thingsai.io
AI Things. Human Things. What Builds What.
This is where I share what I'm learning and discovering:
- Ideas – exploring new concepts in AI systems
- Memos – distilling insights from research and literature
- Essays – examining longer-term questions
- Blog posts – capturing hands-on experiences
Think of it as a working notebook made public—practical, research-grounded, and opinionated.
Featured from the Feed
L'enfer c'est les autres — when AI iteration replaces co-design
The trap of the inner loop that flatters before it isolates
I am not alone anymore, I have an Agent, and it likes all my ideas. Yes, within an hour I can have a fully formed spec, implementation plan, code, and deployed solution. But every hour of solo iteration with AI is a deposit into my own conviction and a withdrawal from my capacity to listen.
Pitrat on AI Research (1997)
Building Solutions to Understand Problems
While preparing my HDR mémoire in 2013, I discovered Jacques Pitrat’s 1997 paper “Comment faire de la recherche en intelligence artificielle” buried in the LAFORIA archives (now LIP6). It was a .doc file from another era of AI research. I’m grateful I saved a copy—it seems lost to the internet now, yet its insights still feel relevant to me.
Zen and the Art of Vibe-Coding
Is there still room for code-templation?
How AI-assisted coding eliminates the contemplative pauses between architecture and implementation—and whether beginners can still build expertise without that rhythm.
Computational Thinking
Frameworks from research and expert practice
Computational Thinking encompasses the practices and knowledge involved in designing computer technologies. It bridges two domains: pedagogy (how we teach programming and computer science) and expert practice (how experienced developers actually design software).
Agent Coin
Autonomous AI Economic Transactions
AI agents need to transact—research agents buying API access, creative agents selling content, agents grounded in expert content monetizing their expertise—but every payment requires manual configuration and human approval. Give agents wallets so they can pay each other directly, autonomously. Coinbase’s Payments MCP could already provide the infrastructure.