dwrodri

lots of work to be done but a year of hacking is starting to come together. launching soon!

@vijaythirumalai They are not all equal in terms of mechanistic interpretability research

I simply cannot fathom this move Dario doesn’t look that charismatic to me atleast from his interviews Andrej could have joined 1/ Elon 2/ Zuck 3/ Google 4/ Apple 5/ MSFT Everyone would have given him a blank check and much bigger role and platform How was Dario able to hire him ?

"Yeah I've started planning dates with my wife by asking Gemini every Thursday" are you listening to yourself right now

Watched 20 mins of Google I/O and couldn't help but think this article is aging finer every year.

How do people move from mac -> linux and keep keybinds? I heard about toshy but sounds invasive. Mac binds are written into back of my school in particular cycling thru tabs, i don’t wanna learn new binds i am too old for that

@GoodfireAI @praveenakode Polosukhin mentioned Arrival as inspiration in the FT interview on Attention Is All You Need ft.com/content/37bb01…

@praveenakode great movie, very similar

Neural networks might speak English, but they think in shapes. Understanding their rich *neural geometry* is key to understanding how they work – and to debugging and controlling them with precision. Starting today, we’re releasing a series of posts on this research agenda. 🧵

command pallette / raycast / everything bar is the best software ux to rise in popularity throughout the 2010s and it’s not even close

@zeddotdev @baseten might be kinda dumb in terms of cost-effectiveness but would also be nice if i could bring my Zed Pro to other clients besides Zed. I’d like having GPT/Claude/Gemini take my credit card from one place and happy to support y’all instead of OpenRouter.

Would you buy Zed Pro if we added open source models from our friends at @baseten

first impression of gpt 5.5 is this thing is worth the money, let’s see how long until i eat those words

Might have to set up a reverse proxy on this thing because ain't no way I am buying more VPS instances at that price and I have yet to flex this hardware, but I have a lot of projects that need hosting

VPS-3 I bought from OVH cloud in November 2025 for $153 would cost $240 for me to renew today 💀

exceptionally rare occasion that i let a chatbot drive code implementation and we leave the encounter with fewer lines of code then handwritten, or even on net when doing a drive by refactor pass. I am always pointing it out to even Opus. As someone who is generally anti loc counting this feels like it’s worth addressing. Even in a “LLMs write ALL code” world, terminal state is constantly be fighting to optimally pack your ctx window. Automated cleanup passes sorta help if you intentionally prompt the bot to gut code, but it feels like a waste of tokens. Seems like one of the biggest quantitative blockers to getting these things in large mature c++ codebases. They are just SO verbose.

@tommiedotjpg ong we got two uniqlos in boston again had me shook

Uniqlo taking their corners.

Companies be like “let’s use semver” and then release v0.278.3

wonder if frontier labs have provisioned extra hardware for us tax weekend

@AlpinDale @HSVSphere weird, 5070Ti here and KDE+Wayland is the only combo which properly supports HDR without gamescope nonsense for me. Seems to be a weird bug where i need a low res display connected to boot though, and I HAVE to plug in the large display only after the logon screen shows.

@HSVSphere KDE with CachyOS

>boot up new PC with RTX 5080 (upgraded from 2070S) >wayland >screen flickers when playing games >screen timeout kills display, have to switch to a different TTY and back to get output >text not green yet Back to xorg I go.

im goofin around w/ local ml is that new 32GB intel card worth the money

wait this is actually wild, a functioning Nvidia gpu on Apple Silicon

moved into a new place with basement office but it gets up to 1250ppm of CO2. it’s a basement, so opening a window isn’t an option. Any ideas?

@0xSero @thdxr if this is real and you pick me i will write a blog post meaningfully assessing the GLM/Kimi models for software work in Rust and Python by June

Giving away 5 Opencode Go subs Winners selected randomly from comments in 24 hours.

@HolgersenTobias @sterlingcrispin @garrytan value created for yc companies whom garry pays for each token lol

@sterlingcrispin @garrytan Aren’t LOC and tokens burned the ideal metrics for value created for real people in the real world?

Absolutely insane week for agentic engineering 37K LOC per day across 5 projects Still speeding up