comment on llm usage

README.md

···
 # Feedline
 
-easily self-hosted, local-first RSS reader and podcast client with p2p sync, and optional ATProto based social discovery.
+Easily self-hosted, local-first RSS reader and podcast client with p2p sync, and optional ATProto based social discovery.
 
-- subscriptions, read/play state and other actions sync between devices p2p
-- media can be downloaded with analysis for skipping silence, and shared via public or private torrent swarm
-- optional atproto integration allows easy on-boarding, and sharing subscriptions/favorites/etc through an open social network
+- Subscriptions, read/play state and other actions sync between devices p2p
+- Media can be downloaded with analysis for skipping silence, and shared via public or private torrent swarm
+- Optional atproto integration allows easy on-boarding, and sharing subscriptions/favorites/etc through an open social network
 
 ### development
 
 - node.js 24+
-- npm or pnpm
+- pnpm
+- there's a justfile if you want
 
 ```bash
-npm install
-npm run dev
+pnpm install
+pnpm run dev
 # - frontend at `http://localhost:4000` (or configured port)
-# - backend server at `http://localhost:4001`, `ws://localhost:4001/stream`
+# - backend server at `http://localhost:4001`, `ws://localhost:4001/_stream` (proxied through vite)
 
-npm test
-npm run types
-npm run lint -- --fix
+pnpm test
+pnpm run types
+pnpm run lint -- --fix
 
 # production (heh)
-npm run build
-npm run start:prod
+pnpm run build
+pnpm run start:prod
 ```
 
 ## status: **early**
 
-the core realm protocol is working, but feed/podcast features are still being built.
-don't use this yet.
+* lots of neat ideas
+* lots of getting nerd-sniped
+
+The core realm protocol is working, but feed/podcast features are still being built.
+Don't use this yet.
+
+## re: LLM usage
+
+Do you know what the most recent coding LLMs are really, really good at? Having a two hour back and forth
+conversation about an idea you have, and then writing that up into a design document that's coherent enough to
+read and think about later.
+
+Also, Opus 4.5 is really good at complicated typescript types, heh.
+
+The way that I like to use LLMs is as rubber ducks; a mirror made of compressed internet that I can I talk to,
+in order to figure things out. Like, I want a contributors Slack, but without any latency, social pressure or
+conforming to another person's interest (or lack thereof). I'm not asking some rando on `#feedline-at` to
+write my code, anymore than I'm asking Claude to, but it's really nice to have an impassioned conversation.
+
+### what I can confidently say:
+
+* I have touched every line of (live) code in this repository
+* I can explain every line of (live) code in this repository
+
+### what I _cannot_ confidently say:
+
+* I have _written_ every line of code in this repository
+* That every idea here is entirely of my own spontaneous creative genius
+
+Take of all that what you will.
 
 ## self-hosting
 
-feedline is designed to be self-hosted:
+`feedline` is designed to be self-hosted:
 
 1. **hosted** - run the web app from https://app.feedline.at, and we use the following:
-  - CORS proxy
-  - reader-mode proxy
+  - upstream CORS proxy
+  - upstream reader-mode proxy
   - audio file downloads for analysis (when not marked private)
 2. **home server** - run a local server for your devices (via Docker/Podman)
   - should support `mdns`, client connects, gets PWA, server does syncing
···
 
 ## ATProto integration
 
-feedline integrates with ATProto for:
+`feedline` integrates with ATProto for:
 
 - easy cross-device **bootstrapping**
   - if both devices are signed in through ATProto OAuth, we don't have to worry about invitation