Ruslan @guyfulla1

Joined November 2012

Tweets

38
Followers

9
Following

85
Likes

339

Ruslan @guyfulla1

2 weeks ago

@VictorTaelin we need a `how many more 9s in the "release next week" probability till release` bet on polymarket 🤓

0 0 2 58 0

View Details

Ruslan @guyfulla1

2 weeks ago

@VictorTaelin @Ceoz_1 but that's a nice ergonomics test - let e.g. deepseek v4 flash code it up, and see if it can figure things

0 0 1 111 0

View Details

Ruslan @guyfulla1

2 weeks ago

@VictorTaelin drop the fckng link

0 0 0 295 0

View Details

@BigBackup27 @Joadys @kimmonismus very well might be cherry picking "look, my new model so good it solve NS by itself, nevermind fine tuning on fresh user data cause we know somebody has the solution by now and we will never let you know what we trained it on"

0 0 0 21 0

View Details

Ruslan @guyfulla1

2 weeks ago

@DrKr00 @wjmzbmr1 ikr, why is nobody asking this OpenAI should come clean why, were they fine tuning knowing Codex sessions have leads to the solution (but hey never will)

0 0 1 66 0

View Details

Ruslan @guyfulla1

2 weeks ago

@kimmonismus kinda sus tbh sounds like a fine tune on Codex maths research logs to beat it to the punch

0 0 0 1K 0

View Details

Ruslan @guyfulla1

2 weeks ago

@markchen90 It'd be curious to hear more about this internal model OpenAI started training end of August (after the rumors) Why use it right away on the specific problem of Navier-Stokes primarily? Otherwise sounds like a fine tune on the research, to beat it to the punch.

0 0 1 943 0

View Details

Ruslan @guyfulla1

2 weeks ago

@mobav0 *on user logs of the people who have allegedly solved it using OpenAI products

0 0 1 10 0

View Details

Ruslan @guyfulla1

2 weeks ago

@mobav0 sure might've been sleepless nights, fine tuning a private model 28th of August, after rumors of somebody allegedly solving it Navier-Stokes and conveniently enough this new checkpoint turns out to be good for attacking the problem and, btw, OpenAI might've trained on user logs

1 0 2 116 0

View Details

Ruslan @guyfulla1

2 weeks ago

@ns123abc

0 0 4 2K 0

View Details

Ruslan @guyfulla1

4 weeks ago

@VictorTaelin lets go!!

0 0 0 182 0

View Details

Ruslan @guyfulla1

4 weeks ago

@relace_ai all good now, thnx!

0 0 0 8 0

View Details

Ruslan @guyfulla1

4 weeks ago

@relace_ai it has happened before, 27th iirc, but today it is really bad

1 0 0 27 0

View Details

Ruslan @guyfulla1

4 weeks ago

@EBorgnia thnx guys nice service!

0 0 0 14 0

View Details

Ruslan @guyfulla1

4 weeks ago

@VictorTaelin glad to see it progressing so far so fast

0 0 0 10 0

View Details

Ruslan @guyfulla1

4 weeks ago

@VictorTaelin o_o oh my, is it still an interactions combinator machine? is it something new, is it something in between? last time I came across your tweet you seemed to be amidst an AI psychosis event a few months back 🥸

1 0 0 47 0

View Details

Ruslan @guyfulla1

4 weeks ago

@EBorgnia where do I report the bug? I'd like to use your service, it's so annoying

0 0 0 17 0

View Details

Ruslan @guyfulla1

4 weeks ago

@EBorgnia hey you guys might want to know Relace is botching up about 50% of its answers rn for my app it returns raw DSML instead of properly parsed tool calls/etc - the model response is perfectly fine seems to happen to "thinking off" deepseek v4 flash, and it's a recurring error