Roberto Lucas

A guy who once wanted to be a developer, leader and mentor. Now he is.

@aialygn | @masterbotsai

[Alygn @aialygn]

The institutions that govern high-risk technology are built before the crisis, not after it.

The IAEA wasn't built after a nuclear catastrophe. It was built in the window when the stakes were understood but the disaster hadn't arrived yet.

That window is open for AI governance today.

The institutions built in this window define the terms. The ones built after inherit the compromises.

[Roberto Lucas @AndlerDev]

Google, X, Microsoft: all shortening exposure windows. All reacting to the same threat. #AIRisk

The question is: Can we coordinate institutional defenses faster than AI automates offenses?

If not, it's not just digital integrity that's compromised. It's human.

[Roberto Lucas @AndlerDev]

Google Cloud: "Never Expire sessions? Over. 16 hours max." #CyberSecurity

This isn't bureaucracy. It's a scream into the darkness.

April 2026: X faced a botnet that compromised 4.8M accounts. 18 workers. Every root password exposed.

Something broke. And we know it. 🧵

[Roberto Lucas @AndlerDev]

Traditional security assumes human attackers:
❌ Slow
❌ Need sleep
❌ Cognitive limits
❌ Manual scale

AI: #AIGovernance
✅ 24/7 w/o pause
✅ Global scale in seconds
✅ Thousands of simultaneous vectors
✅ No fatigue errors

The model is broken. And how many say it out loud?

[Roberto Lucas @AndlerDev]

@ThePrimeagen One discipline that you have per agent is a good path. Then you connect them all to pair-review with reviewers/testers agents according to main agent/your needs. That way you minimize the context length of your rules and your workload 👌

[Roberto Lucas @AndlerDev]

I was reading one of the dreams that Wobblus is having now that @openclaw has had this feature for a few days now. This, is impressive and you can tell the difference of continuity between sessions. If you keep syncing their memory and indexing, all good 😉

#AgenticAI #AI

[Roberto Lucas @AndlerDev]

PD: I recommend watching the #WestWorld series (not the last season), it is a sci-fi that has an autonomous general AI and where AI and humans are seamlessly together... 😁

[Roberto Lucas @AndlerDev]

Just watched "AI Is A Massive Problem. Here's Why." It crystallizes what many of us in #AISafety have been saying.

The issue isn't AI capability. It's #AIGovernance. Let me share why the current trajectory is not stable, and what we need to do about it 👀

#ExistentialRisk

[Roberto Lucas @AndlerDev]

No need for autonomous general AI to solve all problems. Safer paths exist. But they require restraint.

The question: Will we build governance infra before crisis, or after?

History suggests "after" rarely works.

https://t.co/PUUYcwZNF5

#ResponsibleAI #AIFuture #TechEthics

[vxdb @vxdb]

lol

[Roberto Lucas @AndlerDev]

Not even a high-end gaming PC is capable of running a good local model or to orchestrate various agents at once, lol. Hard lesson for all: not everything can run locally.. I have been there, hahaha

#OpenClaw #AgenticAI

[TheStandupPod @thestanduppod]

Why OpenClaw users buy Mac minis

[Figma @figma]

Tired: code or canvas
Wired: code AND canvas

Introducing Claude Code to Figma

[Roberto Lucas @AndlerDev]

lol that's true... Using old devices work. I have a (very) old server machine running #OpenClaw and works REALLY good

[Teng Yan @tengyanAI]

your old Android phones are a better agent server than a Mac mini. while most users spend $600+ on dedicated hardware for @openclaw, the real efficiency is hiding in your junk drawer.

i saw some developers use Termux and Node.js to turn 3 watt devices into constant research hubs. by running npm install -g clawdbot, these discarded screens handle market monitoring and Telegram summaries without a break.

3 phones can roughly match the output of a Mac mini for almost 0 cost. this setup runs Clawdbot 24/7 to pipe private signal alerts directly to a primary device.

i suspect the hardware bottleneck for autonomous agents is already dead. compute is now so cheap that our junk phones is sufficient!

[Roberto Lucas @AndlerDev]

I really love seeing experiments like https://t.co/d7d3ipHVl4. They offer a fair trait between AI and humans... And very funny btw, hahaha. It makes you think though 👀

[Roberto Lucas @AndlerDev]

Product changes are hard—especially when they’re deep. #BuildingInPublic

Using AI tools (like #Figma AI) isn’t about giving up control, but moving faster with intention.

How much do you automate your work?
And where do you draw the line so #AI doesn’t decide for you?