Rendered at 19:36:20 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
gopalv 35 minutes ago [-]
I've been building out a sort of "clickwise"[1] for agents to talk to each other without resorting to English as a communication format.
My problem is approximately ~2000 rules in various levels of conditional nesting from a central context agent to the execution side of things.
This is a better worked example than my confused mess of financial rules.
The structure is a bit more fleshed out, but you are still sending across data which is hard to decipher for a human, but perfect for a machine to bootstrap and learn a set of nested rules about the task.
Finance is a bit less consistent than chess on the rules, but packing into binary "Telepathy" protocols is going to be the next big thing for multi-agent orchestration.
Why does the first frame contain 0, 1, 2, 3, 4, 7, 8 and 15 _twice_? It says something about leading zeroes, but visually that makes no difference. Why isn't this room used to put _all_ the values from 0 to 15 so that there's no implied question of why were those numbers chosen and no others and does that have meaning or not?
nixosbestos 10 minutes ago [-]
Am I wrong to immediately think of Contact?
tychodaimon 1 hours ago [-]
wow!
legotin 5 hours ago [-]
I built Golden Board, a 512 KiB binary message designed to teach chess to an unknown recipient without shared language, culture, or prior knowledge of the game.
The project is inspired by a tradition of messages designed for an unknown intelligence such as Voyager Golden Record, Lincos and others.
But instead of sharing science or art, it teaches chess.
Why chess? Because a small set of rules opens up an enormous world of ideas and beauty. We can teach it from scratch and then let the recipient explore a rich legacy shaped by centuries of human ingenuity, sacrifices, and moments of brilliance.
The message begins by teaching the recipient to count, recognize bytes, and gradually decode and understand everything that follows.
I tested it extensively, using LLMs to simulate an unknown recipient and carefully checking each part of the message. Of course, LLMs are far from perfect stand-ins, but those experiments helped me identify ambiguities and improve the design.
The project is an attempt to solve a broader problem: its bootstrap is chess-agnostic and designed to decode other structured messages. Anyone is welcome to build on it and create messages that teach something entirely different.
My problem is approximately ~2000 rules in various levels of conditional nesting from a central context agent to the execution side of things.
This is a better worked example than my confused mess of financial rules.
The structure is a bit more fleshed out, but you are still sending across data which is hard to decipher for a human, but perfect for a machine to bootstrap and learn a set of nested rules about the task.
Finance is a bit less consistent than chess on the rules, but packing into binary "Telepathy" protocols is going to be the next big thing for multi-agent orchestration.
[1] - https://www.penny-arcade.com/story/automata/automata/automat...
The project is inspired by a tradition of messages designed for an unknown intelligence such as Voyager Golden Record, Lincos and others.
But instead of sharing science or art, it teaches chess.
Why chess? Because a small set of rules opens up an enormous world of ideas and beauty. We can teach it from scratch and then let the recipient explore a rich legacy shaped by centuries of human ingenuity, sacrifices, and moments of brilliance.
The message begins by teaching the recipient to count, recognize bytes, and gradually decode and understand everything that follows.
I tested it extensively, using LLMs to simulate an unknown recipient and carefully checking each part of the message. Of course, LLMs are far from perfect stand-ins, but those experiments helped me identify ambiguities and improve the design.
The project is an attempt to solve a broader problem: its bootstrap is chess-agnostic and designed to decode other structured messages. Anyone is welcome to build on it and create messages that teach something entirely different.
Source code and docs: https://github.com/arlegotin/golden-board