apenwarr
@apenwarr
wvdial, bup, sshuttle, netselect, popularity-contest, redo, gfblip, GFiber, atlogin.net, and now Top Chicken @Tailscale.com doing WireGuard mesh. First search result for "epic treatise."
Deepseek seems to like the (not released yet) thundersnap MCP server so we have that going for us, which is nice.
Yes. And also we’re about to see a whole crapload of new cool things in the first category
I had no idea but this makes a lot more sense than this *not* being a simulation
I’m a little disappointed at the weakness of this argument. It is entirely possible for guardrails to have both false positives and false negatives.
Meanwhile, a particularly wtf section made me double check whether my ability to read legal docs is correct. Answer: no worse than Gemini’s anyway
Protip: do not wake up in the middle of the night and ask Gemini about your dream
“Has anyone involved in one piece ever experienced actual snow”? Apparently yes
Okay this is fun. I tried translating the same sentence to “apenwarr” and it was unchanged; uh, I don’t know what I expected.
AI-driven sales outreach continues to be absolute garbage. Look, there was a 95%+ chance I was going to mark it as spam anyway. But if you don't even say wtf your company does, it's 100%. What's the point of brevity with no content?
Please don't make me search for whatever you're talking about :) Anyway from first principles, gemini thinks it's not a great idea.
Dunno how accurate this is, but it says we’ve invested far less in AI infra, even with the next 3 years included, than cellular infra. How long until everyone is willing to pay as much per month for AI as for their phone subscription?
A modest proposal: measure the median family income in gigaflops (warning, logarithmic y axis)