Software developer · lateral thinker · father of three · ex-Ubuntu & KDE contributor · Devbox, Nix, Neovim, custom keyboards.
Joined October 2009
- Tweets2.8K
- Following4.1K
- Followers468
- Likes26.8K
Rodrigo Belem retweeted
WTH: Sam Altman complaining that individuals (like me!) are purchasing RAM for our PCs wanting to run local inference (and do other stuff), because it gets in the way of him buying more RAM to sell inference at 80% margins… this is barking up the wrong tree
sam: local inference is a terrible allocation of scarce resources
“a lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we're going to be able to run them on our computers, and so we're not going to need all the stuff in the data centre
but from a global allocation perspective, actually it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time
you want maximum utilisation in the cloud if you're trying to deal with this very supply-constrained environment
somehow many people's intuition is the opposite of that”
Cua works great 😃👍
i tested agent-desktop (opus 5.5) vs. cua (opus 5.5) and vs. codex's computer use (terra 6.1)
agent-desktop finished 26 of 35 jobs on my mac
cua finished 33, codex finished 34
the jobs it missed were dragging, clicking on a canvas and scrolling a huge list, all with its mouse-free mode
let it take over the real mouse and it did all three, one try each
i'm sticking with cua and codex's computer use for now, they are more robust and faster
(cua is my own tuned build)
Rodrigo Belem retweeted
Aprendeu com a IA? Feche o chat e explique em voz alta. Se travar em 2 minutos, você copiou.
Rodrigo Belem retweeted
anthropic: having your own AI models is too dangerous. just rent ours, we’ll keep you safe
openai: having your own hardware is a terrible waste of resources. just rent ours, we’ll put it to better use
you will own nothing & you will be happy
Rodrigo Belem retweeted
Vercel Labs is building something interesting: scriptc.
Compile TypeScript / JavaScript directly into native executables or WebAssembly.
No Node.js required at runtime. No JavaScript engine for statically compiled code.
It uses TypeScript’s type information to generate native machine code via LLVM.
Still experimental, with limited JS/Node.js compatibility, but this is a direction I’d love to see evolve.
TypeScript’s developer experience with native execution.
github.com/vercel-labs/scrip…
Rodrigo Belem retweeted
DeepSeek open-sourced a little tool today. Long name: dsh-superagent-prototype.
It watches your GitHub projects for you. Someone files an Issue or opens a PR, and it can take a first pass: tweak things, draft a fix, then hand the result back for you to review. You wake up, skim the diff, approve or reject. No need to dig through a pile of messages from scratch.
Solo side project at night and someone files a bug? It can hold the line for a bit.
Small jobs can run in parallel without clogging your chat window.
MIT open source. Trigger rules, Feishu / GitHub hooks: all yours to tweak.
Handy if you’re shipping a few projects alone at once.
Repo: github.com/deepseek-ai/dsh-s…
Rodrigo Belem retweeted
5090 firmware has finally been unlocked
96GB upgrades are coming.
If you're interested in GPUs and modification and you're not following Greg, you should probably change that.
He is among the only people in the United States performing these types of modifications, which is a big deal.
One of the biggest reasons that people don't order modified cards is because they are afraid they will have no support or warranty once it leaves China, which is most likely true.
Rodrigo Belem retweeted
Great line from the CTO of Proton. I’m paraphrasing and stealing it :
Your job is not to use agents to help you write code. Your job is to lead the agents to help them write code.
Rodrigo Belem retweeted
Friendly reminder to always check your disks when renting a server. Just rented a dedicated server from the Hetzner auction and both drives were degraded. Sent in a support request to replace the drives (should be free). Let's see how long it takes!
Rodrigo Belem retweeted
Someone put Mario 64 on their desk in mixed reality
A weekend build at 90 FPS, made with Claude Code plus the Unity and PICO CLIs, deploying straight to the headset.
reddit.com/r/ClaudeAI/commen…
Rodrigo Belem retweeted
token custando o dobro da folha de pagamento num time de 25
software n ficou de graça, o custo só mudou de linha na planilha
with current token prices i wouldn't really say "cost of software is going to zero"
our token cost is double our payroll which means even though we're 25 it's like we're 75
and i feel like our output is similar to a team of 75 pre-ai. we're just able to do it without all the downsides of a larger team
this is still a good thing. we couldn't hire 50 more people. so much overhead and impossible to retain a high bar
Rodrigo Belem retweeted
Have been using GLM-5.3 for some tasks that opus won’t do
Legitimately had to check and make sure I wasn’t on opus 5.5 by accident it’s that good
First time I’ve tried an OSS model and thought “wow” instead of “meh”
Rodrigo Belem retweeted
Le Chonk runs on colibri.
@MistralAI Large 4, 1T parameters, from @UnslothAI's GGUFs: text, pictures, reasoning and tools, on one 24 GB card with the experts read from NVMe. 1.34-1.36 tok/s.
colibri v2.1.0: github.com/JustVugg/colibri
Rodrigo Belem retweeted
no matter what agent you're using, your voice is the best way to talk to them
and hex is the best way to use your voice:
- fast
- runs locally
- works on mac, windows, linux
stable release now available: anoma.ly/hex
