Fred Oliveira 🧠

Most things that cause me anxiety are coordination problems.

@BardIonson @lisbonai_ I think we would! Let me take this to the rest of the team! Can you email me at fred at lisbonai dot org? Thank you!

@f @lisbonai_ Would you be interested in displaying Ai art as a part of the conference?

we're a few months away from this year's edition of lisbonAI. Awesome location and we're shaping up a sick speaker lineup this year.

@Pze_pos @ericzakariasson you get inside one of these teslers and everything's battery

@ericzakariasson Ask about the battery. Ask About The Battery. ASK ABOUT THE BATTERY

wow i didnt know they made cyberlimos like this

@LukeberryPi é o que dá estar a ficar velho, velho.

e esse maluco tendo um user de apenas uma letra

@shadcn @theo ghostty (or other libghostty-powered terminals) are super fast. That's what I'd recommend.

@theo hmm maybe I need to see to compare. what are you using?

got called out for still being on iTerm2. never found a reason to switch tho. what am i missing

@antirez to keep you sharp, of course

Why macbooks have those blade-alike borders?

if your website has a long animation before it shows anything useful, I am guaranteed to not visit more than once.

@badlogicgames @mitsuhiko yesterday I was reading @deanwball argue against a bunch of people on what the definition of thinking is. We are too quick to cling to absolutes because we hate not knowing shit. The thread: x.com/credenzaclear2…

that's the point! I can make no such claim. But I can mention the uncertainty! Just like we can't provide substantive evidence for what makes our consciousness emerge out of brain activity (go ask global workspace theorists about how they feel about everyone else's ideas), we can't point to the matrices and say "nothing there! just maths!" because with us, we could (absurdly, of course) reduce it to "nothing there! just electricity!"

recommended reading. because today the "model wellbeing" shit came up, and i reduced it to "thems just matmuls". that's obv. an oversimplification to get a point across. personally think sentience is likely substrate independent. our current models' substrate is just insufficoent. a model on its own is not concious/sentient under popular theories/frameworks of conciousness. not because its a big matmul machine. but because it lacks things like continuity/"state", self-maintenance, believe consolidation, embodiement, grounding, feedback loops, etc. pp. we can close some of these gaps. e.g. a model+transcript, into which it can persist current believe, and which establishes a feedback loop does tick some (insufficient) boxes on the way to machine sentience. give the model sensors and you tick more boxes, etc. but there are a ton more boxes that need to be ticked, before we can start worrying about "model wellbeing". bostonreview.net/articles/could…

@mitsuhiko @badlogicgames your post drew too hard a line. In general, I think the tool framing helps ground people, and the psychosis is real out there. But it seems that you are at a 0% of there being anything there there, and I don't think we can be that sure.

@mitsuhiko @badlogicgames that post yesterday was one of the first times where I clearly disagreed with your framing on something. Not that I believe current models are conscious, but because I believe we're already somewhere between "just a clanker" and 2022's "this model is sentient"

@v36372 @badlogicgames you'll be running more models on a 128gb m5, of course. Now, lots of people don't have a particular desire to run local models. If you do, can't go wrong with more memory to stuff them in.

@badlogicgames with current pace of local llm development, do you recommend m5 max 128gb or a 24~32gb vram gpu build

worst part about setting up a new mac

@liran_tal @jxnlco might have recommendations

Who from OpenAI can talk and share about developing securely with the Codex app? I'm referring to devcontainers vs remote environments etc

"a guy" sure, Eric, sure.

@whoiskatrin @badlogicgames @FrameworkPuter @antirez I set up a new m5 max yesterday in around 10 minutes. Scripting everything is the gift that keeps on giving: github.com/fredoliveira/d…

@badlogicgames @FrameworkPuter @antirez i love the feeling of a new laptop but i have to set it up

well, hello there! i haven't had time to setup my @FrameworkPuter yet. hopefully i will tomorrow, then i have two beautiful machines to go full local inference goodness, joining @antirez in his quest with ds4c. this cost more than my first 3 cars combined ...

@chrija an absolute no-brainer. Always make an exception for awesome.

Why We Invested in Anthropic medium.com/point-nine-new…

interpretability work is up there with the most beautiful of pursuits

@HaraldBregu @antirez disagreed on 2 main points: - obviously you can ship interesting things without AI. Will you ship them as fast as the market wants, though? - people will lose their jobs to AI. Lots of things that are human-led today will be unreasonable to not do with AI.

@antirez I don't think people will lose their jobs because of AI. They will mostly lose their jobs because of a lack of skills. AI is just a tool that, if you don't use it, you don't ship anything relevant.

About the conversation around automatic programming, how useful or useless it is. Well, for me it is quite simple: you need to be very good to multiply your output. Retaining quality is harder now, but if you manage you are accelerated 100x. Instead of fighting I enjoy the race.

@elderbieler @future_yas you think those 1m FSD subscriptions are concentrated in places where the regulation is good to go? And that if not, that they will be by EOY?

One year ago today [May 20, 2025] Elon Musk was interviewed by CNBC where he predicted Tesla would have a thousand robotaxis in a few months and by end of 2026 there will be hundreds of thousands if not over a million Teslas with unsupervised FSD (robotaxi + customer Teslas). Today [May 20, 2026] there are a total of 39 unsupervised robotaxis operating only during daylight hours in Austin, Dallas, and Houston. There are 0 customer Teslas with unsupervised FSD. And Tesla still hasn't filed the permit to begin testing driverless cars in California.

@HenrM_Cruz @catalinmpit @patife @luismbat @pcbo @davidgomes now that's some good company

@catalinmpit 👋 also: @patife , @luismbat , @f , @pcbo , @davidgomes

I'll create a list of EU people that work in tech. If you want to be added, reply.