Poctave Plaba

CEO de PessiOVH

[Poctave Plaba @PessiOVH]

@btmclive What are your thoughts about peppy being very political and against AI in the development of osu! lazer?

[Poctave Plaba @PessiOVH]

Can't they do like they did with 5.6 Luna where they make their smartest model optimise its inference itself?

[Poctave Plaba @PessiOVH]

The usage given to Plus users when using Astra is awful, even in light reasoning a simple message that takes Astra 4 minutes uses 12% if the 5h limit. @thsottiaux valve, pls fix.

[Poctave Plaba @PessiOVH]

The usage given to Plus users when using Astra is awful, even in light reasoning a simple message that takes Astra 4 minutes uses 12% if the 5h limit. @thsottiaux valve, pls fix.

[Poctave Plaba @PessiOVH]

@Sardoche On veut pas de toi dans la commu d’osu! mec

[Poctave Plaba @PessiOVH]

@z9a_osu4 sorry but that looks terrible

[Poctave Plaba @PessiOVH]

@jarredsumner @PreslavMihaylov Isn’t the goal simply « make things faster »

[Poctave Plaba @PessiOVH]

@EmperoAI Do we have benchmarks to measure the intelligence lose versus the original model and versus other quantized models like unsloth?

[Poctave Plaba @PessiOVH]

@o7laurence being a goy has been the best decision I've made in recent memory

[Poctave Plaba @PessiOVH]

@scaling01 Why don’t they make it think in Chinese

[Poctave Plaba @PessiOVH]

@saltjsx Vercel is made by israelis so it doesn’t shock me that they let CSAM on their servers

[Poctave Plaba @PessiOVH]

@theo context usage pie when?

[Poctave Plaba @PessiOVH]

Anything but adding basic stuff like a context usage pie, which is one of the main reasons why I haven't switched to t3 code yet

[Theo - t3.gg @theo]

Merged and shipped 48 PRs in 2 days 💀

[Poctave Plaba @PessiOVH]

@kyunhost Holy honeypot

[Poctave Plaba @PessiOVH]

@maria_rcks Prompt injection

[Poctave Plaba @PessiOVH]

@yagiznizipli you are comparing a compiled language vs an interpreted language and runtime, that isn't fair at all, compare two compiled language or two runtimes instead.

[Poctave Plaba @PessiOVH]

@D18K3 @OpenAIDevs 5%?

[Poctave Plaba @PessiOVH]

Hello @thsottiaux codex cannot use 5.6 luna which is apparently made to be used as a subagent, which is that??

[Poctave Plaba @PessiOVH]

@Discussingfoot What is the source of it? I couldn’t find the source on TPB or TorrentLeech, unless this is a telesync?

[Poctave Plaba @PessiOVH]

@ShittyTwittter @ImSh4yy GPT 5.6 medium runs about 6 hours before hitting the entire weekly, that’s about 2 days for me and around 100m tokens

[Poctave Plaba @PessiOVH]

@thsottiaux Being cheaper