Edward Kmett
@kmett
Followers
17K
Following
22K
Media
715
Statuses
13K
Founder/Chief Scientist @Positron_AI Haskell, category theory, AI, and safety. https://t.co/wYaTDISbEB https://t.co/XwbnuaOUzq 🦋 @kmett.ai
Here, Now
Joined October 2008
Given a choice between two things, I always try to do the one that nobody will believe me about later.
9
34
325
Welp. Today's apparently a @ClaudeAI day. The @ChatGPTapp website keeps popping up the same popup asking me if i want to install Codex CLI over and over and over and over and over. Great UX, folks.
0
0
7
.@satnam6502, so er.. how does this stack up as a candidate for the https://t.co/TGFgjuL3kP ? https://t.co/BLh4OA2wT5
fpcastle.com
A representative FP Castle
3
0
24
I am a functional programmer not because of the company I work for, but for the company I keep. -- @satnam6502 "Functional Programming and Hardware Design" is such a generic title for such an incredibly dynamic and heartfelt talk. Well worth your time! https://t.co/A9GeKUZSMY
0
21
166
A new VentureBeat article just landed about https://t.co/FrtDOZm1il.
https://t.co/aR9KfaL1DY
venturebeat.com
The company’s first-generation chips were fabricated in the U.S. using Intel facilities, with final server assembly and integration.
0
4
14
GPUs made training massive models possible, but inference needs better memory capacity, memory bandwidth utilization, more power efficiency, and an architecture built bottom up with transformers in mind. To that end, I'm excited to share that Positron just raised a $51.6M Series
3
12
48
RIP @googlechrome 2008-2025. It was a good 17 year run. Today is the day I'm finally forced to uninstall Chrome. They finally forced uBlock Origin off by "upgrading" behind my back without explicit authorization when I had to restart my machine, in a fashion I can't just
17
14
199
> fp8 is 100 tflops faster when the kernel name has "cutlass" in it kms https://t.co/KpZjwSAkrM
github.com
Rewrite the attention kernel to be persistent. This gives better performance at low-contexts. However, fp16 at large context has suffered a bit due to a ptxas instruction scheduling issue in the so...
27
41
359
I had someone describe my approach to life (and to sleep in general) as an 'any% speedrun' today, and I cannot unsee this.
4
0
29
I can't be the only one who gets this UI bug in ChatGPT almost all the time. It gets stuck thinking it's talking, but it has finished talking. I can't interrupt it with the stop button because it's not talking. And I can't respond because it thinks it is. The chat dies,
4
0
9
One problem with leaning so hard on closed models is that on some days they just take stupid pills. Maybe it is throttling, some kind of meta-level change in the way they do chain-of-thought, whatever. Who knows? The model can't tell me. Today it seems ChatGPT is "being the
10
2
60
It is incredible just how disjoint the support for SystemVerilog language features is between, say, Verilator and Genus. Verilator: What is a parameterized function or type? The only way you can parameterize either of those is if you shove that in an interface and put the
3
1
21
Today I used Sutherland's logical effort to reason about prefix adders. With it I found g (logical effort), p (parasitic delay) and h (load) applied them to paths for g (generate), p (propagate) and h (Ling-style pseudo-carries). No notational confusion ensued. *cough* None.
0
0
10
It seems the Disney live-action remake pipeline has finally made it to Pixels (2015). Bold choice not to have Adam Sandler reprise his role.
0
0
4
I originally wrote this library to er.. re-answer a stack overflow question. It has since been used for everything from analyzing high energy physics data to machine learning to computer graphics to pick and place machines to keeping flying cars aloft with people in them. The
Reverse-mode AD in Haskell? It’s not just for machine learning —it's a game-changer for building reliable, high-assurance systems. We break it down in our latest post: 👉 https://t.co/E00GkSMecG
#Haskell #FunctionalProgramming #AutomaticDifferentiation #DevInsights
3
31
164
Frog put the CoT in a stop_gradient() box. “There,” he said. “Now there will not be any optimization pressure on the CoT.” “But there is still selection pressure,” said Toad. “That is true,” said Frog.
Detecting misbehavior in frontier reasoning models Chain-of-thought (CoT) reasoning models “think” in natural language understandable by humans. Monitoring their “thinking” has allowed us to detect misbehavior such as subverting tests in coding tasks, deceiving users, or giving
21
71
907