Encyclopedia Autonomica

A couple of weeks ago, I picked up GTA IV on Steam for a handful of dollars.

No such thing as bad student, only bad teacher.” — Mr.

In my recent post, I introduced DeepMind’s work on Code World Models and contrasted it with Yann LeCun’s Joint-Embedding Predictive Architecture.

Yann LeCun has argued for years that next-token prediction is insufficient for human-level intelligence and that AI needs world models that learn how the world works.

Jevons Paradox is causing your model bills to grow faster than anyone planned, buried in infrastructure invoices, and distributed across dozens of teams with no coherent view of what they’re buying or why.

“Recursion: see Recursion.” — Dictionary definition joke “To iterate is human, to recurse divine.” — L.

“The feedback loop, not the function, is the unit of intelligence.”

Silly Silicon Valley Executives pride themselves on creating leaderboards who burns the most tokens (tokenmaxxing).

It's been too long since I've written about space.

“For the things we have to learn before we can do them, we learn by doing them… we become just by doing just acts, temperate by doing temperate acts, brave by doing brave acts.”

Let me start by saying something that might be seen as provocative.

There has been much frenzied chatter in the agent community about Moltbot / OpenClaw since it was launched a couple of days ago.

“Joe ain’t vicious, you understand.

Happy New Year!

A key challenge when equipping agents with a large number of tools is that the agent often cannot reliably determine which tool to choose, especially when the tool set is large.

It always fascinated me how LaForge in Star Trek: TNG could easily recalibrate the Warp Drive using only a selection of standard interfaces, a highly efficient “computer’, and a small selection of hand tools.

I have built a lot of software in my life.

I have never been quite satisfied with Retrieval-Augmented Generation (RAG).

Continuing from Post 1 in my series on modular financial agent systems, I finally found time to work on the second part of the series.

For the last quarter of a century, we have developed AI solutions for low-risk environments like recommendation engines, content moderation, or ad-targeting.