Happy Sunday. Big week, for the first time a major lab is openly using the word AGI. Here's what actually mattered in five minutes, then a bit on what's on my mind. Let's get into it.
This week in tech + AI
OpenAI released GPT-6 Astra and said "welcome to the AGI era." Astra is their biggest model yet, trained on more than 100,000 GPUs, and on paper it's genuinely wild. It scores 98.6% on the ARC-AGI3 benchmark, hit 100% on an exploit test, and found two real zero-day security bugs in Chrome's engine while they were testing it. In the demo it built a 3D game, formatted a legal contract and booked a tennis court all at once. Whether this is actually AGI is very much up for debate (a benchmark score is not a mind at the end of the day), and "welcome to the AGI era" is a marketing line as much as a technical one, but the capability jump is real. The timing says a lot too: two days earlier Anthropic shipped Claude Fable 5.1, its best coding model yet and roughly 25 to 45% cheaper to run, so the frontier labs are now trading blows basically every week. Read more
Tesla put its steering-wheel-free Cybercab on public roads in Austin and caught a federal investigation within a day. No steering wheel, no pedals, around a thousand of them, and regulators opened a probe into how Tesla self-certified that they meet safety standards without the basics like, you know, a steering wheel. Future of transport or regulatory headache, depends on the day. Read more
Saudi Arabia launched its national Arabic AI model, HUMAIN M3, built on top of a Chinese open-weights model. It's a 428-billion-parameter model trained on over a trillion Arabic tokens, and the interesting part isn't the size, it's the choice: a US-aligned Gulf state built its sovereign model on China's open stack instead of an American one. A small signal of where AI power and dependencies are actually heading. Read more
On my mind
Is the race to AGI something good? When talking to friends about the development of AI and safety concerns there is one comparison that is almost always mentioned. The race to AGI is the new Manhattan project (the project to build the atomic bomb, if you haven’t heard of this, close this Newsletter now and look it up). And I find that a quite fitting comparison because back in the day people were aware of the dangers that the atomic bomb or nuclear energy would bring with it, but people also knew the potential life changing positive effects of it. Besides that people knew that SOMEONE was gonna build it anyways so the only way to control it was to own it. And I feel it's pretty similar with the AI labs at the moment and governments becoming more and more involved in AI development. I am by no means an expert in this, but my opinion is that AGI definitely comes with a handful of positive impacts for humanity while also having a lot of safety concerns that need to be handled beforehand. I agree with movements towards a centralized institution for AI safety and alignment across nations and hope that smart people will bring something like this to life.
This topic is and has been relevant for a longer time, the release of GPT 6 Astra just sparked a lot of AGI conversations again so I thought this would fit here haha.
More from me
🎬 My latest content:
Master Application tips that got me into Oxford, ETH and UCL, coming from an economics degree. Have a look
I tried a fun new format: the machine learning concepts you could've learned in the time you spent scrolling on your phone. Have a look
👥 Community Section:
I want to make the Newsletter more engaging and give you guys a chance to actively shape it, so I want to introduce the new community section! Here I will reserve 2-3 bullet points for everything YOU might want to share with the reader. I think this could be a great place for book recommendations, resource recommendations, Job Opportunities, Cofounder listings etc. As you know this newsletter now has over 5,000 readers with cracked people from tech companies, Investors and more! If you have something you want to share, please send me an email to [email protected] with the Subject line “Newsletter Community Section Idea” and I will have a look and might feature it :)
🔨 Tool of the week:
Hubspots free ChatGPT Guide
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
Have a great week,
Chris