Evï Anton

Serious thoughts.

[Evï Anton @EviAntonWorld]

Anatoly Fomenko's artwork is my benchmark for AGI for AI imagegen. Nothing can replicate its immensity.

[Evï Anton @EviAntonWorld]

nothing ever happens

[Ben Moll @ben_moll]

Two greats join our bet:
- @dwarkesh_sp on the explosive-growth side
- @Thom_Wolf on my side

https://t.co/NSQZKbfBIy

Total pot now $700k 💰

[Evï Anton @EviAntonWorld]

@DVsefbbfd @_LiaCore Domp probably works bc it increases prolactin.
Corticosteroid helps with with Prolactin routes but is a minor helper so kind of unnecessary given risks.
Pio prolly not a good idea, I don't see mechanisms that would fulfill the intended goal, at least in the literature.

[will depue @willdepue]

im sorry but if claude doesn't refuse to work on your research project it's not cool enough. i don't make the rules

[Evï Anton @EviAntonWorld]

@ElleHasAGock :(

[Evï Anton @EviAntonWorld]

@TechMDAI recipe?

[Evï Anton @EviAntonWorld]

@transkatgirl don't use suno, it's very bad and it all sounds same-ish. use minimax music 3

[Evï Anton @EviAntonWorld]

@chaotictransfem banger

[Ben Koska @BenKoska]

while benchmarking our AutoResearch harness we noticed that there was a ~1.1pt jump in benchmark performance when the agent's name is Ilya. it's kind of funny to think about the set of circumstances that made that happen

[Evï Anton @EviAntonWorld]

@AdithyaNLP How many GPU hours did this cost? Are there plans to advance this model further? I was thinking of doing another two rounds of training.

[Evï Anton @EviAntonWorld]

I didn't realize it was so expensive to distill from a parent model and train a small 3B model. Like $10k, wtf?

[Evï Anton @EviAntonWorld]

A day of training later and it's gotten worse, this might be more difficult than expected.

[Evï Anton @EviAntonWorld]

A day of training later and it's gotten worse, this might be more difficult than expected.

[Evï Anton @EviAntonWorld]

Ugh, it's /only/ 2490 elo.... maybe we do a few days of training on it to get it to 3500 elo....

[Evï Anton @EviAntonWorld]

Ugh, it's /only/ 2490 elo.... maybe we do a few days of training on it to get it to 3500 elo....

[Evï Anton @EviAntonWorld]

Maybe, I finally try to hit NM.

[Evï Anton @EviAntonWorld]

@chromaem @chaotictransfem (don't actually)

[Evï Anton @EviAntonWorld]

@chromaem @chaotictransfem ok so i couldn't find a way to do that, maybe you want to use Meta Muse? 😭😭

[Evï Anton @EviAntonWorld]

@chromaem @chaotictransfem i think in the future hermes and codex probably will catch up to grok/cursor but it seems like that team is iterating much faster than others.

[Evï Anton @EviAntonWorld]

@chromaem @chaotictransfem hermes is local so I actually don't like it anymore! it's nice to be able to do computer stuff when away from my computer.

i use the grok version of dots since i have a edu email so I get it for free. they also have a tailscale integration.

[Evï Anton @EviAntonWorld]

Maybe, I finally try to hit NM.

[Adithya Bhaskar @AdithyaNLP]

The most profound paper of my PhD so far. We truly did something special to make LMs play/explain chess, needing both architectural & algorithmic innovation (natural-language analogue of the Alphazero algorithm) + applicable to many other domains. Please read on and share!

1/10

[Evï Anton @EviAntonWorld]

@chromaem @chaotictransfem oh it might not be out yet for all users? https://t.co/qBN3nukW5o

and https://t.co/C4sBJNpy0o

[Evï Anton @EviAntonWorld]

@chromaem @chaotictransfem hey, you should be able to connect dots to tailscale and tell it to use the CUA skill to do computer use stuff.