aaron retweeted
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
Community note
This post copies verbatim a statement by AI researcher Jacob Coxon, who announced his resignation from Anthropic in September 2026 after conducting pretraining research at both OpenAI and Anthropic. businessinsider.com/anthropic-rese…
166
230
1,134
364,146
guy selling shovels says the gold mines are perfectly safe despite what the mine owners say, colour me surprised
38
until you’ve given codex permission to freely setup cloud resources for a project and told it to setup billing with the card details in your password manager, you haven’t lived
1
55
so 6 sol and Luna drop, similar performance but much cheaper. it’s nice, but a bit boring compared to opus 5.5 - I’m wondering if there’s a bigger fish in store for tomorrow/thurs… 6.1 Astra?
2
3
128
I thought cheap m4 Mac minis were supposed to flood the market now the new M6 is out as people upgrade. where are they tho?
1
133
I think this is my longest running single goal to date? mostly 5.6 sol implementing an epic originally drafted by astra high. im hoping we'll get it wrapped up over the weekend. pretty happy with the results
143
when do we begin going back to analogue controls for critical infrastructure like power grid, water supply, street lights etc? Like nuclear weapons… rogue AI risk aside, with cybersec being how it is, surely at some point it’s justified for truly robust systems no?
1
69
the first two times this came up in my feed this morning I assumed it was a joke or something. guess gpt is a tree hugger?
2
36
I remember this video auto playing in the background, it was half way through before I finally gave it my full attention and wondered what on earth was going on… crazy times how far it’s come already. easy to lose perspective
Can AI Create a Minecraft Hack?
I tried using #ChatGPT by #OpenAI. Here is what happened:
piped.video/watch?v=ukKfAV4A…
2
92
I didn’t know it was possible, but iOS 27 somehow managed to make Siri even less useful?? now it comments on my current screen when I’m trying to place a screen phone call, very nice
1
54
Running malecns:v1.0 end-to-end now. 166,700 reconstructed neurons, full brain + optic lobes + VNC, sensory stimulation in and motor populations out. Current benchmark is needle manipulation / suturing on a grape. Fairly happy w/ early results
57
Can I slow #codex down? It’s producing prs faster than my slooooow CI can handle, so agents get stuck idling/polling. Can I add a 20s sleep between every tool use for example?
42
In the last 72 hours all the big lab spokespersons have suddenly re expressed their concerns about speed of progress. what spooked them? No doubt we’ll find out several months after the fact
43