R Sz

Passionate .NET developer

[R Sz @rob_szach]

@CuleMadafaka @donald_PL_ XD

[R Sz @rob_szach]

@michalmac @donald_PL_ Z tym drugim się zgodzę, ale z tego co piszą to sama poszkodowana była też prowodyrem.

[R Sz @rob_szach]

@Danyyy33b @donald_PL_ Zgadza się, ale to też nie powód by oberwała jakaś działalność, bo to nie tylko jedna osoba.

[R Sz @rob_szach]

@luka_100_ @donald_PL_ Ma pan dowód? Nie widzę by napierdalal, nie ma obdukcji etc

[R Sz @rob_szach]

@NS9822817321 @donald_PL_ Skąd wiesz czy uderzył te panią?

[R Sz @rob_szach]

@SlipperyGem https://t.co/o5ayyc4OxC

[R Sz @rob_szach]

@melvindvivas @opencode Let be frank. DeepSeek API directly is 3 times faster than this in 30-50% of the usage time due to some rate limiting or I don’t know what. Because DeepSeek is so cheap to use directly, I’d rather to stay with that.

[R Sz @rob_szach]

@ZacariahHeim @opencode Next time I will do so, thanks for info! Now, that’s outdated ;D

[R Sz @rob_szach]

@opencode
Can you add a model for opencode go:
webbrain-one/DeepSeek-V4-Flash-0731-Vision-NVFP4

to allow image inputs for DV4 flash 0731?

[R Sz @rob_szach]

@donaldtusk @NawrockiKn XD

[R Sz @rob_szach]

@RBrzoska @Meta Absurd.

[R Sz @rob_szach]

@opencode How is that happen if they have their zCode harness and there we can use it with images/screenshots/browser integrated

[R Sz @rob_szach]

@tianyi What is the purpose to create such app?

[R Sz @rob_szach]

@oTOOLony_steFAN @Zaufana3Strona Albo są Polakami.

[R Sz @rob_szach]

https://t.co/3VIuwoXRJl

Am I stupid or you make me stupid? "1.1 is also dramatically more efficient. In GitHub Copilot tokens stream 25% faster and the model uses 25% fewer tokens to complete a task. " Token per second didn't change. You just decrease tokens. It's not faster.

[R Sz @rob_szach]

@bandyci_drogowi Brona opada sama jak ciągnik jest wyłączony bo nie pompuje hydrauliki. Dodatkowo tak jak ktoś napisał, jak koło wpadnie w koleinę to i tak spadnie ten sprzęt.

[R Sz @rob_szach]

@opencode @thdxr ☝️

[R Sz @rob_szach]

@JamCiTo6 @jack__sparrow5 Na podstawie błędów ortograficznych

[R Sz @rob_szach]

5/ So is this mainly genuine concern about recursive self-improvement?
Or is it also political groundwork for tighter controls that will inevitably hit open / non-US models hardest?
Full statement: https://t.co/D25LvA77yW

[R Sz @rob_szach]

1/ 1,337 employees from OpenAI, Anthropic, DeepMind & Meta just asked the US government for tools to deliberately slow down AI progress.
But they barely mentioned the one thing that makes “pacing the frontier” almost impossible.
🧵

[R Sz @rob_szach]

4/ DeepSeek, Qwen, GLM and others are already closing the gap fast, cost a fraction, and are freely available worldwide.
Once research-automation-level models go open-weight… the idea of controlled pacing starts looking very different.