The Lab: lots of experiments, each standalone, pinned to a release
The person's message, as typed into the session with the design agent on:
Now what I'd like you to do is with the design agent, which is activated on this chat, can you do maybe a new area where we can just do lots and lots and lots and lots of experiments of different type of characters, different type of animations, different types of layouts? Because I think we're onto something really powerful here. Because I think this programmatic way of explaining things is really effective. But I actually think that in sometimes we actually need to go deeper. Like, you know, almost like imagine having one that explains like a mermaid document or a thread modeling or an architecture review or a complex interaction, right? It's kind of more like an animation kind of thing. Also, I'd like you to try different type of characters and different types of, because a lot of this is just figuring out how to design it, right? It's figuring out how to design the little animation and the little character and then the rules and the physics and how it's supposed to behave. And it's kind of like start to think of this pack that separates even like the characters and the storyboard. But eventually what you start to have is these common blocks of stage and actions and bits and bobs that we can glue together, and also we can probably bring one into another and we can explain it. So basically, but for now, make every single one of them a standalone. And actually, you can pin, of course, you can pin to the existing version because all the versions are now pinned, right? But they should all be then standalone. So as we change in the future, you know, they still work. And that's, again, a lot of these is experiments. So we're just exploring. And that's why I added the design agent. So you can basically take a step back and also imagine how some of these works. And let me actually also link here. I can find the risk mandate storyboards, because some of the characters there were quite cool. So maybe you might want to even try different design libraries and different ways. But the point is, it has to be programmatic, right? And also see if you can start to create an API that allows you to control the agents, almost like a game. Sort of like, in fact, take a look at some of the game JavaScript APIs. You know, like how much can we leverage from what exists there, right? Because there's a lot of very powerful APIs out there. So this is our opportunity to really try it out. Because again, a storyboard can have all sorts of stuff inside of it, right? So yeah, let's explore. https://riskmandate.ai/stories/
As the agent read it
| The words | Read as | Done in |
|---|---|---|
| "a new area ... lots and lots of experiments" | /lab/ on storyboards.sgit.ai: one standalone page per experiment, an index generated from each one's experiment.json, the rules in lab/README.md | the Lab |
| "make every single one of them a standalone ... pin to the existing version" | every page loads the kit from /v0/v0.1/v0.1.0/ by path and hash; other libraries vendored under versioned names in lab/vendor/; the gate accepts any immutable release on a Lab page | lab/README.md rule 1; admin/build/validate.py |
| "different type of characters" | an ink pack (thick line, flat fills, riskmandate.ai's stories as the reference for the style, not their pictures), a stick rig, a pixel pack, things with faces, a cast designer | the characters track |
| "different type of animations ... the rules and the physics" | the Stage API (verbs on a timeline, direct mode), a walk cycle, gravity with matter.js, a dialogue engine, a cinematic timeline with anime.js | the motion track |
| "different types of layouts" | a comic page, a filmstrip, a grid, scrollytelling, and a camera across one world | the layout track |
| "explains a mermaid document or a thread modeling or an architecture review or a complex interaction" | a flowchart walked from Mermaid text (dagre lays it out), a threat model probed flow by flow, a sequence diagram delivered message by message; an architecture review is read as the camera over a world of panels, and stays on the list | the deeper track |
| "an API that allows you to control the agents, almost like a game ... game JavaScript APIs" | Stage (the kit's own, hand-rolled), and Room, the cast as sprites in kaplay with keyboard control | Stage; the engines track |
| "common blocks of stage and actions ... glue together ... bring one into another" | the blocks named on the design canvas and the Lab index (actors, props, stage, actions, camera, layout); the camera experiment puts a panel inside a laptop inside a panel | the canvas; lab/camera |
| "the design agent ... take a step back and imagine" | a design canvas of six boards: the map, the Stage, the packs, the explainers, the layouts, the glue; private to the person until shared | the canvas (not on the site) |