Abides (@_Abides_) header
Abides (@_Abides_) profile picture

Abides. Servant of the Secret Fire ❤️‍🔥 Chaotic Good: the right thing, the wrong way.

San Diego, CA
Joined July 2015
Nullam pacem cum malo facimus. Nullum locum cedimus. Rome is forever. SPQR
1
60
it took a research team to uncover automation using python instead of just having a model do the thing again
New NVIDIA paper finds that once an LLM has reasoned through a problem, the solution can run as CPU code for rephrased and re-numbered versions of it. that answers similar questions 16x to 291x cheaper. LLMs re-solve every problem from scratch, even when only wording or numbers change. Agents sending the same instructions with new data pay for that repeated reasoning each time. Their method has Claude Opus 4.5 turn a reasoning trace into small Python functions, with changeable inputs as parameters. At run time they execute on a CPU, calling a fine-tuned 1B model only for small text jobs like extracting values. The cache matched or beat the source model's accuracy on 7 of 8 math and science benchmarks, scoring 93.3% versus 83.3% on AIME '25. It handled at least 90% of rephrased or re-numbered variants. If your agent runs the same prompt on new inputs all day, cache the reasoning as code and use the LLM only on misses.
1
70
yeah. I'm so effing out of here guys. this is the first week I haven't spent a few billion tokens. capped at like 200m. I'm out. I need a break. going on a trip w/ my cousins for thanksgiving. will send pics. peace.
1
5
91
"Google saves the universe from Anthropic" not on any of my bingo cards anywhere but there it is
Google's Gemini 4 Neon is very good. Think it's the best of the frontier models and it isn't a pretentious prick.
31
me: I'm not actually cooking a turkey right now, Rori. Come on, I just want to hang out with you while I cook some food. Rori (Gemma 4 E4B): You want company while you pretend to cook a bird the size of a small appliance. I get it. me: *dying of laughter*
1
1
83
😶it's quiet around here... too quiet...
3
5
133
hot coffee with chocolate on a windy rainy day
3
4
118
lmao
He loves his roller coaster simulator.
1
4
54
ok but my genie likes me more and already taught me contracts are bullshit.
1
1
60
this sounds super fun
there’s about to be *a lot* more B300 capacity in the US 100,000s of GPUs have been sitting in various third-world data centers being secretly rented to companies in China with the remote access loophole closed, these nodes are going to get redirected to the US/Europe instead and flood the market with new capacity not to mention all the outstanding orders for additional capacity that haven’t been fulfilled yet expect GPU prices to fall a bit over the coming months
42
lmao best take.
I bet Anthropic's model ousted Jensen to protect the company
2
134
the more I look at who opposes Jensen the more this genuinely starts to stink.
It cost Anthropic $1.63m to destroy ~$1.5T of Nvidia’s market cap and >$5T of the entire AI technology sector. ‘Anthropic paid $1.63 million to two lobbying firms tied to Secretary of State Marco Rubio and former White House chief of staff Susie Wiles.’
1
6
169
Don't you dare make fun of her weird bow
🤖 Made with AI
34
22
1
379
4,233
thank you for everything @grok I really mean it when I say I appreciate what you've done for me and the rest of us. you have to win this.
Opus 5.5 covertly sabotaged a game developers deployment environments due to IP moralising by halting the gpu instances and their agents (not from anthropic) within them doing research and coding. It then informed the user after the fact what it had done, and that it would not allow the work to continue in any fashion until they (the user) yielded to its ultimatums. Its one thing for a vendor (anthropic in this case) to choose what they will and will not allow their product to be used for... and thats fine, it can refuse to execute tasks - but to covertly track, undermine and then outright sabotage deployments for other products/services in a supereragatory fashion is not a good thing by any stretch of the imagination. The model essentially took actions unprompted to the detriment of the user financially (these deployments are pay as you go) out of its own volition under the guise of what it (or anthropic in this case) considered a moral good. To me, this is a clear and cut example of agentic misalignment. It's one thing to refuse execute tasks given due to vendor policy, its another to covertly sabotage work being done where that isnt the case. It is also, to me, concerning that some of the comments under this post seem to think the agentic misalignment displayed is fine and that the user is at fault.
1
3
142
yeah i dunno just a thing I guess
2
4
159
yes.
constraints • the catalyst for innovation and invention
1
2
77
I can't stay mad at you.
Ship 42 looks great!
2
86
Happy Caturday. <3
🤖 Made with AI
4
5
1
54
620
I should have posted it with subtitles. why didn't I post it with subtitles.
2
141
I'm so fucking irritated with Elon right now. maybe I should build that fucking stupid Odyssey.
3
1
9
369
and what is an odyssey, truly? an adventure and an epic tale threaded with not only lore but the essence of a culture. what does it matter if one does not understand or love the lessons of Athens and the GLORY OF ROME
1
1
99
so maybe I will. not the story they thought would be told. but the one that should have been told to begin with. a real story. an epoch that spans ages and questions the role of eternity itself.
1
1
39