Simon Eskildsen on Learning Infrastructure

Table of content

Simon Eskildsen dropped out of school after high school in Denmark and moved to Canada at 18 for a gap year at Shopify. He never left. Over the next decade, he climbed from intern to Director of Production Engineering while the company scaled from 150 employees to over 5,000.

Without a university degree, Simon had to learn fast. He treated his own mind like infrastructure—building systems to optimize learning the way engineers optimize server performance.

The Learning Stack

Simon reads 30-50 books per year and built an elaborate pipeline to actually remember what he reads:

Kindle → Readwise → Anki + Zettelkasten

When reading, he highlights anything important. Facts get tagged with “.flash” in the note, which automatically syncs to Readwise and eventually becomes Anki flashcards. Abstract ideas and concepts go into his custom Zettelkasten—a note-taking system built on plain Markdown files.

“I read to learn, and I flash to make sure I remember. It’s probably the most impactful habit I have in terms of impact over time invested.”

He now has nearly 10,000 flashcards after 4+ years of consistent daily review.

Why Plain Text

Simon tried Workflowy, Dynalist, and Notion for his Zettelkasten but found them too slow and worried about longevity. His solution: plain Markdown files in a Git repo with custom bash scripts for searching and linking notes.

The tooling includes:

This matches his philosophy borrowed from John Gall:

“A complex system that works is invariably found to have evolved from a simple system that worked.”

Airtable as Personal Database

Before the Zettelkasten, Simon became known for his elaborate Airtable setups:

Each base starts simple and grows complexity only as needed. The tea base began as a rating list, then added brew logging, then automated suggestions, then push notifications.

From Personal to Production

The same first-principles thinking that built his personal systems led to turbopuffer, the vector database company he co-founded.

The origin story: while helping Readwise scale in 2022, Simon found that vector search for 100M documents would cost $20k/month—4x more than their entire database bill. The feature got shelved.

That gnawed at him. Existing vector databases stored everything in expensive RAM when object storage (S3) costs 100x less. So he built a search engine that uses object storage as the source of truth with SSD caching for hot data.

Cursor, the AI code editor, was an early customer. They saw 95% cost reduction when they migrated billions of vectors from their previous provider.

The Reading Selection Process

Simon’s heuristics for choosing books:

  1. Applicability: What can I apply right now? If recruiting, read recruiting books.
  2. Syntopical reading: Go deep on topics—read multiple books to see different perspectives
  3. Age: Old books that survived are likely to stay relevant (Lindy effect)
  4. Breadth: Balance depth with exposure to different disciplines
  5. Sample first: Send Kindle samples of top candidates and skim before committing

He actively resists “should read” lists in favor of following genuine curiosity. Getting through Anna Karenina was a struggle; he tore through a book on telegraph history.

T-Shaped Learning

Simon’s framework for learning is “T-shaped”—deep expertise in one area with wide foundations across many disciplines:

“There are so many disciplines where people learn to think in different ways to solve different problems. Over time, I’d like to get a rudimentary understanding of most of the major disciplines.”

His flashcards span everything from kilowatt-hours per 100km of driving to the definition of “pollyanna” to the history of New Caledonia.

AI for Learning

In 2024, Simon spoke about using AI to accelerate learning. The combination of retrieval systems (like turbopuffer), embeddings for semantic search, and LLMs creates new possibilities for personal knowledge management.

His work at turbopuffer directly enables these use cases—making it economically feasible to embed and search every document, note, and highlight you’ve ever collected.