← Feed

Posts

• Post #1725155

  @CrabbyIT I am not normally a fan of developing in virtual environments but this strikes me as a potentially valid use case. Discourage running LLM tools directly on developers' real hardware; instead, have a template VM that you can spin up, with all the tools pre-installed (make it easy and low-friction!), and then sandbox those virtual workstations at L2 (VLAN) / L3 (subnet) on the network. I am not sure if I would trust purpose-built LLM sandboxing tools currently, and LiteLLM in...

• Post #1725153

  Dominos Pizza apparently DDOSed their own website to death with their NFL Draft promotion. How do you think they apportion the blame between the Marketing department and the IT department? #fail

• Post #1725151

  @hannah and people say journalism is dead

• Post #1725149

  Welp, so much for #Github #Copilot Pro — it was a pretty good deal at $100/year, and the billing based on prompts made more sense to me than billing based on "tokens". Just to ensure anyone who wants to ride out their annual plan gets thoroughly screwed, they're doing away with the "0x" (free, mostly older) models and increasing the multipliers on the remaining ones. Nice touch. Quite the #enshittification speedrun. Bravo, guys.

• Post #958350

  I've read a couple of breathless articles about Donald #Knuth's use of #LLM systems, and found (unsurprisingly) that he has a paper about it, if you prefer your news without someone else's hot takes: https://www-cs-faculty.stanford.edu/~knuth/papers/claude-cycles.pdf (Sidebar: apparently he's working on *another* volume of #TAOCP? Good lord. I still haven't worked through the exercises in Vol. 1.) #AI #Programming #ComputerScience