Zebin

Wake up Azathoth!

[Zebin @ZebinWilson]

@vishnuvig Did not expect cache hit at 0.003. Nicely done!

[Zebin @ZebinWilson]

@alfcnz Everyone loves floating pins, till they start RF blackmagic

[Zebin @ZebinWilson]

@ptrschmdtnlsn Top tier shitpost

[Zebin @ZebinWilson]

@hdarshane This would suggest we'd need plasticity over the layers to approximate gradient descent. But damn that's gonna take some time to get it right

[Zebin @ZebinWilson]

@JFPuget @YouJiacheng The urge to write the inference engine with fixed params but shitty perf just for the interview

[Zebin @ZebinWilson]

@alcuintweets Meeper is back
yaoi stocks going back up

[Zebin @ZebinWilson]

@difficultyang You cannot just say this and dip without the sauce :(

[Zebin @ZebinWilson]

@ArielKwiat Direct editing? You mean update the weights?

Wait a second, why havent folks tried this!

Current llms can run tracers for the logits and should be able to update weights directly

[Zebin @ZebinWilson]

@jwt0625 You should do these more lol

[Zebin @ZebinWilson]

@shri_shobhit I think it has more to do with customer data which has higher value, compared to your agent shopping/interaction data (which probably changes for each update).

Makes sense for existing companies but for new companies, llm friendly setup would be so much better for discovery.

[Zebin @ZebinWilson]

@frontier_foid @tugot17 Why do you say that? I've seen MES and ring timeout, but those are firmware issues, right?

Amd can patch those, but whether they will is a separate question.

But have you encountered literal hardware issues? If yes, how did you identify or trace it back?

[Zebin @ZebinWilson]

@vaizaragorn And it'll be charged at least 1000 per piece

[Zebin @ZebinWilson]

@henrylhtsang And another for model routing within the family :)

[Zebin @ZebinWilson]

@johnloeber You can't do this aaaaaa

2nd post having fun problem without giving it away

Also if it's a Haskell challenge, no

[Zebin @ZebinWilson]

@_imdawon Yeah true. I can't just wait to die and start over in the next run!

[Zebin @ZebinWilson]

When do we get cheap transparent displays aaaaa

https://t.co/lP3wMSASVc

[Zebin @ZebinWilson]

@raysan5 I'd argue the opposite. Because agents are writing most of the code, searches for documentation or errors would reduce, without an increase of search for other languages.

I think search for languages would go down (this index's ratings as well) as models get better.

[Zebin @ZebinWilson]

@haydendevs High chances of shorting itself/potential fire. BMS usually cuts it off below temps.

iirc it's from lithium building spikes on the graphite plate and short circuiting

[Zebin @ZebinWilson]

@zackslab

[Zebin @ZebinWilson]

@DeGatchi Usually taking meds or breathing exercises?

I just completely dip my head in a bucket of water and hold it there for sometime. Usually clears the head