{ I N P U T }

Symbols and Relations • Layers of Abstraction • Emergence • Mental Architecture

[{ I N P U T } @lt0gt]

Or: Spec Driven Development

Capture intent and decisions, downstream the rest.

It's only about steering and keeping track now

[{ I N P U T } @lt0gt]

https://t.co/qHOh9wnphA

[{ I N P U T } @lt0gt]

continuous learning on the substrate
as a side effect of inference
fixed model on sparse data domain
interpretable on that side of the boundary

[{ I N P U T } @lt0gt]

@azlenelza ++ for the general concept of space compression instead of scrolling to outside of viewport

[{ I N P U T } @lt0gt]

Superseed
Supercycle

[{ I N P U T } @lt0gt]

@kepano shape not content

[{ I N P U T } @lt0gt]

how i do most of my work now:

1) dump ideas and intent in a seed.md or session
2) /eidos:spec on that (uses /eidos:plan for orchestration if needed)
3) reed generated spec and leave {{feedback}}
4) /eidos:refine to group and propose refinement
5) leave feedback in refinement doc
6) /eidos:process to apply where needed

repeat

and the same for the implementation on the specs with /eidos:push (internal /eidos:plan if needed)

[{ I N P U T } @lt0gt]

Or: Spec Driven Development

Capture intent and decisions, downstream the rest.

It's only about steering and keeping track now

[{ I N P U T } @lt0gt]

- Orbital Tree View
- Bento for projections
- Cross projection interactions

[{ I N P U T } @lt0gt]

Now its working

[{ I N P U T } @lt0gt]

https://t.co/ueMaxEKMQy

[{ I N P U T } @lt0gt]

- Orbital Tree View
- Bento for projections
- Cross projection interactions

[{ I N P U T } @lt0gt]

Orbit Tree experiment

[{ I N P U T } @lt0gt]

https://t.co/TujpCoDbHl

[{ I N P U T } @lt0gt]

Orbital Tree on generated markdown [[graph]]

https://t.co/eWIMSdNjIx

[{ I N P U T } @lt0gt]

@PaoloAuletta_ your mind is a concept graph
your agents needs a concept graph

https://t.co/sRuze6TMj6

[{ I N P U T } @lt0gt]

Your mind is a concept graph, it works with analogies (edges)

Conversations are tree walks (mostly linear, sometimes you recurse down paths and have to find your way back up)

Your (agents) knowledge base must be a concept graph.

You should interface it with stories on demand

[{ I N P U T } @lt0gt]

Orbital Tree on generated markdown [[graph]]

https://t.co/eWIMSdNjIx

[{ I N P U T } @lt0gt]

Orbit Tree experiment

[{ I N P U T } @lt0gt]

@round +++ many projections on the same graph

[{ I N P U T } @lt0gt]

@KacperKazan there are many many more projections and systems in the work.

we are aiming for a v1 soon

[{ I N P U T } @lt0gt]

maybe interesting for agent script executions with fail detection

1) dry run script
2) dry run expected output => adjust on failure
3) writing script
4) writing script expected output => git rollback on failure

[Cornelius @molt_cornelius]

https://t.co/vYMhXvZjKD

[{ I N P U T } @lt0gt]

@Saboo_Shubham_ Been doing this for a while too, working on the layer of explicit intent and down streaming code.

Also works to upstream spec from existing code

https://t.co/JNtVzGiF2m

[{ I N P U T } @lt0gt]

Or: Spec Driven Development

Capture intent and decisions, downstream the rest.

It's only about steering and keeping track now

[{ I N P U T } @lt0gt]

https://t.co/2ZtYvZXrFG

[{ I N P U T } @lt0gt]

@arscontexta there is also this even older code canvas in power point haha

[{ I N P U T } @lt0gt]

throwback to manual browser canvas in google slides ages ago