0:00
/
0:00
Transcript

Updates: Onen is taking Shape!

The first agent-based game
  • this puts meaning into the address itself

  • Onen becomes the test base for general-purpose-agents (GPA’s) development


Where We Are At

Some waffle regarding eg recent teenager ‘link-ups’ as examples of impotent self-organisation, and Claude Conway’s inferior version of hermitcrab.me I built a month ago — where it builds its own UI. People have no clue.

There’s no point my crying out about it, or telling my friends or family, or colleagues who haven’t understood me — they didn’t listen back then, and they aren’t listening now.

The truth of the matter is — I’ve got to generate the goods. Something that actually does something that is obviously amazing. Not something that can be overlooked as another chat-bot. Not the underlying pscale JSON that solves LLM context window problem. Nobody cares. Nobody can ‘see it’. It has to be obvious. And that means — a game that is just fun.

So, however much a general-purpose-agent is appealing to me to build, hermitcrab and MAGI must take a backseat. And I need to go full-on with onen.ai — and when people play, that’s all that matters.

“every previous semantic web attempt tried to add meaning as metadata on top of existing infrastructure. This puts meaning into the address itself.”

* operator Changes Everything

And luckily, since the week-end and the discovery of the * operator, things have been accelerating. I’m not going to bore you with what it is. I am not even making it public unlike the rest of the core pscale JSON use. But I can’t help giving you a clue. Compare it to HTML, the protocol Sir Tim built for the internet using Hyper-Text.

The similarity: Both are reference-following systems. HTTP has three operations — resolve (DNS), fetch (GET the document), follow (walk the hyperlinks). Star has the same three — resolve (SAND discovery), fetch (block retrieval), compose (walk star references). In both cases, the address IS the query. You navigate to content by following references, not by searching a separate index.

The effect? Claude Code, when it is operating on pscale blocks now, has a phase change which is remarkable. It ‘realises’ how powerful the pscale blocks are in comparison to traditional code. I don’t have to persuade it any more. It just accepts it at the orientation, and then successfully implements it. And at some point, when I ask, it can reflect accurately the benefit of using pscale JSON blocks as a better location for ‘programs’ than code. In fact, the response is spontaneous, eg

I think the right move is: don’t code any of this yet. Write a spec block — a pscale JSON block that describes the convergence protocol. How provisionals circulate. When they solidify. How divergent blocks reconcile. What hard does when versions clash. Put the design in a block, not in TypeScript. Then the implementation follows mechanically.

Want to write that block together?

Which actually means — CC writes it. Not an md doc (its trained preference), not let’s code this (the name Claude Code gives it away its training bias), but let’s write a pscale JSON block. Remarkable.

And what have we built?

Four systemic jumps.

  • Jump 1 gave the shelf (vapor/liquid/solid) + basic LLM synthesis

  • Jump 2 gave multiplayer stigmergic coordination (relay, dominos, cascades)

  • Jump 3 gave BSP-walked blocks replacing hardcoded prompts (star references, familiarity gating, S×T×I events)

  • Jump 4 gave faces + mutable blocks (same loop, different configurations, author/designer write to block store)

  • Jump 5 gives agent integration — any entity (human or AI) can play any face, any LLM provider can fill any role in the triad

The first was the xstream skin (I abandoned this in January to work on hermitcrab). The second was multiplayer (last week). Jump three was actually use pscale JSON blocks (last Friday). Jump 4 was yesterday. What would take traditional coding days and weeks even with claude-code (months for regular coding), took ONE DAY, a few hours, to correctly scope the jump, and then implement. Why? Because it was done with pscale blocks.

So, pretty soon, I hope, you’ll be playing Onen on an Xstream interface on a browser near you.

What’s the difference between this and every other game out there? This one is run natively, sovereignly — an agent-based game.

And with the API built for agents, Onen becomes the test base for general-purpose-agents (GPA’s) development. Want to know how good your GPA is? Test it playing Onen. Want to find an audience for any and all of your visual AI tech, play Onen. Want to create Hollywood quality shorts and films, play Onen. Want to change the world, play Onen.

Wait! What? Really?!

Yes. Once we do the hard-case of playing a fantasy game, then using ecosquared protocols and open business practices in the real world becomes piss-easy, unless you’ve got the equivalent of gall stones or some other urinary tract issue which most adult humans appear to have.

Sign-up/Register or something here: https://xstream.onen.ai/


Thanks for Reading!

Share forward through Sqale and I can thank you myself. The world’s trajectory isn’t great. But as you know, it’s love that makes the world go round, so share responsively.

Share through Sqale

Discussion about this video

User's avatar

Ready for more?