Youssef Wasfy

Joe
Software Engineer - Investor - AI Enthusiast

[Youssef Wasfy @YoussefWGerges]

@Aslex @TimJayas Grok 4.6 used to be really good but lately it’s almost unusable, really very bad

[Youssef Wasfy @YoussefWGerges]

@StefanoGPT Leave the reset date unchanged for sure. This will help me more

[Youssef Wasfy @YoussefWGerges]

@VadimStrizheus Ok it’s a good model but infinite usage?!

[Youssef Wasfy @YoussefWGerges]

@pcshipp Astra is theoretically available in pro plan , but can you do anything with it?

[Youssef Wasfy @YoussefWGerges]

@AlexDochioiu @thsottiaux Did you count the resets in?

[Youssef Wasfy @YoussefWGerges]

@Teknium Oh your saying that the agi is failing to orchestrate? Or maybe is not in the mode to orchestrate, agi is different.

[Youssef Wasfy @YoussefWGerges]

@Da7_Tech All the compute goes to 4.7 now

[Youssef Wasfy @YoussefWGerges]

@blkmtlemblems66 @buildwithrajath With Astra and fable 5.1 you can share your budget with the model and it will try to solve the problem within the given budget

[Youssef Wasfy @YoussefWGerges]

@mark_k Add me please

[Youssef Wasfy @YoussefWGerges]

@YURIWinchest @buildwithrajath Yes exactly and this is the best thing about it

[Youssef Wasfy @YoussefWGerges]

@thsottiaux Astra medium in Codex burns very fast through the limit. Can you please investigate this? It's amazing, but it's really very token-hungry, and this is inside Codex. Inside other harnesses, it's better for me.

[Youssef Wasfy @YoussefWGerges]

For me, Astra medium burns very fast through the limits. I am on a 20x limit and use Luna subagents! but its noticably faster through the limits.
Second, in other harnesses, it burns slower for the same task than in Codex. I think some kind of optimization is needed.
Overall, I am very happy with Astra, and it's really, really a good model, the first model that I feel its my assisstant not just a model.

[Youssef Wasfy @YoussefWGerges]

@aapakari Yes, I also think so. Just have the feeling after they cancelled 3.5 pro that they might want to drop a surprise to come back. They should keep the expectations very low this time

[Youssef Wasfy @YoussefWGerges]

@mattyp @jonoringer Excited for this model. Do you think it will be way better than 4.6 , which is already number 1 for me for everyday coding tasks

[Youssef Wasfy @YoussefWGerges]

@chanduuu_cs Physical intelligence is the future.

[Youssef Wasfy @YoussefWGerges]

@Henryf1w A human to pay for Astra to keep it running! A human to delegate the work to Astra, and a human to pay for the Output and make use of it!

[Youssef Wasfy @YoussefWGerges]

@KyleLoraine92 @adhamuxi not really unless you have Luna subagents doing the real work, which means lower quality but you get the work done, but if you have sol agents working around the clock, you are done with the limit in max 24 hours

[Youssef Wasfy @YoussefWGerges]

@catalinmpit Astra can write in your voice and sound like you. This is one of its strengths