toreb_000

[toreb_000 @toreb_000]

@JustMicrock @scaling01 You also dont get it. If you reach superintellegent models you can create all of that in 10x better whenever you want. If you are a hardcore god-ASI believer like anthropic, you dont need anything except the best model. Everything else can be a byproduct later on.

[toreb_000 @toreb_000]

@scaling01 Arent you German? How are you surrprised when theres maybe 10 people in all of germany that are AGI-pilled 😹

[toreb_000 @toreb_000]

@scaling01 DeepSWE has been saturated for months now, everything above like 68% doenst say anything lmao, it has even been shown that ~30% of the benchmark is flawed which lines up exactly with the scores non-benchmax models get. About as useful as SWE-Bench Verified/Pro now.

[toreb_000 @toreb_000]

@KDHabibi Deutschen sagen sie sollen nicht kleingeistig sein 😹😹😹😹😹😹😹😹😹😹😹😹

[toreb_000 @toreb_000]

@scaling01 who the fuck works at anthropic and says AI cant write "routine emails" XD

[toreb_000 @toreb_000]

@VoidStateKate They are searchong domain experts in non-AI fields. What does being hardcore pro AI have to do with anything? Lmao have you ever met a physics domain expert that deosnt have a high academic physics degree? Same for non software-engineeting fields?

[Alexandr Wang @alexandr_wang]

somehow the corposlops refused to use my video

[toreb_000 @toreb_000]

@scaling01 Worse than Opus when Fable 5.5 and Astra 6.1 are (probably) close to release... But im curious how it performens beide benchmarks...

[toreb_000 @toreb_000]

@SimonasLTU1 „i need you to create a LAVA-lamp, i need you to make it really beautiful and really realistic and i need you to use 10 gazillion subagents to do so“

[toreb_000 @toreb_000]

@JaydenDavisNC I think its time for the „a bike riding a pelican“ test.

[toreb_000 @toreb_000]

@AdamHoltererer Buying multiple plans is against TOS afaik. As soon as enough people start to abuse it, they will just start enforcing TOS.

[toreb_000 @toreb_000]

@MichaelTrazzi @ethansdyer

[toreb_000 @toreb_000]

@Da7_Tech They should have compared it to Sol instead of Luna XD

[toreb_000 @toreb_000]

@peterpme You couldnt spend a single minute to look up actual costs for agents with caching on openrouter? XD Opus is less than a dollar per million. Every model except Fable and Astra is.

[toreb_000 @toreb_000]

@GaryMarcus @aviromanoff @sidereal111 god damn you are a pathetic loser holy

[toreb_000 @toreb_000]

@EducatingwithAI @daniel_mac8 As a not so wise man once said: "Everything is computer". Except the physical world. Except except when robots (which are computer) are catching up.

[toreb_000 @toreb_000]

@Bidiert @Anthony_Bonato MS Word equation editor?!

[toreb_000 @toreb_000]

@bridgemindai What are you talking about GPT 6.1 Sol is definetly not GPT6 Sol with RL lmao Its a looped transformer architecture like Astra and completely different vom 6

[toreb_000 @toreb_000]

@theo Can someone explain wtf that even means? You could use your Plan for anything you wanted before? So what does "you can now use your Plan in other harness" Change?

[toreb_000 @toreb_000]

@forloopcodes Before Astra Codex 200$ plan was ~13.000$ (!) api usage equivalent. While claude has been reported at about 8.000$ for several months now. Therefore, its definetly not even close to 500$ now. Still sad Change though