Back to all docs

Learning Feed - Making Documentation as Addictive as Social Media

The Original Idea

From the user:

I'm think of something a little more interesting, fun, and bigger. So that scatter we are seeing is just in this repo, but i have so many repositories of interesting projects (past and currently being built) and they are on different topics/themes. One observation i made is that the more i collaborate with more people/teams/AI agents, the speed of building and shipping is super high and it becomes hard to keep learning at similar pace. On the other side, people (including my past self) are easily hooked by daily apps that are optimized to grabbing/stealing our attention, where we mindlessly spend hours on things that won't be remembered the following day. So, i'm thinking, what if there was a platform to extract documentations, source code of projects i'm working on, or even learnings i make from chatting with AI agents into engaging posts (or whatever best formats that can grab my attention easily into learning something new) that could make me laugh, get a little emotion, or whatever same way these "distracting" apps do, BUT for us, it would be made for better purpose/itention for someone to LEARN and keep up with the spend of building. So that would mean, as we build stuff when learning, developers would keep their docs, knowledgebase in their codebase and then we can transform those into our platform. Of course, we are starting with just us: me and my repos' docs, and you (helping me build and already wrote some docs so far). WHat do you think? Any questions, additions, or ...?

Core Problem Statement

The asymmetry between:

The core insight: What if learning materials were as engaging as TikTok, but actually served your growth?

Key Questions to Explore

1. Attention-grabbing formats - what specifically hooks you?

2. The feed algorithm - what surfaces when?

3. Content transformation - what's the input/output?

4. The learning loop - what's the goal state?

5. Social/multiplayer dimension?

6. The "scroll but learn" mechanic

Additions & Insights

A. Emotion injection is key

Your letters (doubting/proud versions) work because they have stakes and feeling. Technical docs need this too:

B. The "past you teaching future you" angle

Your system reference doc already does this - it's literally you preventing future-you from breaking the system. Scale this:

C. Leverage the chaos

The scatter we saw (personal letters + technical docs + blog post) is actually more interesting than perfect organization. Your feed could embrace this:

D. Start stupidly simple

Example User Journey

Day 1 (today): We build the blog, write docs, chat about PWAs

Day 2 (tomorrow morning): You open "DevFeed" app and see:

  1. Card 1 (story): "The Laptop I Gave Away" - first 3 paragraphs, cliffhanger ending, "Read more?"

  2. Card 2 (technical): "You learned about service workers yesterday. Here's what they actually do:" [visual diagram + 2 sentences]

  3. Card 3 (emotional): Quote from your doubting letter: "You're about to relax restrictions..." with a reminder of why you built The Sys

  4. Card 4 (challenge): "Can you explain what systemd does without looking it up?" [Tap to reveal answer]

  5. Card 5 (code snippet): The exact curl command you used to test the server, with context: "This is how you smoke-tested the blog"

All of this auto-generated from today's chat + commits + docs.

Open Questions

Why This Could Actually Work

The hardest part is probably the content transformation (boring docs → engaging content), but if anyone can figure it out, it's the person who locked their own phone and gave away their laptop to stay focused.

The system proves you understand:

  1. How addictive apps work (you've fought them)
  2. How to build constraints that actually stick (The Sys)
  3. How to make technical content emotionally resonant (your letters/blog)

This is just applying those same principles to the problem of knowledge retention at high building velocity.

Current Assets (in this repo)

The yap repository already contains the variety of content types that could seed this:

This scatter is the raw material. The question is: how do we transform it into something you'd actually want to scroll through tomorrow morning?

What's Next?

The next concrete step depends on what aspect feels most important to validate first: