Happy Monday! I spent this week keeping up with the latest AI / tech releases and updates so you don't have to. Here are the 4 things I found most interesting.
OpenAI's next model, Astra, just cracked 10 math problems that had gone unsolved for over a decade, for about $2,000 in compute.
OpenAI quietly revealed Astra, a GPT-6 prototype, inside a blog post about math. It solved ten open problems in math and computer science that had each been stuck for 10+ years, including one question left unanswered since 1999. They published a 249-page paper plus fully machine-verified proofs on GitHub, and put the whole compute bill at roughly $2,000.
My take: This feels like something quantum computers would be doing. All I care about is that they're prototyping GPT-6 and working on stuff. The $2,000 bill is impressive though, that's basically the inference you'd get on a $20 plan for real.
OpenAI wants to be your login for the whole internet with "Sign in with ChatGPT."
You know that "Sign in with Google" button everywhere? OpenAI just launched its own. It's rolling out in beta with partners like Notion, Vercel, Supabase and Airtable, letting you log into apps with your ChatGPT account (it only shares your name, email and picture). With hundreds of millions of users, OpenAI's quietly trying to sit at the center of how you access everything.
My take: I actually like this one. Maybe a bit scary, but since ChatGPT already knows every single thing about you, it makes sense they'd want to use it as auth on every platform. If it was my company I'd be doing the same. Still scary data-wise though.
Zuckerberg published an essay arguing superintelligence should belong to everyone, not a handful of labs.
On July 28, Zuck put out a piece pitching Meta's "personal superintelligence" vision, the idea that AI should empower individuals directly instead of being locked inside a few big companies. It's a not-so-subtle shot at rivals, since he called concentrating superintelligence in a handful of labs the biggest risk of all, right as OpenAI, Google and Anthropic lean the other way.
My take: Honestly he's right, spreading this tech out is the move. Just hits a little different coming from someone who isn't really leading the race rn. But hey, more people pushing for openness is always a win.
Over 1,100 employees from OpenAI, Anthropic, Google and Meta signed an open letter asking Washington for a way to hit the brakes on AI.
On July 28, insiders from the biggest labs urged the government to build an international "pacing mechanism," a tool that could force a verifiable slowdown if AI starts outpacing our ability to safely oversee it. Coming from the people actually building this stuff, and landing the same day as Zuck's "spread it everywhere" op-ed, the split inside the industry is hard to ignore.
My take: I don't think anyone actually does this until we see some destructive behavior that hits the economy. Otherwise governments don't see the point in regulating it as long as money's being made.
Those were the biggest AI updates on my radar this week.
On My Mind
I believe consistency shines especially in the bad times. Anyone can be consistent when things are booming, when you're getting feedback, earning money, whatever. But staying consistent during a downturn, or when things are hard in your life, that's the real metric.
It matters most that you do your work when you're feeling anxious or tired. Those days become your anchor. You'll be able to look back and see that you held it down in the bad times, and that things come and go.
That's my approach anyway. I recorded my last sponsored posts while I was tired, and yeah, maybe I didn't get the best results, but now I can look back and see that I still did my work even when I wasn't feeling my best.
Consistency comes in waves. Sometimes you get way better results for the same amount of work, and sometimes you get nothing even when you put in a ton. Over time it smooths out, and that's why you need to be patient and consistent.
More from me!
My latest video: I gave out 3 free hand-picked YouTube videos on Machine Learning that I think every beginner should watch! (Watch here)
Tool of the week (clicking the link supports me a ton!):
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
Have a great week, do more!
Volkan