaire who profits either way this goes, and he's still saying slow down.
That should carry more weight than another manifesto from a CEO with a product to sell.
The real tell is the trillion dollar line. Everyone downplaying risk right now has a fundraising round riding on the optimism.
Watch this show up in enterprise contracts soon, in the fine print about who eats the liability when things go wrong.
QUICK HITS
Meta just paid $17 billion to avoid finding out what a jury thought of its internal Instagram data: The company settled with California and 28 other states days into a trial that included testimony from a former safety engineer who said Meta ran on a "don't ask, don't tell" policy about harm to kids. Meta now has to roll out daily usage limits, nighttime blocks, and a ban on plastic surgery filters nationwide, all while still denying any wrongdoing.
Google's new transcription model fixes the thing that makes voice typing insufferable: Gemini 3.5 Transcribe cleans up the "let's meet Tuesday, no wait, Wednesday" moments in real time, cutting filler words and self-corrections before they ever hit the page. It's live now in the Gemini API, Gboard's new Rambler feature, and the macOS Gemini app, with Chrome support coming soon.
OpenAI's first in-house chip is named after a pepper, and the heat is not just marketing: The company says its new chip, Jalapeño, beats Nvidia's GB200 and GB300 systems on both speed and power efficiency, pulling 1.5 to 1.9x more AI work per watt and cutting latency by up to 3.6x. OpenAI even used its own models, including Astra and Codex, to help design and program the chip in under a year.
Buy On Blue, Sell On Red..
Buy on Blue. Sell on Red. This 2-time award-winning hedge fund algorithm from Mark Helweg captured a 6,566.5-point Nasdaq move, worth over $131,000 per futures contract. Now a fresh Blue Dot BUY signal is live on a little-known ticker. Get your FREE live entry directly on the chart today.
YOUR BEST CLAUDE DESIGN WAS A FLUKE
Building one gorgeous design in Claude is easy… but building the same quality of design ten times in a row is the actual problem.
⏱️ 30-45 min | 🔧 Claude, a skill file, an image/video generation tool of your choice
Why build this? Most people run a design loop, get something they love, then hit a wall the second time. The output looks close but never quite right. That is because the win never gets written down. Nobody stops to ask why the good version worked. So every new attempt starts from zero instead of building on what actually worked last time.
The fix is treating your best design like a spec, not a one-off. Once you break it down into rules Claude can follow, you are not hoping for a good result anymore. You are just running the system again.
Steps:
Find a design you actually love. Doesn't matter if it's a poster, a website, or a short motion graphic. Save it or screenshot it.
Use a design loop to build your own version. This runs multiple critique passes in the background so the output gets checked and refined instead of accepted on the first try.
Once you have a version you're happy with, have Claude reverse engineer it. Not describe it, actually measure it: spacing, ratios, color rules, what's missing, what breaks it.
Turn that breakdown into a skill file. This is the step everyone skips and it's the whole point.
Reuse the skill any time you need a new version. Same rules, new input, consistent output.
Expected outcome: You end up with a repeatable system instead of a lucky result. The next design you make in that style takes a fraction of the time and doesn't need ten more rounds of "make it better." You built the taste once. Now Claude just applies it.
Full tutorial:
TOOL OF THE DAY
TBH, our founder Art built this one. And he built it completely on his iPhone, using Replit, which is honestly the more interesting part of the story.
The idea is simple. There are 1,000,000 pixels up for grabs. Claim a chunk, put your brand, your joke, or your face on it, and it's there for good. Think internet billboard, except you actually own your square of it.
Right now it's early, which means the good real estate is still cheap. Once this gets discovered, prices go up and the map fills in fast.
Get your pixels while there's still room to be seen!
