Research as a portfolio
Over the past year, I've been soliciting advice for how to pursue "transformative research." The cornucopia of counsel I've received is worthy of its own essay, but today I wanted to focus on one nugget of wisdom that's been particularly challenging to follow: pursue a portfolio of bets.
Some relevant background about me: I'm a one-thing-at-a-time guy. For me, reading Paul Graham's "The Top Idea in Your Mind" was not a revelation about priority; it was a description of lived experience. In a User's Guide to working with me, I wrote:
I like to focus on one big initiative at a time. When I have to divide my days between two or more large initiatives, I start to feel drained. I can only have one top idea in my mind, and I'd prefer to not have competition for that slot.
Which leads me to the advice I received about pursuing "transformative research": don't lock in on a single project too early. The thinking goes as follows:
- Transformative research projects tend to require 3-5 years or more of directed effort before they yield significant insights.
- Not all research projects have the same difficulty or leverage in producing insight; some have obvious low-hanging fruit once you start to dig in.
- Therefore, evaluate multiple projects in some depth before settling in on one.
I was talking with Andy Matuschak about his early forays into research. At first, he said, he was very resistant to settling down on a single project. He spent about a year taking more of a portfolio approach, but it was a frustrating experience. He was thinking about projects in terms of "making progress on the problem" rather than trying to get signal on whether the project was worth betting on. His advice?
- Build well-scoped prototype projects, cutting scope to the smallest thing that might feel "juicy."
- Avoid the tendency to "do things right" early on in the project. It's a bet, remember?
Andy reflected that if he'd been thinking in bets, it may not have actually changed his behavior in practice, but it would definitely have changed his psychology.
This leads to a few interesting questions:
- How to scope a prototype project appropriately?
- What signal would indicate that a particular research project is fruitful?
My portfolio so far
The last few weeks I've been exploring a variety of different bets, which has been challenging when I really want is to go all in on one thing. Here's my current "portfolio":
- Spaced repetition and knowledge management - I'm investigating the integration of spaced repetition prompts into a knowledge management system. The narrowly-scoped project is syncing prompts from Markdown files into Andy Matuschak's Orbit. I've also been tinkering with various spaced repetition systems (Fluent and Execute Program, specifically).
- Collaborative operating systems - My hypothesis is the only way to improve collaborative knowledge-making is to build a new operating system that's a multi-user experience from the ground up. I'm looking at Croquet as a delivery mechanism for that collaborative OS, and I have my own project called Tapestry.
- Whiteboard digitizer - I have a whiteboard in my office with a 4K camera pointed at it. I often use that setup on video calls or while streaming to share my whiteboard. I also have written a small set of command line scripts to take a picture of the whiteboard, crop and rotate, and color adjust. I'm looking into turning those digitizer scripts into a simple Rust CLI utility.
- Blockchain and NFTs - While learning about how Blockchains work, I'm increasingly convinced that Blockchain represents a unique and versatile technological solution. I'm still unclear what problems it will be most capable to solve. I'm currently working with a friend to build tooling to understand and explore NFT markets.
- Personal knowledge management systems - Each week, I am thoughtfully tweaking my own knowledge management system. I haven't written extensively my ingestion and writing process, but I do have a video where I draw out my process on whiteboard.
Did I mention that I like to focus on one thing at a time? Well, I think you can see the challenge. The breadth of taking a portfolio approach is daunting and draining . I'm still experimenting with how to organize my time and architect my environment so that I don't end up spread like "butter over too much bread."
Things that caught my eye
- I came across a fascinating overview of generating natural language to describe a complex system using text. The video goes into a bunch of the interface and technology challenges that he had to overcome in order to utilize the technology in a game called Knights of San Francisco.
- Have a Mac? If so, chances are you use homebrew to install software. I set up a new Mac this week, and I was reminded of brew's ability to install a set of software from a
brew bundle dump and
brew bundle install. You can check out my Brewfile in my dotfiles.
- In crafting novel tools for thought, you have to overcome the "first mile problem": how do you onboard new users seamlessly to the experience and present a UX which requires almost no explanation? Ultra-powerful tools like Roam introduce new concepts. Those new concepts need simple UX in order to be adopted. Cortex Future reflects on their experience teaching Roam to beginners in this thread.