Professor of AI @ITAM_mx teaching at @profrodai | I teach how modern AI works and love making ML research understandable and implementable
Joined September 2015
- Tweets2.4K
- Following197
- Followers645
- Likes8.3K
AI is supposed to be liberating tech. We work less. AI does everything
Not my observations. Everyone I know sleeps less, spends more time on a device, builds more, has more side projects
Combo of dopamine hit + first 80% being really easy with AI but last 20% extremely hard
Tech skills becoming irrelevant? Coding agents bring complexity from day 1
Suddenly we're re-reading Leskovec and reasoning about dist. systems
Non-techies miss this entirely: you don't know what you don't know
Agents don't optimize, they tell us to throw hardware at problems
I tried building native macOS apps with AI. I’ve now moved to webapps.
In macOS, signing, certificates, builds, etc all added friction.
I couldn't find libraries + examples for what I wanted.
There is less open knowledge on what makes native apps good, what are good UIs, UXs
I wanted the model to look at the native app and improve it.
I ended up building a little app to take screenshots of the other app and feed it back to have a smooth loop for the agent:
Make a change.
Run it.
Inspect.
Try again.
What's impressive is that the final result is not that different from what your indie creator with Seedance can achieve today. Voices still sound extremely AI.
GenAI is a great democratizer. Tools you now have access to aren’t that different from the ones available to top labs
Watch Episode 1 of our Odyssey series made with Grok Imagine.
Let us know in the comments if you'd like a full breakdown of how it was done!
Creative collaboration with @WonderStudiosX and director Dev Walker @glumlot_
Who would have said that coding agents require so much storage?
Between caching, and just being sloppy, cloning and generating worktrees, they fill your drive so fast!
Just got a new 4 TB SSD. Let’s see how long it lasts.
All this is not helping bring prices for storage down
In the PR or not PR debate, I’m in team PR.
Code must be reviewed by other sessions and vendors.
The code that Sonnet vs Fable generate is quite different, let alone GPT vs Claude vs Gemini.
Ideally, you want them reviewing each other during PR before merging to main
Whereas for LLMs we can expect more or less the same results by now, the variance we get from image models is impressive.
I wanted @profrodai to blink; it should be super easy to generate with any model.
Or so I thought...
I prompted @recraftai & this is the result👇
AI doomerism is popular. But Europe does have top AI labs, academic & industrial.
There are credible regional players in cloud computing.
European universities are leaders in AI research.
Openclaw was invented in Europe!
Things aren’t as bad as Twitter wants us to believe
I am worried about the future in Europe right now. There are no foundation AI labs here. There are no GPU fabs here (despite having ASML) and no plans to build or upgrade any foundry with sub 7nm capabilities (despite EU CHIPS act). There is no energy policy for the power AI data centers need. There are pockets of excellence but the strategic foundations are totally missing. Doing any work at the technological frontier of tech from manufacturing to robotics to AI is dependent on Asia and the USA (try making an autonomous robot without chips and GPUs from Asia or training one without AI models and compute from USA). Why is Europe so out of the loop (building 12-28nm fabs in Dresden, WTF are we gonna do with these?)? How do we get on track?
Rod Rivera retweeted
Why does your AI agent demo beautifully and break later? It's the harness, not the model.
Starting Friday at 11 AM EDT, free weekly office hours with @Nebius_Academy and @rasa_hq builders on exactly this.
Share your questions. I will answer live on Zoom for 60 minutes.
RSVP: lu.ma/7ci6q3zp
Rod Rivera retweeted
I pulled an all-nighter to finish and release the final project for our course at Nebius Academy.
One part of the project was open-sourcing the Sovereign Agent.
You can check it out here: github.com/sovereignagents/s…
This is my first fully open-source contribution to PyPI, and I really hope others will find it useful.
Over the past few weeks, I’ve been studying how modern, always-on AI agents are built in 2026. I call these 'sovereign agents.' Think of tools like OpenClaw, Claude Code, Aider, and similar projects.
My goal was to make something simple enough to help people learn, but also practical for building real projects.
These days, agents seem overly complex and packed with dependencies. For example, OpenClaw has hundreds of thousands of lines of code, but it doesn’t have to be that complicated.
For those familiar with it, my inspiration was the fast.ai library. It was designed for learning how Deep Learning works, but many people ended up using it for their own projects.
Since this field is moving fast, I already have plans for version 2 and hope to release it in the next few days.
I’m also working on documenting all the patterns these agents use.
If you don’t want to miss any updates, please star the project and share your comments or feedback!
For selected examples, the results are impressive, but as soon as you try ChatGPT's latest image model with prompts that have less data, it breaks.
Anyone living in London will confirm that this map has many inaccuracies.
However, it is a pretty hallucination!
🤖 Made with AI
Rod Rivera retweeted
You’ve built your Rasa agent. But no matter how much you test, you can’t predict every path a real user will take. Nobody can.
That’s exactly why Arklex AI and Rasa are teaming up.
Rasa’s CALM architecture gives your agents the flexibility to understand users and the reliability to act predictably. Arksim closes the last mile, automatically running quality tests across thousands of simulated conversations to catch errors before real users ever do.
No more holding your breath at launch.
Try Arksim on your next Rasa project:
github.com/arklexai/arksim/t…
#rasa #AIagent #opensource #AIEval
The problem with naming your product after the current thing is that it ends up sounding dated fast!
Anything with "GPT" appended feels stuck in 2023.
And if you are a "Copilot," it sounds even older.
Same will happen to all those products with "Claw" in the name.
Have you heard the term "harness engineering" before? And do you have thoughts on the topic?
H.E is an emerging discipline in AI development focusing on building infrastructure, tools, and wiring necessary for AI agents to autonomously build, test, and manage software products
