Fable helps optimize performance- it outputted a report

SSR render perf, avoiding thrash: https://github.com/seed-hypermedia/seed/pull/851

Editor extensions are loaded on demand

Notifications load before the underlying data

Comment from horacio

Full-markdown intermediate step?

Current: EditorState -> HMBlockNode

Proposed: Any Editor -> Markdown -> HMBlockNode

Consensus: we are staying with HMBlockNode and will introduce empty/fragment block. We need to work on the tooling around it.

Tech debt: we have two converters from markdown to HMBlockNode

Two big directions - use agents to experiment with both directions to see which has a better

• Introduce "blank node", "fragment node", keep our existing editor

• Rewrite editor with prosemirror, blocknote, proseKit, introduce markdown intermediate. Write our own extensions

Everyone please read/comment on Horacio's comment here

First approach- full support for noSchema data editing

• P1: Doc Metadata

• Raw blobs read+write

• Block Metadata

We can construct schemas on top

• Are schemas documents, blobs, something else?

tech stack? defer to agent?

• Server resource utilization

• Perf regressions

• Uptime

• Notification server errors

• New VPS for dashboard

• Coolify?

• Standalone server?

Shared instances vs reserved? Cost differences!

What features we want in the future?

• Sub agents

• Memory

• Code execution

• Whatever it takes to implement import/export workflows!

• Agent introspection?

 had agents firing unexpectedly

Julio can't build desktop on his machine, but it works in CI?