Happy Sunday. Big week, so let's get straight into what actually mattered in tech, then a cool announcement from my side!
This week in tech + AI
Anthropic officially filed to go public. Anthropic submitted its S-1 to the SEC and plans to list on the Nasdaq under the ticker "ANT". Share count and price range aren't set yet. The timing is interesting: the market is nervous about AI spending right now, with Alphabet having its worst day ever this week (down 12%, around $500 billion gone) and reports that Anthropic might not get an investment-grade credit rating after the IPO. What it means: this is the first real test of whether public investors will pay frontier-lab prices for an AI company, and how it goes will shape every AI IPO after it. Read more
An OpenAI model claims to have disproved the Exponential Time Hypothesis. ETH is a conjecture about how hard Boolean satisfiability is, and hundreds of computer science papers are built on it being true. OpenAI says its model found the disproof, checked it in the proof language Lean and published everything. Two of the researchers who originally proposed ETH are now leading an independent review, and they're clear that this isn't a confirmation yet. If it holds, a whole chapter of theoretical CS gets rewritten by a model. Read more
Binance lost around $6.3 billion in the biggest crypto hack ever. Attackers drained wallets across ten blockchains, most of it in about 20 minutes, roughly four times the size of the Bybit hack last year. According to Binance the entry point was one compromised engineer's laptop, and blockchain analytics firms link it to North Korea's TraderTraitor group. Binance paused trading, says it will cover all losses, and Bitcoin dropped about 5% to under $80k. The scariest part isn't the number, it's that a single laptop was enough. Read more
Google released Gemini 4 Carbon, its new top model built for cybersecurity. It ranks #1 on the Artificial Analysis Intelligence Index and sets new records on benchmarks for finding, exploiting and patching vulnerabilities, just two weeks after Gemini 4 Argon. For now it's only available to vetted cyber defenders, with a wider rollout coming later at $2 per million input tokens. Same week as the biggest hack ever, the strongest security model drops. Read more
On my mind
A lot of the great things in my life came from risky (and slightly stupid) decisions. So here I am on an island building out an idea me and my friend had for months all while I should be in my lectures at ETH. I could've waited till the semester was over, but to be honest there will never be a perfect time to start. And I'd rather try it out now than keep wondering if this would've worked. But enough of that, what are we actually building?
A new type of hackathon focused around quantitative finance in the age of AI. Mirkovicdev and I love competing and we've competed in lots of quant finance competitions but none of them have caught up to the age of AI. People get judged on whatever Claude gives them as output, and a lot of them don't even read their own code submission anymore. We want to change that with challenges and evaluations that are AI proof (or at least they can’t ace it). First competition is planned for Q1 and we're working hard to get the coolest partner companies on board for you. If that sounds like a competition for you here is the special waitlist link for my newsletter readers. Sign up here!
Hope to see y'all in there.
More from me
🎬 My latest content:
TechMaxxing Ep. 2: how I got into a technical master's with a business background, the question I got asked the most. Have a look.
TechMaxxing Ep. 3: How to set up a good context system for your AI. Have a look.
👥 Community Section:
Got a book, a resource, a job opening or are you looking for a co-founder? Send it to [email protected] with the subject "Newsletter Community Section Idea" and it might end up right here.
Have a great week,
Chris