A1RHE4D

Y'all are a bit dense, still Im the lightest in the head airhead!!
Shitposter
Here to be annoying

[A1RHE4D @A1Rhe4]

@HumorOverSorrow @davepl1968 IDK man I think it needs a new car. The alignment of new tire would be off, not to mention its age w.r.t other tires and the axle of the car.

[A1RHE4D @A1Rhe4]

@beffjezos Bandwidth still limited by eyes.

[A1RHE4D @A1Rhe4]

@suchenzang Also hacks to propagate feedback properly to lower layers which are justified later in hindsight lol

[A1RHE4D @A1Rhe4]

@snr_boost @basedjensen They just needed to fund uni labs. NVIDIA (before cuda took off so without AI money) used to do it, and so do some other companies. A lot of uni lab groups used to be more capable than industry, until Uber created a precedent (poached robotics lab in CMU) and others did so too.

[A1RHE4D @A1Rhe4]

@TVMohandasPai Lol crazy bro it was supposed to be invested in 10 years ago amortized over the years. Doesnt cost 60b more around 5.
The huge bill upfront now is just the cost of sitting pretty while other countries invest in capabilities.

[A1RHE4D @A1Rhe4]

@CooperZurad Does the hardware you run on keep running forever?
Hardware behaviour is one of the assumptions that goes into the design of software, as the assumptions evolve(culture, market behaviour, preferences, habits) so does the need for software to catch up. Same with models.

[A1RHE4D @A1Rhe4]

@basedjensen Routing to serve queries via distilled models is silent degradation. On top of the cost efficiencies built inside the models themselves...

[A1RHE4D @A1Rhe4]

@HououinTyouma This says more about your cohort than humans

[A1RHE4D @A1Rhe4]

@Ressniee This is why Kierkegaard was depressed.

[A1RHE4D @A1Rhe4]

@aliceisplaying They've only become great since dec 25. Many ppl didnt come back after they tried during the hype era..

[Teng Yan @tengyanAI]

basically: anthropic sneakily turned down how hard claude thinks before editing code, changed the default from "high" to "medium" effort, and hid the reasoning from session logs. all without telling users.

an amd director had 7k sessions of telemetry to prove the degradation was real and measurable (not just vibes). anthropic admitted to the changes. there's a workaround (use "/effort max"). the uncomfortable part is most users had no data to notice it happened at all.

[A1RHE4D @A1Rhe4]

@wsdd2_fhc @butterflyfact8 You might have regretted the PhD as well- such is the mystery of life. We obsess over optimizing over a single dimension while life itself is experienced in multiple facets.

[A1RHE4D @A1Rhe4]

@davepl1968 Tastes like medicine which is good as I can avoid all soda

[A1RHE4D @A1Rhe4]

@HSVSphere And hasn't heard of arena allocations

[A1RHE4D @A1Rhe4]

@MadelaineLucyH I also like to read the ramblings of a sick mind in search of companionship in my madness.

[A1RHE4D @A1Rhe4]

@icpolicy @VinciRSS the apis. Ever heard a file try to send data to you?

[A1RHE4D @A1Rhe4]

@icpolicy @VinciRSS And they are against the unix philosophy of everything is a file. The win isnt because of adhering to the unix philosophy but by being practical.

[A1RHE4D @A1Rhe4]

@MichaelAluya3 @bindureddy Its making the 10x engineer a 5x engineer. And its mandated to be used. Gonna be fun.

[A1RHE4D @A1Rhe4]

@bindureddy Its all by design. Please buy more Claoud that comes with the distilled experience of REAL CAGE FREE™ programmers raised in an AI free environment. Unlike clodex slop which combines experience across synthetic and high cortisol programmers. Why pay full for less?

[A1RHE4D @A1Rhe4]

@MWagner1971 @AnotherArtist2_ @StefanMolyneux It took industrial revolution to create a middle class, a class that rose purely because capital needed labour to create wealth. Automation helped increase productivity that led to increased wages.
Now you wont need any human in the loop, the multiplier is gone.