Gregor Tokarev

21 y.o. Software engineer. building Became teamlead at age of 20.
Currently work at biggest e-com platform in Russia

[Gregor Tokarev @gregortokarev]

@theo Only 2 and 5 not look like shit, but "needs you" and "running" grouping from 3 is great UX

[Gregor Tokarev @gregortokarev]

Github should do these things if it don't want to lose me and all future developers in the future:

- keep SLA above 99.9%
- faster clones and pushes
- faster and chiper actions
- cli that not sucks
- moderation tools for issues and pull requests

[Gregor Tokarev @gregortokarev]

kiru video editor export feature

[Gregor Tokarev @gregortokarev]

Проблема в том, что университет не учит тебя думать. Вообще ничто не может научить тебя думать.

Ты рождаешься с определенным IQ, и ты с таким до конца жизни, если ты не сделаешь его меньше, путем ударов теменем об стену. Все, что ты можешь сделать, чтобы эффективнее выполнять какое-то дело, это получить опыт.

Университеты всегда нужны были, чтобы получать знания. В этом было их уникальное свойство, Сейчас все знания в ИИ.

[Gregor Tokarev @gregortokarev]

I think the Fable and OpenAI's 5.6 SOL ban are actually great for the community. Look at it from a different angle: frontier labs like OpenAI and Anthropic have been winning largely because they won in the past. They had more RL data to steer their models, and every time open source (or any other lab) caught up with their latest model, they'd release a new one. So everybody kept using OpenAI and Anthropic.
But now open source finally has time to catch up. More and more people will start using it, which means these labs get more RL data. Nobody can restrict you from using these models, and inference for SOTA will be way cheaper. Sounds pretty awesome to me.

[Gregor Tokarev @gregortokarev]

@jvqtil Why i should use it instead of make?

[Gregor Tokarev @gregortokarev]

@aidenybai yeahhh finally more dependencies

[Gregor Tokarev @gregortokarev]

@jvqtil @meshtimes_ just don't like this types of attention bytes

[Gregor Tokarev @gregortokarev]

@gonka_ai new https://t.co/j6EHRzEvwa is not enforcing limits

I spend all my welcome bonus 20 mins ago. My agent is still running.

[Gregor Tokarev @gregortokarev]

@DaLiberman @gonka_ai Is it subsidized?

[Gregor Tokarev @gregortokarev]

@DaLiberman @gonka_ai How are you even able to achieve these prices at these speeds? How is kimi 450x cheaper? GPU providers will probably spend more on electricity to generate these tokens.

[Gregor Tokarev @gregortokarev]

https://t.co/YICOqmV8uK

[Gregor Tokarev @gregortokarev]

Free transcription that runs 100% in your browser.

🌟 Drop in any audio or video
🔒 Your files never leave your device — no uploads, no servers, no accounts

I built this for everyone who just needs a transcript without paying enterprise prices.

[Gregor Tokarev @gregortokarev]

@PenguinWeb3

[Gregor Tokarev @gregortokarev]

@bruvimtired @raycast voxfusion, give it a try

[Gregor Tokarev @gregortokarev]

@b_nnett @jewprobablyknow Why do you think Rust resulted in performance issues? Sounds counterintuitive.

[Gregor Tokarev @gregortokarev]

just added parakeet tdt model to voxfusion

Transcriptions now can be 50% faster

[Gregor Tokarev @gregortokarev]

@tsoding @memoakten I don't like this take.

1) An analogy is never proof.
2) We have the same level of confidence that LLMs don't have consciousness as we do that any random person on the street doesn't.

[Gregor Tokarev @gregortokarev]

@saltyAom Several months ago, I subscribed to Brilliant SWE, who created Elysia. Now I'm unsubscribing because I can't stand faggot pics anymore.

[Gregor Tokarev @gregortokarev]

bro anthropic will fuck u over soo hard

[Luis Catacora @lucatac0]

you can just build things