@statusfailed

founder @ https://xitter.cf/t.co/DuFsvLWGF0 - verifiable AI (catgrad & ???) - theorem provers (Metacat) - combinatorial syntax for compilers!

statusfailed.com
Joined May 2012
  • Tweets1.8K
  • Following801
  • Followers737
311Media
cyclic open hypergraphs are tricky to draw. Here's ONE SIMPLE TRICK to run your acyclic layout algorithm on cyclic diagrams! GraphViz Users Hate This! (kidding i love graphviz) blog.hellas.ai/blog/cyclic-h…
1
3
1
13
1,174
your regular reminder that X is highly contested information space extra-spicy contested near election times
0xSero
@0xSero9h
I need to clean my feed up No more politics, I suggest you do the same
32
statusfailed retweeted
statusfailed
@statusfailedOct 25
type-system design is the feature engineering of formal verification
Arjun Guha
@ArjunGuhaOct 24
Will we see a meta result? Consider some compiler (eg, rustc). On an OS+CPU model, prove that for all input strings, the compiler either rejects the string or produces an executable that does not produce memory errors on the same OS+CPU. At that point, the compiler is safe (on that platform) and source language semantics were not needed.
1
2
23
883
statusfailed retweeted
Remco
@recmo13h
Replying to @gakonst
Basically revisit any idea that was blocked on software-engineering complexity!
1
6
50
4,122
honestly good prompting is still valuable you have to understand your own thinking first "Writing is nature's way of letting you know how sloppy your thinking is"
1
147
statusfailed retweeted
Léo@Leik0w011h
Consumers taking their ownership back from greedy companies is so wholesome.
Xuan-Son Nguyen
@ngxson21h
Printed the first "Hello world" from my open-source firmware 😁😁 It's reversed (pun intended), but we are close!
18
598
the lesson is more bitter than we could possibly have imagined
1
6
245
you have never been more able to just do things
50
my hard drive is like 50% model weights, i gotta delete some models
1
4
185
Replying to @lfgist
Can't even reverse entropy.
5
5
272
6,855
ironically now we have autoregressive ML writing code, making software now feels like diffusion
1
7
307
type-system design is the feature engineering of formal verification
Arjun Guha
@ArjunGuhaOct 24
Will we see a meta result? Consider some compiler (eg, rustc). On an OS+CPU model, prove that for all input strings, the compiler either rejects the string or produces an executable that does not produce memory errors on the same OS+CPU. At that point, the compiler is safe (on that platform) and source language semantics were not needed.
1
2
23
883
this is probably a good time to launch my data parallel proof checker but it's not READY YET
Jason Gross
@diagram_chaserOct 24
The bitter lesson of ML is that scale eats all. The bitter lesson of PL is that [efficient] brute force eats all.
1
3
130
i am no longer a fanboy of any language (sorry Haskell) but Rust + LLM is really pretty nice for getting shit done
5
334
statusfailed retweeted
Alex Kavvos
@lambdabetaetaOct 25
Replying to @statusfailed
Don’t @ me I’m Clauding
1
1
3
99
a lot of my euromoots awake suspiciously early (or late?) this morning... lookin at you @lambdabetaeta
1
1
244
today i'm writing pseudo-RISC-V in order to better understand category theory life is strange
2
7
408
the future has never been more unevenly distributed
2
124
documentation-oriented programming
1
1
4
237
a superset of paper-oriented programming
1
1
26
just accept that your program is a morphism in a symmetric monoidal category and enjoy life
Lewis Campbell@LewisCTechOct 23
If you really think about it, having both "functions" and "operators" is bloat. If you really can't deal with concatenative "1 2 +" or LISP "(+ 1 2)", then make every function infix "1 + 2", "array ref i" "Unary" function? second arg is ().
1
2
10
535
Load more