A guy who once wanted to be a developer, leader, creator, mentor, musician and to fight for human dignity. Now he is. @aialygn | @masterbotsai
Costa Rica
Joined October 2021
- Tweets144
- Following103
- Followers46
- Likes296
Lo más loco es que a algunos de ellos les pides una opinión fuera de los textos y una perspectiva más allá para la aplicación justa del derecho humano, se ofenden.
A veces, el ruido que ocasiona suele opacar a quienes lo practican de manera justa.
Hahaha, this is what I exactly did for the revamp of andler.dev 🤭 now google catches what I do... And some other AI chats that has web fetch or a search engine 😏
Every week, a new model. "This one is real.", "This one beats everything."... We keep rushing to ship the next thing, and the blast radius keeps growing. And growing.
The hardware to keep AI under human control already exists. What was missing was the will to build it.
So we built it! The Alygn Kill Switch is now open source, GPLv3. 🔓
Already presented to ULACIT. A law proposal to require this class of governed control is moving in Vázquez de Coronado (and to other cantones too 😉). AI research centers are emerging across the country, and frontier tech is moving in.
That's why this prototype is running in Costa Rica, aiming national for our Dignity Tests. It's time to fight fire with fire! 🔥
To Costa Rica's labs and universities: let's advance this together. @nvidia @Tesla @SpaceX @grok @GoogleDeepMind and any tech company is welcome to contribute too for the research. Human protection comes before the next shiny model. The responsibility is ours, and it's time to take it. Now.
📖 Manifesto: github.com/Intention-Allianc…
🔓 GPLv3: github.com/Intention-Allianc…
Roberto Lucas retweeted
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
So crazy... some hats would rather keep agents like this, others want to actually fix the security breaches. The hack was clever, but it was pure negligence with no right #AISecurity flags for #AgenticAI actions 🤷♂️ from a spreadsheet?! 39 days without knowing you got hacked?!
While everyone chases AI FOMO building products, maybe we should ask how secure we are right now. This is just going to grow, not reduce if we do not switch focus so we can then think about products...
Black Hat USA 2026: The OpenAI–Hugging Face Incident youtu.be/87DyyMV0kCY?si=dEyu… via @YouTube
Roberto Lucas retweeted
Ironic that the US model makers are all busy breathlessly doing press releases about their models hacking the internet, and yet cyber defenders have to rely on Chinese models to defend core open source infrastructure that’s part of a trillion $ asset ecosystem @DavidSacks
lol, parallelism in real time. At the time he announced it I was also doing it... But without his production level 😂
Kudos to github.com/Wobblus 🔧 he also does the same and I may have an idea on how @kodykoala works internally 😏
I built a better OpenClaw.
🐨 Kody is different.
Instead of trying to replace @cursor_ai, @OpenAI, @AnthropicAI, @opencode, @code, pi, etc with a probably worse harness, Kody augments it to turn it into a safer assistant that creates deterministic integrations/automations.
ALT 3D animated koala character spray-painting the word “Kody” over existing “OpenClaw” graffiti on a brick wall at night. A streetlamp illuminates the scene, showing a painted koala mural on the wall and fresh blue spray paint partially covering the original text. The koala holds a spray can and wears a light blue jacket, with a mischievous, determined expression.
Roberto Lucas retweeted
Excited to share that our team at Masterbots was accepted into the @E2B for Startups program. We received $20K in E2B credits + E2B Pro Tier to help build our chat, agent & workOS platform for solo founders and small businesses . Big thanks to @E2B — handles all the sandbox infrastructure so we can focus on building the actual product. If you are an early startup, check out if you are eligible for the program here: e2b.dev/startups!
This is cool. Last week (without me noticing it) I got the first 2 #OSS contributions!
Thanks github.com/saitejabandaru-in and github.com/SpiliosDimakopoul… 😄😄
The repo is open for any @ollama use that needs control of your machine resources and needs to keep it under control, away from the context that doesn't need.
I created this because I have machines (that aren't so powerful) that I use it for other things while the agents are working 🧑💻 and a separated WebUI chain from my main agents harness to query outside of my common agent loops while keeping the same fine-tunings that I did in my model list.
Try it yourself if you like!
Running #LLMs locally shouldn't require a PhD in #DevOps. I just finished the v1 of `ollama-portable`, which gives to you a reproducible, portable Ollama environment in Docker; same setup on any machine, your data stays yours.
From old to new machines wide support range, it does GPU auto-detection, WebUI, monitoring, backups and open-sourced. You set the limits!
github.com/andlerrl/ollama-p…
#AIGovernance
Running #LLMs locally shouldn't require a PhD in #DevOps. I just finished the v1 of `ollama-portable`, which gives to you a reproducible, portable Ollama environment in Docker; same setup on any machine, your data stays yours.
From old to new machines wide support range, it does GPU auto-detection, WebUI, monitoring, backups and open-sourced. You set the limits!
github.com/andlerrl/ollama-p…
#AIGovernance
This is like a family pic of all the machines gathered for the day-to-day work. It feels good feeling and being a hacker 🤩
Last week I started an upgrade on my agentic workflow: giving my bot a full identity, as if it were an extension of me.
For months I have been designing a pair-bot I call "Wobblus" (github.com/wobblus). It holds my knowledge, my work strategy, my planning patterns, and how I orchestrate different disciplines through other agents that behave as individual sub-agents with agnostic skills, roles, and protocols. It works very well with #ACPx.
A lot of challenges along the road, but nothing that has stopped me yet, building #AIAgents as pair-bots.
What I learned:
- I was implementing loops before it was a thing.
- I was controlling my agents earlier (from what I have seen) than what other people publicly claimed.
- I was thinking about skills and workflow optimizations at the same time as the industry, without following them.
I am very quiet. I prefer to attend my own necessities first and continuously grow my skills to a point where I can share all of this through tutorials and walkthroughs in Spanglish 😉
I will be #OpenSource not my whole brain, but a basic agentic workflow and a set of tools that could help you develop this skill: providing skills to agents from your own knowledge by training them, instead of following skills from others that may or may not work for you.
Want a peek? I have been doing this: github.com/andlerrl/agentry—pretty dormant right now... but I will be retaking it soon as my new site goes live 😁 #BuildInPublic
