Julia Evans

How does knowledge get locked up in people's heads?

or, “Building expertise at work”

I mentioned yesterday that I joined a new team at work this week, so learning and expertise have been on my mind. For the first time in a while, I’m working on systems that I don’t always understand very well, and where I often can’t figure out what the system does on my own.

Something that has really bothered me at work, for a long time, is – there are a bajillion people who know amazing things, and often that knowledge gets stuck in their head and other people don’t end up learning it. For years, sometimes! How can this be?!

I’m trying to understand this, so as usual I decided to write about it.

In this post I’ve come up with a couple of reasons sharing information is hard, and a few things I’ve seen actually work for making your team have more experts. All of this assumes that everyone is on the same side and genuinely wants to share information.

Discovery: who knows the thing?

So, if you’re going to learn awesome stuff from the people you work with, first you have to actually know that they know it. here’s an handful of examples of stuff I know that other people know.

  • erik knows a lot about jvm performance (as does ianoc, I think)
  • to get information about scala I’d just go to the #scala slack channel because there are a ton of scala experts. scala information is easy to come by.
  • nelhage writes the most about the linux kernel and evan knows a ton too
  • cory just wrote cool blog posts about java garbage collection so it turns how he knows about that.
  • jessitron knows about java concurrency models and her and jorge wrote a great article about implicits
  • alyssa has a PhD in statistics and can answer basically any stats question I have

I have a bajillion pairings of people <-> information like this in my head. But as I’m writing down these examples, I’m realizing how not obvious it is who knows what! For instance, suppose I was running into a problem with JVM garbage collection 4 weeks ago. I would definitely 100% not have thought to ask Cory. But turns out he knows a lot about that!

And Ray wrote Cloudflare’s DNS server. He won’t tell you this if you meet him so you have to go look at his LinkedIn or something and read about it. This is a super relevant thing to know if you work with him!

If I had a question about Go, I’d probably look up who has committed to Go services internally, and maybe I’d come up with Carl or Colin, but I also happen to know that Aditya knows quite a bit about Go and gave a cool talk at Gophercon. I don’t know how I’d discover that without googling “$coworker golang” for every single one of my coworkers. And that wouldn’t even do it because lots of people don’t have an internet presence. Or maybe I wouldn’t notice Mark’s great Go By Example site. There are probably at least 7 people at work who are really good at Go and I have no idea who they are. I just found out this second that Mark has a ton of interesting Clojure repositories on github. There is so much I don’t know about my coworkers’ expertise! Eek.

So, discovery. In general I know (or can easily find out) what projects people work on at work. That is no problem. But. I’m constantly shocked by how often I find out that there’s a topic I need to know about for work and someone I work with either

  • has a ton of previous professional experience with the topic (“oh yeah i worked on networking for 6 years”)
  • maintains an important open source project in the field that I had no idea about
  • has been learning about the topic in their spare time and now knows a lot about it
  • … maybe most importantly, don’t know a lot about the topic and are really really eager to learn about it and we could totally learn together and be wizards

and I had no idea. I basically think this is a travesty. People occasionally run internal talks about topics they know about which is THE BEST. But honestly sometimes I wish every developer (to say nothing of my other great colleagues!!!) I work with came with a resume or document or something saying HERE IS WHY THIS PERSON IS REALLY IMPRESSIVE AND AWESOME AND THESE ARE THE THINGS THEY ARE AN EXPERT ON.

Okay, so that’s discovery. I don’t know how to fix it but I think it should be fixed, because not knowing that you have internal experts on a topic is just silly.

how do people become experts?

Here’s the easiest way.

  • person builds system from scratch
  • person is the expert forever

This sucks. The person who built the system has to answer questions about it forever, which puts a lot of pressure on them, and other people can’t modify the system if the person is on vacation (“bus factor”).

I find this a little mysterious because there are internal systems that are important which have a Single True Expert but that I honestly don’t feel are so complicated – they’re often a few thousand lines of code, and have some history and complexity, but I feel like somebody else could learn almost everything about the thing in 6 months. But often nobody does.

Like everything in this blog post, I don’t know the answer to this one. It’s easy to be an expert on a system that you built from scratch.

I think there are maybe two reasons for that (and probably more!)

First, if you built it then you have a really strong sense for the history of the system. Systems don’t come into being fully formed, and if you try to understand them that way it doesn’t work.

Second, you end up with this sense of… entitlement, or something, to knowledge about the system. If someone asks me a question about a system I started, my basic assumption is “yeah, I can answer that, no problem!”. And even if I haven’t looked at it for 6 months or other people have done significant development work on it afterwards, I still expect to be able to just ask them questions about what they did and figure out the answer to arbitrary questions about the system.

But for systems that I didn’t build, I don’t always feel that sense of “uh yeah I can definitely work that out just give me a minute / an hour / 3 hours”. Even though often it’s not that hard to figure out!

I kind of feel like telling someone that they can’t understand a system or that it’s “too hard” or like “yeah that’s weird only $other_person” knows about that is like… a crime. We should all expect to be able to understand the software systems that we work with! right????

Now that we’ve talked about problems, here are a few ideas for solutions.

to make someone an expert, ask them questions

One thing I’ve noticed is that sometimes you have Person A and Person B who know roughly the same things but somehow Person A ends up being more respected and becoming the Internal Expert Who Everyone Goes To just because they’re better at communicating those things / advertising that they know it.

I think some people just like answering questions about a topic more, which is fine. But I think it’s more than that.

If everyone always asks Person A questions about the System, then that is itself an organizational investment in Person A’s knowledge. They’re constantly practicing answering questions and updating their knowledge and checking their understanding of the system.

This was a huge thing on my old team – for a long time, a small subset of people were responsible for answering questions on behalf of the team. Then we brought a bunch more people into the rotation. And after that those people reported feeling like they understood the system way better! Because they had to actually answer questions and figure out the answers if they didn’t know.

Kiran often says that you can build people who are “caches” for information – put them in charge of answering questions, they slowly start to accumulate knowledge and eventually become experts. Give them code reviews to do! Take their feedback seriously! Put them in charge of an architecture decision! Believe in them =D

I think the One True Expert has a role here too – if other people are going to become experts alongside them, then they need to sometimes hold off on answering questions and doing debugging work, even if it would be faster for them to do it or if they know the answer instantly. I always find this hard, but I think it’s worthwhile.

I think it’s worth considering who you encourage people to go to with questions. Is it always the same person? Why?


Dan Luu gave me some feedback on this post and he mentioned that an apprenticeship model has been successful for him in the past. Here’s how he described it:

  1. Someone new comes in
  2. Experienced person sits down next to new person and walks them through every “new” task for some period. At the first place I worked, this would be weeks or months.
  3. During and after, experienced person makes themselves super available for questions at any time.

I find this interesting because it’s a longer-term commitment. As time goes on new things becomes less and less frequent, but I think it’s really important to observe that that can actually take months. Sometimes I think there’s an expectation that somebody is done learning after just a few weeks which really isn’t true.


Let’s talk about internal talks! I LOVE talks. I find it kind of curious that outside of work I give tons of talks (“want to talk at montreal ruby next month? sure!!!!”) but at work I give almost no talks, and most of my colleagues don’t either. What’s up with that???

I think it’s easy to assume that if you propose an internal talk series, people will just show up with great talk ideas, but this actually isn’t true at all! You need to prod people into “hey remember this thing that you know really well and is really important? OTHER PEOPLE DO NOT KNOW IT!!!” and they’ll be like “oh yeah right!!” and then give a killer talk about it.

This week I remembered that not everyone I work with knows how to use tcpdump, and it’s a useful tool to have in hand, so I organized a tcpdump workshop for next week! Yay!

“how did you find that out?”

This is a MAGICAL TACTIC. I forget who I learned it from but maybe Alyssa.

Sometimes I ask someone a question “hey why is the database having timeouts” and the answer is “oh sorry I fixed it”. This is sometimes great and sometimes my LEAST FAVORITE ANSWER.

I often follow up with

  • how did you do that?
  • how did you know that?
  • what command did you run?

This is a great way to extract information from experts – often they don’t realize what they know, but if you ask them what they did FIVE MINUTES AFTER THEY DID IT they can usually remember.

From the other side, when people ask me questions I also often try to include the source of my answer – “the answer is $blah but I found it by doing X Y Z”.

building expertise is hard!

All this stuff is really hard to get right, but it feels to me really really worth it – if you have a Single Person who is the only expert on a system for a long time, I think it’s ultimately really bad for the system and for the organization.

I would love to know how you make this work! Have you become an expert on a project that someone else started and is 2 years old? Did you successfully hand off a large body of work so that you basically never answer questions about it anymore? (without quitting?) Do you have a really strong sense for who knows what in your organization and you feel like they teach other people really effectively? How did you do it?

thanks to Laura Lindzey and Dan Luu for their comments!

CDNs aren't just for caching The Etsy site performance report is amazing