⚠️🔧⌨️🔥
a land of contrasts
Joined August 2007
- Tweets2K
- Following1.2K
- Followers315
- Likes6.1K
Aaron Miller retweeted
636606729769440499166579950236036751749912014371509557713570027508971809534551913252252094954941974952859310861988904737359709200557919
is a factor of RSA-896
saweis.net/posts/rsa-896.htm…
Aaron Miller retweeted
Is all of your work single threaded? The moment you start using subagents or spec dec or exposing any form of parallelism, SRAM-only machines become a much worse tradeoff.
The market segment for 100x price for 10x speed does not exist yet, at any sort of relevant scale.
Aaron Miller retweeted
Ayyy go have some fun with DiffsHub.com, a crispy clean demo app showing off our new CodeView component for rendering millions of lines of changes with all the fixings—virtualization, streaming, scrolling, annotating, and more. Just replace `github` with `diffshub`.
Aaron Miller retweeted
This is why PR diff speed matters. This isn't a dunk on GitHub specifically, because GitLab, Forgejo, etc. are all equal or worse. But this is the kind of thing that drives me nuts, because this is a core workflow and its slow enough I literally take my hands off the keyboard.
Btw, when my mouse jiggles on the left, its because the page is literally skipping frames and I'm instinctively shaking my mouse to see if it'll respond. And on the keyboard input you can literally here me finish typing before a letter even shows up.
For someone like me who is an expert at these tools, my brain navigates the tool dramatically faster than it can keep up, and that is not good. The tool should not get in the way.
Aaron Miller retweeted
Are there any actual success stories for "software factories" like gastown? Anything I can see running, and preferably see code?
Aaron Miller retweeted
Today, we're launching AEC-Bench — the first open, multimodal agent benchmark for construction.
196 tasks across real construction documents. Full agent harness. Automated evaluation. Apache 2.0.
We benchmarked Claude Code, Codex, and our own agent. Here's what we found 🧵
Aaron Miller retweeted
If I were a threat actor exfiltrating a ton of company data I’d set my user agent to “ChatGPT”
Aaron Miller retweeted
I feel stupid for even paying attention to this stuff in the first place. I should know better by now.
Aaron Miller retweeted
Why do 16k GPU jobs fail?
The Llama3 paper has many cool details -- but notably, has a huge infrastructure section that covers how we parallelize, keep things reliable, etc.
We hit an overall 90% effective-training-time.
ai.meta.com/research/publica…
Aaron Miller retweeted
i used to find this kind of thing intimidating - ML is so popular, competitive, etc
but I realized how different every paper is. each presenter here traveled down one single esoteric rabbit hole
high-dimensional spaces are so sparse; there's still plenty of room for your ideas
ImageBind uses a *frozen* OpenCLIP (w/ different final projection) - so you can embed a cat pic and car horn noise with it, adjust the *unprojected* image emb to be closer to the audio (after projection), then put it through *CLIP*'s final layer instead to search the laion index
Aaron Miller retweeted
I’d phrase it as this: “Ponzi schemes have an unbridled urge to expand, and eventually the expansion of the fake books is so large that it should produce a visible change outside those books, but does not. This fact is often noticed casually by someone uninterested in scheme.”
Aaron Miller retweeted
Them: “Here’s my perpetual motion machine!”
Me: “It doesn’t work. You have to put more energy in than you get out.”
T: “If you’re so smart, where’s YOUR perpetual motion machine?”
M: “I don’t think it’s possible to make one.”
T: “So you’re against free energy then. Got it.”
Aaron Miller retweeted
periodic reminder. wishlist norco for gnostic ditch ceremonies, sentient swamps, fugitive cosmic androids, antebellum petroleum labyrinths, psychoactive urinal crust, etc. it's a zine written in C# to reconstruct a forgotten memory of the flooded american south.
Aaron Miller retweeted
I don’t know why, but this reply has really stuck with me. “web3 is promising because I once wasn’t allowed to squat on 1,200 twitter usernames that I planned to flip for profit”. It’s just… so revealing
Aaron Miller retweeted
Apart from a few notable exceptions media crypto coverage mostly fails to follow this journalism tenet:
teapawt.net/Klujypop/status/101821…
