In partnership with
Weird week.
AI cracked a math problem that we (humans) couldn’t for almost 200 years. The same day, one of the people building it quit and told the world it might kill us all.
Honestly, that's the whole story of AI right now. Lot more has happened, hence a roundup today I could bring your attention to.
Before we dive in:
When you saw "AI might kill us all" trending this week, you felt:
Use it this week
3 things you can open and use today.
OpenAI shipped a big upgrade to image gen inside ChatGPT.
Sharper images, much better at editing a photo without wrecking everything else, and a new Sketch feature where you scribble a rough layout and it builds the image from your doodle.
2. Meta Muse
Meta launched its first personal AI agent. that goes and does things for you, like books tickets, fills forms, orders stuff, sets appointments.
It runs its own little browser in the background so it can click around the web on your behalf. There's a free tier plus $20 and $100 plans. US only for now.
My take: great direction, but do not connect your payment methods or main inbox yet.
Meta's own engineers reportedly raised flags about it mishandling sensitive data, and this is week 1 of a brand new product. Make it earn trust on the small stuff first.
The Chinese lab DeepSeek dropped a new model that's faster and cheaper than the one it replaced, with image understanding built in and prices cut across the board.
My take: DeepSeek keeps dragging the cost of "good enough" AI toward zero, which is fantastic for your wallet.
One caveat: it's hosted in China under Chinese law, so keep anything client-confidential or sensitive off it.
The big story
OpenAI says one of its AI models solved the Navier-Stokes problem.
Before this, only 1/7 had ever been cracked. OpenAI pointed 10,000 AI agents at it for 88 hours and said they found the answer, and they're not even claiming the prize money.
Here's why you should care though.
For 2 years the standard knock on AI was "it just remixes stuff humans already wrote”, but a genuine new proof would be the first sign that AI can do original expert work.
That shift is what’s rattled Tech Twitter and the “Will AI take away my job” crowd.
Although it's disputed.
Serious mathematicians including Terence Tao have said pump the brakes, the wider maths community hasn't verified it yet, and there's an ugly fight brewing over whether OpenAI got a head start from other researchers' unpublished work.
So file this as a "huge if true.” and relax.
But if you want to read more:
The other big story
They laid out three futures:
Mild: AI is a nice productivity bump, the economy barely moves, life stays mostly normal.
Middle: solid growth, but knowledge-work jobs start shrinking (they estimate roughly 4% fewer).
Wild: the economy grows at a pace we've basically never seen. but white-collar work takes a hit, and the share of national income going to workers instead of company owners falls.
The scenario that creates the most money is also the worst one for people who work for a living.
My take: don't panic, yet.
The only takeaway imo is the direction. The safe place to be over the next few years is still the person telling AI what to do. That's the whole game.
Figure out which side of that line your work sits on, and if it's the wrong side, start moving now. (yes, that's literally what we teach at Outskill. the point holds even if you go learn it somewhere else.)
The rest
Stuff that’s happened this week but doesn't need a full essay:
Apple announced a feature that cryptographically proves a photo is real and not AI-generated. As fake images get flawless, proving something is genuine becomes the product. Smart bet.
Anthropic put out a report on how criminals and state hackers are already trying to misuse its AI, and how it caught them. Good reminder that this arms race is live.
The Seattle Times and Newsday sued OpenAI and Microsoft for using their journalism without paying. The "who owns the training data" fight keeps snowballing.
OpenAI added a well-known safety researcher to its board. Small move, mad timing given everything else this week.
XPeng, a Chinese car company, started building humanoid robots on a production line. The robots-plus-AI thing is officially leaving the demo stage.
The thing to ignore
The "AI is going to kill us all" panic that took over your feed this week.
That's a serious, long-term conversation but what headlines chopped off is that the same researcher said the risk from today's AI is low.
The worry is about a future superintelligence that doesn't exist yet. so the chatbot on your phone right now isn't plotting anything.
Hold both thoughts at once.
Take the long-term question seriously, ignore the short-term panic-farming.
Before you go
AI is now good enough to do things that felt impossible a year ago, and the people closest to it are the ones most nervous about where it's headed.
This is just what being at the frontier of tech feels like.
Don’t pick a team between hype and doom. Instead, stay close enough to use this stuff, and skeptical enough that neither camp can play you.
That's the entire reason this newsletter exists.
Go try the image tool, skip the panic.
See you next week.
Vaibhav 🤝
If you read till here, you might find this interesting
#AD 1
Get more meetings by removing the LinkedIn limits.
When you need more leads, you want to do more, right?
Adding mailboxes to cold emails.
Placing more cold calls
Running more ads.
But what do you do if your leads are on LinkedIn? LinkedIn outreach still depends on the limited number of profiles already on your team.
Aimfox Avatars lets you rent dedicated, customizable LinkedIn profiles that start conversations and hand interested replies to your existing sellers.
Add outreach capacity without hiring another SDR or turning every employee account into a prospecting channel..
#AD 2
Porkbun is the domain registrar trusted by folks who want low prices without the nonsense. Get most domains at cost, free features like WHOIS privacy & SSL certificates, plus real 24/7 human support. Get $1 Off Your Domain
