Our Vision
Build a mind that
compounds.
We’re not building another note-taking app. We’re building a persistent cognitive layer — a system where every piece of knowledge you encounter makes everything you already know more valuable.
The broken model
You read a thread about transformer architectures. You watch a lecture on attention mechanisms. You skim a paper on arXiv. You bookmark a discussion about vector databases.
What happens next is the failure mode. Each insight lives on a different platform. Your brain connects “that tweet about KV caching” to “that paper on efficient inference” — but no tool does. After two weeks, you remember you learned something, but not where, when, or how it connects to what you knew before.
This isn’t a filing problem. It’s a compilation problem. Your brain does the synthesis but can’t persist it. Existing tools persist the raw data but can’t synthesize it.
Compilation, not storage
Every knowledge tool you’ve tried is built around the same broken model: store now, organize later, search when needed.The problem is that “later” never comes, and search only works if you remember what to search for.
Storage tools
Source #50 sits next to source #1. No connections. No synthesis. A graveyard of saved links.
Compilation
Source #50 enriches sources #1–49. Concepts deepen. Connections multiply. Knowledge compounds.
How compounding works
When you add a new source, the AI doesn’t just summarize it. It reads your entire existing knowledge graph and asks: what connects? What contradicts? What deepens?
A blog post about attention mechanisms gets linked to the paper you read three months ago about transformer efficiency. A podcast episode about scaling laws gets cross-referenced with the arXiv paper that measured the exact scaling coefficients. The concept page for “Retrieval-Augmented Generation” evolves with every new source that mentions it.
This is what compounding means. Not just accumulation — synthesis. Every piece of knowledge makes the whole system more valuable. Connections you’d never make manually emerge automatically.
Auto-linked concepts
Contradictions flagged
Cross-references built
Questions compound
Where this is going
The product you see today is Phase 1. The long-term vision is much larger.
Phase 1
Now
Personal knowledge compilation
You feed it sources, it compiles your knowledge. Concepts extracted, entities identified, cross-references built, contradictions flagged. A personal wiki that grows smarter with every source you add and every question you ask.
Phase 2
Next
Portable knowledge identity
Your compiled knowledge becomes a portable profile. When you interact with any AI agent, you share your knowledge graph as context. The agent instantly understands what you know and what you don’t — tailoring explanations to your actual expertise. No more re-explaining your background.
Phase 3
Future
Connected intelligence
Knowledge graphs that connect across people. Research groups that share concept maps. Teams where organizational knowledge compounds across every member. The collective intelligence layer for how humans learn together.
There is room for an incredible new product instead of a hacky collection of scripts... An LLM reads everything you’ve consumed, and constructs a wiki on your behalf. Knowledge is compiled and kept current — not re-derived every time you need it.
12M+ views. 2,100+ GitHub stars in 12 hours. Karpathy described a pattern — we built the product.
What makes this different
This isn’t about replacing your tools. It’s about adding a layer that does the work humans abandon after two weeks.
Compiled, not stored
Knowledge is synthesized once and kept current. Concept pages evolve as new sources arrive. Cross-references are pre-built.
Connected by default
Every source is automatically linked to related concepts, entities, and prior sources. The graph grows with every ingest.
Human curates, AI maintains
You choose what to learn. The AI handles summarizing, linking, and the bookkeeping nobody does after week two.
Every question compounds
Ask your knowledge anything. Good answers get filed back as new pages. Every question makes the whole system richer.
Yours to keep, forever
Standard Markdown files. Obsidian-compatible. Exportable anytime. Your knowledge in your format, on your machine.
Retention, not just recall
Spaced repetition built on top of compiled knowledge. Fight the forgetting curve with the things you actually learned.
Stop consuming. Start compounding.
Your knowledge should be a compounding asset, not a leaky bucket. Start building the cognitive layer you deserve.