The 512GB step on the new M5 Ultra Mac Studio costs $15,500 on its own. TechRadar worked out that's about 3.9 times more per gigabyte than the same step on the M3 Ultra last year.
My M3 Ultra with 512GB just got a lot harder to replace. Holding on to it a bit tighter now....
Reboot your Mac every 50 days or the network dies. Ok Apple. Thats fine, totally normal for a 2026 computer 🤷♂️
Did you know that running your Mac for more than 50 days without a reboot will fuck with your internet connection?
Yes, really. There is a 32-bit integer timing your TCP connections and if you let it overflow MacOS will run into all kinds of fucky issues with its TCP stack.
For example I was unable to upload things bigger than 4MB. I was wondering why my Claude sessions started to hang seemingly at random.
The fix? Literally just reboot your machine. That's it.
What the fuck
photon.codes/blog/we-found-a…
Readers added context they thought people might want to know
The TCP issues after ~50 days uptime were a kernel bug only in early macOS 26 Tahoe versions, fixed by Apple starting in 26.4; current systems are unaffected.
nevyn.dev/blog/20261030-…
photon.codes/blog/we-found-…
portal.greenmini.host/help/macos-tah…
1
47
New model out. This will be fun. GPT-6.1 Astra has been my main frontier model for a long time. Upgrading to 6.2 now.
Some of you have been asking very politely. Some of you have just been asking.
GPT-6.2 Astra and Sol are now available.
Super efficient and built to perform where it counts—the real world.
openai.com/index/introducing…
40
The EU is training its own open model at almost a trillion parameters, on a supercomputer we paid for. Did not see that coming from Brussels 👏
No weights yet and the big training run hasn't even started. Ship it before the 400 page compliance guide!
‼️ The EU's open AI model, EUROPA, is being tested at nearly 1 trillion parameters, more than twice the size the EU asked for.
Domyn, the Milan company the European Commission picked in June to build it, ran the trials on Germany's JUPITER, Europe's first exascale supercomputer.
The test model has 951.9B parameters against a 400B minimum. It splits the work across 384 expert networks, so only about 32.8B run for each token.
Training got about 14 times faster, from 90 to 1,293 tokens per second per GPU.
There are no weights yet. The licence, training data and release date are unknown.
38
Someone under my Mac Studio post is still kicking himself for only going 256GB on the M3. I get it. Whatever amount of RAM feels like too much when you buy it, the models you actually want to run show up a few months later and need more.
So my advice is still the same. Max RAM. But I know what that costs right now 😱
3
91
On the same day I have the most popular post I’ve ever had on X, the exact same thing happens on LinkedIn for a different post there. What are the f’ing odds 😂 Both were about local AI which seems to be all I post about lately.
I’ve been replying to comments across the two platforms half of my day 😍
Thank you for following me here! … if you don’t, and want to hear rants about AI, Omarchy or nerdy tech stuff, you know what to click 🙏
1
104
They were honest about using AI and got "100% AI generated garbage" on their Steam page for it. So the next dev just won't say anything, and the same people will happily buy that game without a clue.
I openly disclosed using AI to help create my game, and someone still came to my game’s Steam page to comment “100% AI generated garbage.” You knew AI was involved because I told you. What’s the point, punishing me for being honest? Then people wonder why devs stay silent lol.
84
















