Maurice Nsabimana

Happy being, inner engineer, #quantum neurohacker & #data #nerd ≡ #AI #compassion #ExO #globaldev #policy #shibumi ▌RT: ≠ Endorsement

[Maurice Nsabimana @mutabazi]

The future of work depends on better data.

📢 Call for papers for the Better Data for Better Jobs and Lives conference (Dec 8–9, DC & Online).

Submit by Oct 31: [email protected]

🔗 More info: https://t.co/dQ1yVfUrx2

#DataForDevelopment #SurveyMethods #AIForGood

[Ansh Bhatnagar @iansh04_]

Claude just dropped 13 FREE AI courses (with certificates).

No $500 course needed.

No “guru” required.

Just real skills — straight from Anthropic.

Here’s the full list:

👇

1. Claude 101
  https://t.co/JahbpNxCuF

2. AI Fluency: Frameworks & Foundations
  https://t.co/W0woGjqWxz

3. Introduction to Agent Skills
  https://t.co/QTkHwO49YI

4. Building with the Claude API
  https://t.co/Jr8iVHSODg

5. Claude Code in Action
  https://t.co/4sIUOan0zs

6. Introduction to Model Context Protocol
  https://t.co/H6NKuic8i9

7. MCP: Advanced Topics
  https://t.co/LSaF6Do778

8. AI Fluency for Students
  https://t.co/0UtkUWAGS5

9. AI Fluency for Educators
  https://t.co/wgQ4qXlnIa

10. Teaching AI Fluency
  https://t.co/4fiHBRIXpe

11. AI Fluency for Nonprofits
  https://t.co/0A363Ms47f

12. Claude with Amazon Bedrock
  https://t.co/xMqBPnjoB5

13. Claude with Google Vertex AI
  https://t.co/It5xFtjmUj

—

If you go through even HALF of these…

You’ll be ahead of 95% of people using AI.

Most people won’t.

Because they’re still:

• Watching random YouTube videos
• Buying overpriced courses
• “Learning AI” without actually building

Don’t be that person.

Do this instead:

1. Save this post (you’ll come back to it)
2. Pick 1 course → start today
3. Share it with someone who needs this

Free. Practical. No excuses.

[Maurice Nsabimana @mutabazi]

AI applications have become mission-critical in recent years. Why isn't the engineering that drives them as well?

https://t.co/euAYX1F9oJ

[Maurice Nsabimana @mutabazi]

@bradmenezes @sparkcapital @kleinerperkins @MeritechCapital Superblocks

[Brad Menezes @bradmenezes]

I’m excited to introduce Clark, the first AI Agent to build internal enterprise apps.

We’ve raised $60M, including a fresh $23M from @sparkcapital @kleinerperkins, @MeritechCapital and Greenoaks.

Unlike consumer vibe coding tools like Lovable, Replit and Bolt that only generate prototypes, Clark builds production-ready internal apps — enforcing your enterprise standards:

🧩 UIs generated using your design system
🔌 Integrations with private APIs, databases and SaaS apps
🔐 Permissions mapped to @okta & @azuread groups
🛡️ Security with audit logging, secrets management, and vulnerability scans

Clark is designed to operate exactly like a human internal tools team. It's built on a state-of-the-art multi-agent architecture, emulating your Designer, IT admin, Engineer, Security Operations, and QA employees.

When Clark generates an application, you can modify it in 3 ways:

1. Natural language - talk to Clark
2. Visual – Edit it like in Figma
3. Code – Use your IDE like Cursor or VSCode

Global enterprises in regulated industries like @Instacart (CART), @carrier (CARR), and @cvent (Blackstone) already run their mission-critical apps on our platform.

Our customers save $5m on average. Book a demo and we'll save you $5m too. And we're so confident that if we can't, we'll donate $5,000 to a charity of your choice:

Book a Demo – https://t.co/oG5LaD10V9

---------

As part of this launch, we’re giving away the system prompts of leading AI products like Cursor, Manus, and Codex.

These 6,000 line system prompts have enabled them to become billion-dollar companies on top of foundation models.

Retweet this post and comment ‘Superblocks’ below, and we'll send you the link so you can engineer world class prompts yourself.

👇 See Clark in action in the thread below

[Maurice Nsabimana @mutabazi]

What happens when China sends a message without saying anything... 💥
A silent move just roiled trade and the U.S. felt it.
🎯 Watch the moment that changed the tone: https://t.co/XiZVTFpkiJ #Geopolitics #ChinaUSA #TradePower

[International Institute of Holistic Health @i2h2wellness]

What do @realNDWalsch, @DrJoeDispenza & @EckhartTolle have in common? 🧠🌌
Turns out… #QuantumPhysics might be the missing link.

Ever wondered how science and spirituality collide? This one's for you.
🎥 Watch now: https://t.co/3Ice6FZ1N9
#Consciousness #SpiritualAwakening

[Maurice Nsabimana @mutabazi]

Everyone's chasing logic. But what if *intuition* is the secret edge? 🧠✨🦋
This video explores the symbols life throws at us—if we just slow down enough to see them.

🎥Watch here: https://t.co/98xknEcCA7 #Intuition #Symbolism

[Maurice Nsabimana @mutabazi]

Eliminating #USIP is like banning vaccines: short-term politics, long-term disaster. Peace takes work. And funding.
#PublicService #GlobalSecurity #VeteranVoices https://t.co/mhJdUP1Xnx

[UNStats @UNStats]

Join us @UNStats event to honor the founders of the ICP, celebrate its 50+ year legacy of global economic measurement & look to the future with global, regional, and national experts. #purchasingpowerparitydata #ICPPPP #UN56SC @worldbankdata. Register ➡
https://t.co/RTbt7zKWig

[Jim Fan @DrJimFan]

We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive - truly open, frontier research that empowers all. It makes no sense. The most entertaining outcome is the most likely.

DeepSeek-R1 not only open-sources a barrage of models but also spills all the training secrets. They are perhaps the first OSS project that shows major, sustained growth of an RL flywheel.

Impact can be done by "ASI achieved internally" or mythical names like "Project Strawberry".
Impact can also be done by simply dumping the raw algorithms and matplotlib learning curves.

I'm reading the paper:

> Purely driven by RL, no SFT at all ("cold start"). Reminiscent of AlphaZero - master Go, Shogi, and Chess from scratch, without imitating human grandmaster moves first. This is the most significant takeaway from the paper.
> Use groundtruth rewards computed by hardcoded rules. Avoid any learned reward models that RL can easily hack against.
> Thinking time of the model steadily increases as training proceeds - this is not pre-programmed, but an emergent property!
> Emergence of self-reflection and exploration behaviors.
> GRPO instead of PPO: it removes the critic net from PPO and uses the average reward of multiple samples instead. Simple method to reduce memory use. Note that GRPO was also invented by DeepSeek in Feb 2024 ... what a cracked team.

[Maurice Nsabimana @mutabazi]

For the first time in the history of #meditation studies, scientists have discovered something profound happening inside the body when a simple formula of brain and heart #coherence is practiced #source #drjoedispenza #mindfulness #quantumhealing #mindbodyconnection #mentalhealth

[Maurice Nsabimana @mutabazi]

Discover the Mind’s Extraordinary Potential to Influence the Body. Powered by #InnerScienceResearch, this powerful film highlights the incredible potential that lives within us and the work of @drjoedispenza, a global community, and ongoing research.
https://t.co/rDiaX6zl5z

[Maurice Nsabimana @mutabazi]

🌟 Simplify and scale content production using advanced AI models like Stable Diffusion and Flux directly in the browser with MimicPC! #TechMadeSimple #CloudInnovation https://t.co/8FiPlc6xbn

[Maurice Nsabimana @mutabazi]

✨ Experience the transformative power of the Sacred Blue Lily. From ancient rituals to modern wellness, this flower’s gentle mood lift and deep relaxation are legendary. #SacredBlueLily #SpiritualWellness #Relaxation

https://t.co/vo9BCu5wKI

[Yann LeCun @ylecun]

GREAT IMMIGRANTS * GREAT AMERICANS

"Every Fourth of July, Carnegie Corporation of New York celebrates remarkable Americans — all naturalized citizens — who have enriched and strengthened our democracy through their actions and contributions. The 2024 Class of Great Immigrants is comprised of citizens from 16 countries. "

I'm very honored to be among them this year, along with Jensen Huang, Esther Duflo, Tino Cuéllar, and others.

I am sad to see immigrants being vilified in the current election campaigns in my two home countries, France and the US.
https://t.co/Evbew7zVpw

[Microsoft Research @MSFTResearch]

Researchers at Microsoft emphasize the importance of globally equitable AI. Join us at Research Forum on June 4 to learn more about data representation, the latest foundational advances, and how inclusive AI creates impact.
Register now.

[Yann LeCun @ylecun]

To qualify as Science a piece of research must be correct and reproducible.
To be correct and reproducible, it must be described in sufficient details in a publication.
To be 'published' (to receive a seal of approval) the publication must be checked for correctness by reviewers.
To be reproduced, the publication must be widely available to the community and sufficiently interesting.

If you do research and don't publish, it's not Science.
Without peer review and reproducibility, chances are your methodology was flawed and you fooled yourself into thinking you did something great.
No one will ever hear about your work.
No one will pick it up and build on top of it.
No one will build new technology and products with it.
Your work will have been in vain.
You'll die bitter and forgotten.

If you never published your research but somehow developed it into a product, you might die rich.
But you'll still be a bit bitter and largely forgotten.

[Maurice Nsabimana @mutabazi]

Discover the World’s Economic Pulse! 💼 New ICP #data reveals pivotal information on the #GlobalEconomy and #LivingStandards. Find out where your country stands in our international ranking.

👉🏾 Key findings from new #ICPPPP data & interactive #dataviz: https://t.co/u5VyJsnb3k

[Andrej Karpathy @karpathy]

# On the "hallucination problem"

I always struggle a bit with I'm asked about the "hallucination problem" in LLMs. Because, in some sense, hallucination is all LLMs do. They are dream machines.

We direct their dreams with prompts. The prompts start the dream, and based on the LLM's hazy recollection of its training documents, most of the time the result goes someplace useful.

It's only when the dreams go into deemed factually incorrect territory that we label it a "hallucination". It looks like a bug, but it's just the LLM doing what it always does.

At the other end of the extreme consider a search engine. It takes the prompt and just returns one of the most similar "training documents" it has in its database, verbatim. You could say that this search engine has a "creativity problem" - it will never respond with something new. An LLM is 100% dreaming and has the hallucination problem. A search engine is 0% dreaming and has the creativity problem.

All that said, I realize that what people *actually* mean is they don't want an LLM Assistant (a product like ChatGPT etc.) to hallucinate. An LLM Assistant is a lot more complex system than just the LLM itself, even if one is at the heart of it. There are many ways to mitigate hallcuinations in these systems - using Retrieval Augmented Generation (RAG) to more strongly anchor the dreams in real data through in-context learning is maybe the most common one. Disagreements between multiple samples, reflection, verification chains. Decoding uncertainty from activations. Tool use. All an active and very interesting areas of research.

TLDR I know I'm being super pedantic but the LLM has no "hallucination problem". Hallucination is not a bug, it is LLM's greatest feature. The LLM Assistant has a hallucination problem, and we should fix it.

</rant> Okay I feel much better now :)