Want to get notified about new posts? Join the mailing list and follow on X/Twitter.
Who does what:
- Alex oversees all our content and manages the team of content developers who write the entire mountain of content from scratch (as of 2024: thousands of lessons, tens of thousands of worked examples, hundreds of thousands of questions).
- I develop the quantitative/algorithmic components of the system (all the AI & science-y stuff, knowledge graph management infra, and any algorithmically complicated back-end infra in general). More info here.
- Jason develops the rest of the tech (all the UI/application-level stuff in the system, anything that interacts directly with users) and is the “big idea” guy behind the product vision.
- Sandy does all the business operations, customer support, etc., pretty much every other thing that is vital for us to stay afloat besides tech and curriculum.
Follow-Up Questions
So besides content, there are just 2 people coding and 1 doing customer support? How do you get so much stuff done?
Yes. As you might expect we are very time-constrained, we’re working around the clock, and we have to be very discerning in what we work on at any given moment. We try to stay focused on core functionality. Our the rate of incoming to-do items far exceeds the rate at which we can get them done, so we just have to take a hardline approach to prioritizing what we’re working on, always working on things that will move the needle in big ways.
Why is the software team so small?
(Added 2025) We have previously made efforts to expand the software team, in an effort to move faster, but have repeatedly and ultimately come to the conclusion that the fastest way forward for us at this time is for us to stay as small and streamlined as possible.
Have you heard of the barrels vs ammo analogy? If you add ammunition when you need more barrels, it doesn’t work at all.
(Added 2025) Yeah, it’s a good analogy, barrels are indeed the bottleneck for us. But they’re also much harder to find (especially in a deep/uncommon niche and on a budget) and harder to evaluate initially, so it takes a lot of effort to find an expected barrel – and even when you do, developing them is still a big, risky, long-term time investment that puts a drag on your near-term progress. We’ve been burned by barrels that looked promising initially, but didn’t pan out, wasting many hundreds of hours from true barrels in the process.
One trend we’ve noticed is that all our functioning barrels (myself included) started out doing more ammo-like tasks, doing an amazing job at those tasks, jumping at the opportunity to take on additional tasks (some slightly more barrel-like), doing an amazing job at those tasks as well, and compounding this cycle over a long period of time. So, for the time being, that’s our approach to find/develop barrels. (Though unfortunately, software doesn’t lend itself to ammo-like tasks nearly as well as content.)
Can you elaborate on the division of software between you and Jason?
- Jason is more of a full-stack / systems software engineer and I’m more of a “quant” engineer in the sense of a Wall Street quant. (But because there’s just two of us doing the coding, I spend half my time getting pulled into a lot of other stuff outside my comfort zone, which has been really good learning.)
- Loose analogy: It’s kind of like we run a medical practice, Jason is the generalist who diagnoses/remedies most of the body & heart & signals transmitted from the spine. But as you into the spine itself and then into the brain, things quickly become complicated enough that a specialist is required, which is me. If it were just me then the MA system would basically be a brain in a jar, unable to interact with the outside world. If it were just Jason then the MA system would be behaving less intelligently.
Want to get notified about new posts? Join the mailing list and follow on X/Twitter.