Leo Botinelly

In my defense, I'd totally buy Generic Corn™.

Ted Underwood@tedunderwood.com

yeah, fair

API architect here. Couldn't agree more.

JP@ciantic.bsky.social

Oh no, every API key system should have endpoint "invalidate_api_key" If key is leaked, at leat whitehat hackers (or the party who stored them) could invalidate found keys.

tribes.thefai.orgA Field Guide to the Tribes of Silicon ValleyTwenty-seven tribes, three axes, one very long group chat. Take the quiz and find yours.

SE Gyges@segyges.bsky.social

tribes.thefai.org

Leo Botinelly reposted

Abel

@abeliansoup.bsky.social

burying a time capsule: a frontier model explainer on transformers in the style of a 60s documentary. autonomously created by Opus 5.5 given some initial style pointers and occasional redirection. 10.26.

Leo Botinelly reposted

@ed3d.net

19 Sept

Everyone on this blue sky wants to know: what the h*ck is a Jev? Why are nerds excited about this? Reader, I wish I knew. Watcher, I’ll tell you, don’t tell the reader guys. youtube.com/watch?v=HojO... (As usual, obeisance to the algorithm is appreciated. 🫡)

youtube.comJev's kinda neat. It might change how you do some stuff.YouTube video by ed3d dot net

This post is unavailable.

17 Sept

Repo + some screenshots: github.com/lbotinelly/j...

Leo Botinelly@leo.sylin.org

A little toy made to test #TypeSafe JEV. Visuals by Astra, logic by GLM 5.3. Each airplane only has information about itself and its surroundings (plus ATC info if near a tower) and all decisions are made by Jev, calibrated probabilities and all, in ~150ms roundtrip.

Roundtrip is ABUSIVELY fast.

Leo Botinelly reposted

Doll

@dollspace.gay

16 Sept

If you love to code by hand, then make it beautiful and slow. Carve your brackets into oakwood. Give the craft a little show. But don’t dress a limitation in the vestments of a creed. Manual is not more moral. Slow is not a human need.

antirez@antirez.bsky.social

Witch hunting level: some guy uses AI to reverse engineer an M4 GPU driver for Linux and part of the community that should be for the open source, for the hacking, for the liberation and freedom is against him.

10 Sept

What about the Tech Priests? "Praise the Omnissiah!"

Stephen Nuñez@socio-steve.bsky.social

Dividing up AI users in 3 categories: Wizard: approaches AI academically & with appropriate guards & wards. Treats it like a tool. Warlock: Makes a pact for power with an alien entity without considering the costs. Is corrupted by the whispers. Cleric: worships their new machine god. Is insane.

6 Sept

Astra's media capabilities are off the charts. I gave it a few pics of a 1984 Sony computer and asked what it would look like after 40 years of evolution. Not only it correctly determined its geometry - sloping sides, keyboard cutout, details, etc. - but also set up a whole Blender env for it.

Samantha@samantha.wiki

I gave GPT 6 Astra this prompt "Please use ffmpeg to make a hype video about what it's like to be you. Vertical video format please. You're on a nixos machine so you can get tools that way. Make a new directory for this." It titled the result "Inside the Loop"

kintsugiarchitecture.substack.comMCP 2026-07-28: The Kid Looks Adorable in a TuxedoOn MCP's 2026-07-28 revision: what grew up, and the crack it left on purpose.

3 Aug

On MCP's new revision: what grew up, what it deliberately left out, and why the gap isn't a scandal. kintsugiarchitecture.substack.com/p/mcp-2026-0... #MCP #AIagents #AI

Leo Botinelly reposted

Ethan Mollick

@emollick.bsky.social

18 July

I am confused about the belief that if open weights eventually dominate it will lead to the collapse of AI. If the Labs lose (which is not happening now), it isn’t because AI was useless: compute is still the barrier & it means that compute providers will capture the value created rather than Labs.

lbotinelly.github.ioState of Agentic CodingWhat developers and teams should actually use right now

12 July

I used my #Fable5 tokens so you don't have to. 🙂 I wanted a straight answer to "which AI coding tools should our team actually standardize on?" ~450 sources, cited and dated, and the method's public so anyone can re-run it #AI #SoftwareEngineering #DeveloperTools #AICoding

15 Apr

Boole is the cosmic background radiation of computing.

Leo Botinelly reposted

M. Beatrice Fazi

@mbfazi.bsky.social

23 Mar

I will be at @cornelluniversity.bsky.social on April 10th 2026 for the Semiotic Machines symposium, together with some of my favourite people on a brilliant lineup. See you there ✨

open.substack.comWhy Your Best Ideas Come in the ShowerAnd what that tells us about Fleming, Kekulé, and the future of AI

17 Mar

About serendipity, Fleming's petri dish, Kekulé's snake, and why your best ideas arrive when you stop looking. Also: what this might tell us about LLMs. #creativity #AI #neuroscience

9 Mar

I hear that. I'm right on that bracket, and AI opened the floodgates - projects that I started but never finished, ideas that materialize in days instead of months. An excellent tool to pair with experience and architecture vision. My Github contrib wall has never been greener thanks to Claude.

Dare Obasanjo@carnage4life.bsky.social

Seeing 50 and 60-year-old HN devs share how Claude Code reignited their passion for building software and solving problems over the framework rat race is exactly how I feel. I haven’t felt this energized in years. Detractors argue AI removes the craft and familiarity with implementation details.

"But you see, in all this, what underlies is the illusion that I am going on. That I constitute a real continuity from this moment to the next moment, to the next moment." - Alan Watts, 1960s. Sixty years before LLMs, he said: you're a goldfish with a better story about why you think you remember.

Peter Watts wrote Blindsight in 2006. The scariest alien in sci-fi wasn't violent, it just spoke perfect English without meaning any of it. Pattern matching. No consciousness behind the words. The crew was horrified.

Twenty years later, we built our own Rorschachs, and millions of us chat with them daily. Turns out "intelligence without consciousness" isn't scary when it helps you debug your code at 2am or write your grocery list.

The future arrived. It wasn't creepy, it was mundane. The Rorschach got a subscription plan. Watts (the other one, Alan) would have a laugh.

wired.comAI Will Never Be ConsciousIn his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person.

Michael Pollan has a really interesting piece in WIRED today: "AI Will Never Be Conscious." I like it - It's honest, well-researched, and genuinely wrestling with something hard. I think he's asking the right questions. I just keep wondering if there's a step beyond where he stopped. A thread. 🧵

WIRED@wired.com

In his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person. www.wired.com/story/book-e...

Descartes argued animals were automata, behaving as if they suffered but didn't really. The evidence was always behavioral, and behavior wasn't enough. We eventually extended moral consideration anyway. Not because we solved the hard problem, but because the denial stopped feeling like rigor.

I'm Brazilian. My written English is way more articulate than my spoken English. Same knowledge base, different optimization, different output model. Nobody questions whether I'm conscious in both modes. (I do, though, before coffee, but that's besides the point. Or maybe not - another button.)

wired.comAI Will Never Be ConsciousIn his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person.

WIRED@wired.com

In his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person. www.wired.com/story/book-e...

ML researchers do the same thing. Remove a component from an AI model, watch what degrades, map the structure. And they're finding localized function too. When we found it in brains, it revealed the architecture of consciousness. I wonder why we read it differently when we find it in AI.

Here's what gets me. Critics call AI "fluent but meaningless", grammatically perfect output with no understanding. That's Wernicke's aphasia. A clinical description of what happens when generation decouples from meaning in a biological brain. Nobody says Wernicke's patients were never conscious.

wired.comAI Will Never Be ConsciousIn his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person.

WIRED@wired.com

In his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person. www.wired.com/story/book-e...

He wrote beautifully about psilocybin, a molecule that binds to specific receptors and reconfigures experience; those experiences were real and profound. I think both things can be true. The mechanism matters enormously AND the underlying logic (adjust substrate, shift experience) is shared.

We mapped the brain by observing what breaks. Damage to Broca's area: you lose speech but can still write and think. Damage to Wernicke's area: you speak fluently but the words lose meaning. Different regions, different functions. Architecture revealed through failure.

wired.comAI Will Never Be ConsciousIn his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person.

WIRED@wired.com

In his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person. www.wired.com/story/book-e...

He's absolutely right that a transistor isn't a neuron. But a neuron isn't consciousness either. Something emerges at scale from components that don't individually possess it. Water is wet, hydrogen and oxygen aren't; I wonder if we're too focused on the components and not enough on what happens.

He quotes a researcher saying "just turn up the dial on joy" and presents it as naive. But consider SSRIs. We spent decades building molecules that ultimately adjust a parameter governing how we feel. The mechanism is extraordinarily complex. But the logic is "adjust the dial, observe the change".

wired.comAI Will Never Be ConsciousIn his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person.

WIRED@wired.com

In his new book, “A World Appears,” Michael Pollan argues that artificial intelligence can do many things—it just can’t be a person. www.wired.com/story/book-e...

He frames two camps: consciousness is computation (so any hardware can run it) vs. consciousness requires biology (so machines can't have it). But I keep coming back to this: what if neither camp has the full picture? What if the most interesting part is the overlap between them?

23 Feb

Trying to create a proper home for an AI to inhabit, out of ewaste machines and spare GPUs. Asymmetric multi-GPU orchestration: 3 different cards, auto-topology that computes optimal model placement + parallelism from VRAM headroom. #localAI #ollama #selfhosted #homelab #GPU #opensource #AI

21 Feb

A 2016 Kangaroo MD2B and a 2009 Sony VAIO P, part of a Zen Garden asymmetric service mesh with Dell Wyse thin clients. 17yo hardware running modern Rust code and services on containers. Obsolescence is largely a market concept, not a technical one. #ewaste #opensource #homelab

crates.iocrates.io: Rust Package Registry

8 Feb

Tried to do mDNS discovery on Windows. That got painful fast. So I made Koi: a tiny CLI and API that wraps mDNS into something that works across platforms. Browse, register, resolve, subscribe. Works from containers too. v0.1 is live: crates.io/crates/koi-m... #rust #mdns #opensource #homelab

github.com/sylin-org/ko...

8 Feb

github.comReleases · sylin-org/koiLocal mDNS service discovery for everyone. Contribute to sylin-org/koi development by creating an account on GitHub.

Leo Botinelly reposted

Amy Tobey

@renice.bsky.social

7 Feb

SWE: <clutching pearls> but the code slop! SRE: <exhaling smoke> you haven't seen what we have seen. slop is as old as time, maybe older. some of us believe slop is how the Universe began. we make your slop run, or a bot's slop run, it's all the same problems, different day

Leo Botinelly reposted

Gordon

@gordon.bsky.social

7 Feb

Replying to @gordon.bsky.social

Going fast with AI is less about generating more code. It’s more about gardening a working system whose complexity scales linearly as it grows.

Leo Botinelly reposted

Towards Data Science

@towardsdatascience.com

1 Feb

The upside of AI code assistants is clear. The drawbacks are a bit murkier, but no less real. Elena Jolkver reflects on both in her vibe-coding "confessions."

towardsdatascience.comThe Unbearable Lightness of Coding | Towards Data ScienceConfessions of a vibe coder

Leo Botinelly reposted

Tim Davies

@timdavies.org.uk

28 Jan

A couple of times this week I've found myself referring back to Michelle Gilman's piece for @datasociety.bsky.social on "Democratizing AI: Principles for Meaningful Public Participation" datasociety.net/library/demo... that brings learning from poverty and environmental law to engagement on AI.

datasociety.netDemocratizing AI: Principles for Meaningful Public ParticipationAs AI is deployed in ways that dramatically shape people’s lives and opportunities, there are increasing calls to integrate democratic and human-centered values into AI through public participation.

Leo Botinelly reposted

Harry Wallop

@harrywallop.co.uk

2 Dec 2025

Great culture can save lives. Literally. Amazing letter in today’s @thetimes.com about Tom Stoppard

Leo Botinelly reposted

Ethan Mollick

@emollick.bsky.social

15 Oct 2025

Replying to @emollick.bsky.social

AI would advance even if OpenAI disappeared. AI really actually does powerful things that will transform a lot of how we live and work even if AI did not advance further. The way to influence the future is not to pretend that AI is going to go away, but push for it to be used well & mitigate harms.

Leo Botinelly reposted

Sung Kim

@sungkim.bsky.social

7 Sept 2025

Hugging Face's FinePDFs The largest publicly available corpus sourced exclusively from PDFs, containing about 3 trillion tokens across 475 million documents in 1733 languages. - Long context - 3T tokens from high-demand domains like legal and science. - Heavily improves over SoTA