Dream

Bitcoin cures us of our Oneirophrenia. That which should not exist, and yet does. The lines are blurred. A Dream.

[Dream @DreamBTC]

This. https://t.co/mVfelYSTjJ

[Tone Vays @ToneVays]

Just read about #TheDAO & had to update the graph. Almost there!!! #bitcoin #blockchain #ethereum $BTCUSD $ETH #btc

[Dream @DreamBTC]

@chrismd10 You are correct!

Source: my 5 year old having the time of his life

[Dream @DreamBTC]

@atmoio @Sentdex No there are still definitely a lot of unsolved or unoptimized things to fix

The LLM wouldn't know dijkstras or memoization or ray tracing if someone didn't invent the concept first

There are still infinite amount of those discoveries out there. Current models can't do this yet

[Dream @DreamBTC]

@atmoio Great video man thank you for sharing. I def sympathize with how you're feeling, but I don't feel the same way at all.

I basically took the bargain because the end state was always more important to me than the journey.

As for the way out it would take longer than a tweet.

[Dream @DreamBTC]

@pronounced_kyle Hell yeah love culture series

[Dream @DreamBTC]

@awesomekling 4 is plenty appropriate. Good job unc

[Dream @DreamBTC]

@bradmillscan The models, because of the way they are trained, do have a tendency to tell you "what you want to hear", or "what is most entertaining" according to its understanding of what you're saying.

Both of those tend to take precedence over "what you actually want" in subtle ways.

[Dream @DreamBTC]

@badlogicgames I am having so much fun

[Dream @DreamBTC]

@Jack_Raines It achieves the same thing or better for 99% of use cases, with less setup or resources.

So at worst super directionally correct

[Dream @DreamBTC]

@csuwildcat @mmatthias Followup @csuwildcat here's a no bullshit technical blog post from a stripe engineer.

-Atypical stack (no rails ruby)
-big scale and $
-they focus on beginning and end, let LLM do the middle

and they even used Block's goose as harness!

https://t.co/b5vSTorBxZ

[Dream @DreamBTC]

@csuwildcat @mmatthias Source: I work for fortune 50 medical device company. pretty complicated rbdms with multiple inputs to sanitize and standardize

I feed that into our cloud and create an ai chatbot for non-technical business users based on that data. Not crazy scale but nontrivial

cc writes 99%

[Dream @DreamBTC]

@bradmillscan @AlexFinn @MatthewBerman It is inspiring somewhat but also consider that they are making polished videos that DO NOT show you all the time it's taken to set all that up and deal with the same pitfalls you're having.

Ask any technical ppl in Bitcoin you trust that's messing with this stuff. It's not you

[Dream @DreamBTC]

@matt_mcclure lol that sounds delightful

[Dream @DreamBTC]

@jmrphy @codegraph Nope you got it. And yes it's awesome

[Dream @DreamBTC]

@boristane @thdxr Good canary to check to see if he's actually using AI to write posts?

[Dream @DreamBTC]

@bradmillscan @rot13maxi It is a retard in many ways. It's "50 first dates + Memento" as karpathy says.

You will be constantly fighting against this. It's built in. It's not figured out yet.

If you want it to use git, you have to set it up and test it in a basic way first, before any projects start.

[Dream @DreamBTC]

@liamzebedee @zetalyrae @mathlover3b1 Any of the options mentioned so far is definitely better than Ubuntu.

Nix is awesome but would have the annoying learning curve up front.

Omarchy would fix everything faster but be more opinionated.

Bootstrap arch or Debian somewhere in the middle.

In age of LLMs all glorious

[Dream @DreamBTC]

@bradmillscan Openclaw can be great but yes you have to know where the sharp edges are.
It requires an understanding that people usually find with guess and check like you are right now.

Kinda obvious but yeah start small. Check comprehension often with hypotheticals and test runs.

[Dream @DreamBTC]

@VictorTaelin Use pi!!!

[Dream @DreamBTC]

@Cromwelp It wasn't me but I just went through the whole journey myself late last year.

Absolutely incredible.