Skip to content
498: Season 2 Recap episode artwork

Mar 17, 2026 37 min

498: Season 2 Recap

from The Bike Shed Hosted by thoughtbot

Our hosts come together to wrap up season 2 and share some of their highlights along the way. Our trio expand and recap on some key topics covered over the last few months, Sally and Joël provide updates on their work with typescripts and LLMs, and Aji accidentally stumbles into an idea for a new keynote topic. — Your hosts for this episode have been thoughtbot’s own Joël Quenneville, Sally Hall and Aji Slater. If you would like to support the show, head over to our GitHub page, or check out our website. Got a question or comment about the show? Why not write to our hosts: [email protected] This has been a thoughtbot podcast. Stay up to date by following us on social media - YouTube - LinkedIn - Mastodon - BlueSky © 2026 thoughtbot, inc.

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

498: Season 2 Recap

0:00 37:23
of matches

Transcript Auto-generated, may contain errors

Hello and welcome to another episode of the bike shed a weekly podcast from your friends at the on about developing great software I'm aji Slater, and I'm Sally Hall, and I'm Joel Kimville and together we're here to share a bit of what we've learned along the way and This episode is the recap for our season two of the bike shed How has the season been for everybody good? I've enjoyed this season. Yeah, you know We joined season one which started episode 400 something so season zero. It's very long and I feel like season two I'm starting to feel like yeah, I had a podcast I can say words. It's working That's true, that's true I feel like there has been for myself a lot of learning and becoming comfortable and like finding a groove to this sort of thing And it seems like Sally that was your experience as well. I'm curious then Joel. How has this felt for you?

I really enjoyed this I do think that we've sort of found a little bit more of our groove I think the rotating between hosts which is a format that we introduced late last year when we did our season one Sort of rebooted the format of the podcast has been working really well So I'm enjoying the way that we're doing this. I'm curious for you two in this past season were there episodes that you're sort of feel particularly tied to like that was one of my favorites or I really enjoyed recording that one That stood out to you. I think for me There was one with each of you that really kind of stand out as highlights for me as far as season two does with Sally It was recording our AI episode that has started some conversations and different things internally at thoughtbot as well as between each other and Also Joelle with you the personal big biography episode was a lot of fun And just kind of like sharing things that were meaningful for us in our journey And that was really great getting to find out some what I found to be a shock that you had never read fooder

That is still is etched into my memory for as much as I absolutely love Sandy mess's work I have not read principles of object oriented design in Ruby the correct name wrong I might be misremembering this but does Sandy mess live near you Sally is that right? Yeah, she lives in my town and I'm a fan girl who've never seen her and I feel kind of weird admitting this publicly But like Sandy if you're listening we both do Ruby we're both queer women. We both live in terms of friends I love it. I love it. Yeah, I also enjoyed the AI one Well, I'm looking through the ones we did together and I'm trying to choose a favorite and I really like all of them I think maybe talking about what makes a code base welcoming was particularly fun Hmm, I think before we really got into that conversation the idea of like a welcoming code base hadn't really solidified in my mind But I think about that way now sometimes it's not the framing question, right?

And so once you have that sort of that frame of it's not just oh, how do you make a code base well documented or Easy be productive in or something but saying specifically How do we make it welcoming all of a sudden it sort of suddenly shifts the way you think about the way you start your things? Yeah, and the way you think about new people joining or I don't know I'm thinking this whole metaphor of like a code base is like a home and you know, I'm hosting you in this code base How can I make sure that when you come over to visit or when you move in that? This is a place that you can feel comfortable to and that we can now have shared ownership and hopefully not have the Monica's Closive code nobody wants to look at can I pitch a new mental model here? Yes The idea of code hospitality. Yes, yes I mean now I want to talk about that for an hour which might just be a rehashing of the what makes a code base welcoming But yes, let's let's keep this mental model going. I want this to live on

Yeah, I think my project team is going to be welcoming some new folks on to the project in a while And I will have to keep that in mind as as to how to make the project a welcoming place and revisit your episode to re listen to that So I can have it in front of mine even the whole like the people who used to live here Why did they paint the wall that way? We actually were painting our bathroom this weekend and like several moments of like what? Good idea. Only there were comments or pull requests where we could see why are they did this? Then you find out it was it was you that was all that was us two codes of pain to go. Oh, yeah

are there any projects or Concepts that you've carried through this season that you've learned some or grown or like seen some movement on that you started earlier on in the season Sally I feel like you know, we touched on this a little bit in season one But I think that's carried over into season two as well your sort of journey with react and especially TypeScript I think you've gone pretty deep on in the season. How do you feel sort of before and after here about those technologies? Yeah, I will say I miss rails. I miss Ruby. It's been a while since I've been on a like I Had two weeks of like doing some Touch-ups on an old rails project but other than that it's been a long time since I've been in my comfort zone of Ruby and rails so my most recent project is Textrupt but not react kind of node and I feel like I'm finally starting to feel comfortable I still want to go home to rails, but I guess I'm not as homesick as I was and I'm really like I'm the idea of Types it's growing on me more and more Especially I don't know for code that I didn't write or for code that I'm less familiar with than a while ago

It's just it feels like it's so helpful to have the shortcut of like What is that representing again? Oh, it should look exactly like this and it's like such a helpful guy post and like safety net maybe Like types of documentation basically yes types of documentation. I really I really appreciate that the project I'm working on We're doing there's a couple different packages that like essentially gathering data from a browser and sending it through a pipeline And I'm doing the part gathering data from the browser and someone else is doing the pipeline part But we're sharing types so that you know we can have a it sort of helps solidify like the interface between the various pieces of this a Little better having the share types which has been a discovery for me. That's awesome So feel like I might be in a bit of a metaphor mood today. Yeah, let's do it And I heard you say that it like feeling a little homesick for rails now I'm almost imagining the idea of being like a foreign exchange student in a different project That's very much what this feels like I'm studying abroad and Java's plan

I'm learning so much about the culture and some ways maybe it's fun and exciting but also novel and overwhelming Maybe this homesickness when I get back to rails. I'm gonna be that kid who said he brought once and it's like well But I was in JavaScript You see it's pronounced Barcelona You may not know this but JavaScript actually has nothing to do with Java Well, I believe misunderstood Now I just want maps on the walls of like JavaScript land and Ruby land and I feel like this sort of thing could almost fit in the type of things that you do as a conference talk. Oh, it very oh, no, you're right And my oh no, I mean that is very exciting I'm not a big one of the things person I suspect that one or both of you might enjoy Lord of the Rings and it feels like There might be some quest metaphors here. Yeah, that would be a lot of fun I feel like there that has similarities to Joel didn't you do something? There was a town like a JavaScript town in one of your talks

I think it was Ruby town. This was at the final rails conf I give a talk on 10 of the most influential blog posts in the last 20 years of Rails development And I flavored it as I'm like walking to or through a like historic town center where each building that I was pointing out Corresponded to a influential article. It's interesting You were talking about types because I feel like types have come up for me during the past season as well Yeah, and I know I probably a little bit in season one but I've been really playing around with the idea of What people get out of types and what people see as the value of types so sort of different Maybe mental models are sort of what is your perspective on why even have a type system?

What are you hoping to get out of it how you tend to see it interact with it Ranging from a sort of loose hierarchy of things right? So ranging from a Taskmaster to be satisfied all the way up to like a it's a pair programmer that Speaks to you in type errors and acts as like positive pressure on your design kind of thing also a thing. I'm sort of working on potentially a Conference talk on that sort of thing has come up I think a lot on the project that I'm on right now because there's a lot of types script But a lot of the people writing the type script are coming from more dynamically typed backgrounds And so sometimes conversations about like what are we even trying to do with the types here or oh?

We got a type error Here's the way we should try to fix it lead to broader conversations that have slightly deeper philosophical underpinnings are we just trying to get the like compiler shut up or Does there some deeper meaning that we should address here? And if we see a type error does that imply that we need to get better at our types or does that imply that there is an underlying bug and The implementation that we need to fix and where do you tend to look first as a developer? Do you find that there is one that needs to come before the other or sort of? Obscures the other I'm thinking maybe an example where the types aren't very Strong or well defined or they're a little bit loose And so they're not giving you that feedback that you might have a bigger question here Right and so you just kind of patch at the top level and never see that sort of larger conversation come up It's a little bit of both right I think in order to get to the point where you're like in conversation with your type system and all that Yes, you will want richer types then just strings and numbers everywhere that said conversations are too way

the compiler the type checker speaks back to you in type errors, but you speak to it by using richer types and so it's a little bit of a back and forth and The level of types that you use are a little bit the level of conversation You're trying to have with your type checker as you use fancy your types oftentimes allows you to have more nuanced conversations with your type checker And so you might want to move into a deeper conversation with it But you can try to have conversations with it at a more broad level as well It's like the beginning of your home stay you're kind of like pointing to the refrigerator and then your belly to say that you're hungry But by the end of the stay with the family you're able to talk about the you know particular must urns or whatever that they've found at the farmers market last week And how they hit your palate a little differently So you know what?

I do if I do give a talk on like hierarchy of levels of interactions with a type checker I might just re-steal the metaphor from you That's all right. I'll share it on your computer I feel like the way that I think about the type system is it feels a little bit like a game for me with different levels of difficulty Oh, so like base level inter-level of difficulty is just like get the type system to stop yelling at me And then like the next level is like okay all those places where you just shoved any in you can't use that anymore Like now you got to get it to stop yelling at me without just basically telling it to shush And then like the next level I think is like yeah, it feels sort of where the conversation kind of starts that you're talking about It's like okay What are these what is this type system telling me about what I've created and like because I did this recently Like we didn't have the type checker enabled correctly and I was just kind of going through my day and then Fix the setting was like oh right kind of forgot I was using types grip I should do this and it's gotten me to the point where I'm like thinking more deeply about okay

But what is this object actually like I've been treating it like it's just a pile of key value pairs But it's not it actually has meaning and what is that and how does it relate to the other kinds of things? We've got here and so it feels like I've you know the difficulty setting of my type script game Yeah, I'm percent I feel like this is also we'll have to go back and listen to this feels very familiar to some Similar kind of experiences that I was having a little earlier in a type journey of my own And I think this was before season one But Joelle and I talked about the command the like growing Friendship and companionship with a type checker as you kind of learn a type system on the bike shed before and so this is this is a conversation journey I feel that just continually evolves along a career and We've all been at different places in it and different levels in it as we've gone along and no doubt something that we will continue to discuss How we've gone grown and explored next season as well. I feel like the way you just spoke about that It makes me wonder have you ever given your type checker a name like come on Dave I feel like the one that most easily falls into being understood as a conversation is the the Elm compiler

Does the Elm compiler have a name that feels like something the community might have done just not have a name? I think in many ways the Elm community has a lot of overlap with the Ruby community in terms of like focusing on developer friendliness and All of that but one way in that it diverges from Ruby is It tends to prefer sort of straightforward naming as opposed to sort of fun quirky names that we like in Ruby So a lot of the packages the official encouragement is just name it for what the package does for easy discoverability and don't That give it some sort of cute name that no one will have to hear about from word of mouth So it does generate these sort of intermediate compilation artifacts and the finally extension for those is Elmo Love that But speaking of giving names to things Sally you released a library with a Appropriately cute Ruby theme name in the past week. Yeah, I think last month

I released Michelle which is named for the unhelpful concierge from Gilmore girls If you're familiar with it, if you're not go watch that show but keep in mind It's a product of its time The character part of his job is to set up a desk and answer the phone and schedule bookings And he hates doing that and hates talking to people and sometimes he just kind of pretends to look at the calendar And then says no we have nothing which is sort of how it feels trying to build self scheduling into an app Especially like a healthcare app Especially trying to you know coordinate with a an e mr and get that to help you No, we have nothing to get to yeah, don't want to talk to you So my Michelle is actually the solution to that It is a generator that helps you create a materialized database view and update classes So that it becomes very straightforward to query for available times for scheduling appointments And i'm excited about it I Have a whole bunch of ideas for what I want to do next and i'm really hoping that my next project Or a project in the future will give me an opportunity to you know Give it a try in the real world and implement self scheduling faster than it has ever been implemented before

What And nothing else is a good Gilmore girls conversation starter I'm so excited that it's finally out because I feel like on one of your first episodes you showed up on the bike shed before you were A host and you just came on as a guest we talked about this and some of the frustrations you had about managing Booking appointments and the fact that you were sort of working on a at the time private Gem to abstract some of these pain points away and I feel like it's come up a few times On various points where you've been on the show like oh, yeah, and i'm doing this thing as a side project One of these days it'll see that the light of day that days now that day is now and it's it feels very good Yeah, and I was we were actually recently had a talk among some of the team members in the america's here. I thought but we got two teams on the america's team about Extracting libraries and gems from client work and sort of the different motivations for some people It was that they were just like really excited about this thing that they had done for a client And they wanted to abstract it and then build on it work on it more because I just really enjoyed it And I was like, yeah mine was actually the opposite of that Which is that I worked really hard on getting self scheduling to work for this project and then thought I never want to solve this problem again

Let me find a way to abstract this and pull it out so that I don't ever have to do this again And that feels like one of those solutions or one of those challenges that it comes up so often that preparing yourself with a tool at hand Is going to come in in handy. I feel like that's the story behind a lot of classic thought thought open source because as a consultancy We work on a variety of projects we tend to encounter the same problems over and over and so creating standardized solution Is something that really comes to us easily and i'm thinking of things that even are like now Out of the box part of rails but it weren't back in the day so things like file uploads We create a paperclip as a like hey, we keep doing file uploads on different clients Let's create a gem that's just built in file uploads And we had paperclip we clearance as a way to handle authentication And now you know, there's built-in file upload support that's built in authentication in more modern versions of rails That's sort of maybe validation But also a bit of sweet and like the thing that you've done is really successful But then it's almost like it becomes so clear

This is the problem that everybody's trying to solve that it makes sense to make it first party Yeah, I think that is something that happens a lot in thought bot open source currently That doesn't get sort of the thought bot name or stamp on it or anything like that because so often a lot of our solutions that we're kind of redoing and Re-encountering our first recourse is to upstream it into rails right away And so there are a lot of things that have come from thought butters doing these kind of Perennial challenge again again and it becomes a PR to rails that has over the time gotten merged or kind of morphed into a different Concept and become part of the framework and so while there might not be as many thought about gems kind of grabbing the headlines out there There is still a lot of work that is coming from thought bot and you know Miriam number of other shops like ours that are just making that framework come stronger and more fully featured along the way And I really love that upstream first mentality and I take a minute to celebrate every time I get a chance to do some work like that I think that's one of the things that I really enjoy about consulting rather than just like being on a product team and working on the same product for years on end

Is that you get a chance to sort of iterate on problems and solutions and get better and better at it Right if I was working on the same product for a long time, you know, maybe I implemented self-surgical like a couple years ago and don't really love the way it turned out But it's there and it's fine and good enough. Yeah, good enough. It's doing its job. Whereas as a consultant I can go okay I know what I did last time and what I didn't like about it How can I do it just a little bit better this time? How can I solve the problem more cleanly or more universally or You know be proactive and like skip over that step where we had it working But it was really slow and everybody hated it and the adoptimized and go straight to the good one Although it does feel like Part of me is like but you don't understand how hard this one Have this shiny answer that I have refined over years and many clients But like how can I communicate to you that also like this is not simple?

That's the your talent as a software developer, right? Taking a really complex domain problem and providing a solution that seems simple Yeah, yeah, that's true And really I don't need the client or other developers or anybody to acknowledge that it wasn't simple I mean I kind of need y'all to so like thank you but like There's really nothing wrong with presenting a client with a beautiful solution to a complicated problem and having them underestimate how difficult it was that's fine As the single tier runs down both Joelle and his face. It's like yeah, it's fine Great I'm not here to get pats on the back for solving complex things all the time. I guess I guess We'll put each other on the back there you go. Yeah, we did it. You did it. We've all done it

It's a little bit of a you know standing on the shoulders of giants moment Absolutely, and you can be your own giant sometimes yes, yes Just the same way that the developer in the past can be yourself that you can look forward not only to your teammates But also helping yourself in the future out you can also solve the thing for yourself in the future Which is I think what Michelle is doing a little bit for you Sally One thing that I'd like to hear about Joelle a few weeks ago I can't remember exactly when you talked about using an LLM for get history in context And we had mentioned that we wanted to revisit that and see how it's going so how was that going Joelle? It is a thing that I've now sort of incorporated into my regular workflow It's not like a huge game-changing thing, but it's definitely something that I find pretty useful So two variations of this that I'll use is one just ask a like plaintext English question But pointing towards get history and then having it do all the like necessary It's the thing to come up with answers to my question with commit shots to have receipts on that the claim that's making This is often helpful if I'm debugging or if I'm telling something weird or sometimes even it's just I know that we

Did a change that's relevant to what I'm Working on and I might want to look into it before as I have pre-work Or things that I often like to do when I'm writing commit messages is just put commit shots Of other things that I'm building on that are related to it that an author that's reading my commit might want to also look at So I might say this is building on this work or a bug was introduced in this commit and this is a fix to it Or we tried this approach in this past commit and now we're pivoting to a different approach Just so that people can like see a little bit more of that continuity But sometimes finding the exact commit is just a little bit of drudgery And so people just say hey, you know like the thing that where we did that one that one time We did that one thing find me the commit and then it'll like give me a commit shot Maybe two or three if it is a few that might fit I'll look at them and like oh yeah that one and then I get to pick it So that's been really helpful And then I've also been using when I'm having LMS that need context on work It was done just pointing it to a commit shot and saying hey read this commit and Immediately sort of loads into the context window all of the relevant things about that change

So maybe it's a read this commit that was a refactor trying to set up we're trying to do here We're building on that and now we're going to try to do the thing You know, I'm doing a make the change easy that make the easy change Now that technique requires you to write good commit messages LMS are decently good at looking at let's say a diff and getting a sense of what was done But I find that the really crucial context that I often wanted to have are the types of things that I find valuable for human to have In a commit message So the why was this change made what were some paths that we looked at but chose not to go down And why do we pick this particular implementation? What are the gotchas in this change that sort of thing? And so with sort of one line of just like check this commit shot so just like proper conversation and all of a sudden it's able to Discuss the work I want to do intelligently with little to no effort on my part I can imagine using that tool and doing like who made this decision and why on earth did they do it this way and having that?

Oh, I'm coming back and be like, well, you did that so I've had very similar experiences where the the LLM is actually the one throwing shade at some code and saying who wrote this and I say to the LLM You wrote that Really blurring lines here the case what we're doing. Yeah Yeah, that's really great I love any tool that sort of in the process of using it or in the requirements of using it Well rounds out other places that are of benefit to yourself and the team so Introducing a tool or a process that not only completes its own purpose But also has kind of knock on effects throughout the work is always something that I'm giving I out for and I love finding new versions So that's not really cool. I feel like the flip side of that is with a lot of AI stuff I find that techniques that are good for humans are often also the ones that benefit AI tool So things like writing good commit messages like I deeply value that as a person because I find myself looking through get history A lot and so I get a lot of value out of that

And I strongly encourage teammates and clients to get into that practice But also to think that I think now maybe compounds a little bit once there's an automation layer on top of it I could see that going both ways. I could see someone saying like I am gonna write a really good commit message for this so that when I ask An LLM about it later. They have all the context I have in this moment and can give it back to me Or on the other hand going I don't need to write a good commit message because The LLM can just read through this and figure it out. Uh-huh I think a lot of people are thinking that way and this is where I would strongly push back and say no You can look at the diff and there's a lot of things you can infer from the diff But there are a lot of things that you cannot what you want to commit message are guidelines on how to interpret the diff And that's good for people and like I want that because I'm gonna look at a diff and be like huh And I want those guidelines and that context for myself But also if you can have automated tool interpreting the diff It can go in so many different directions and you want to point it in the right direction And something I found and this is something I've added to my note-taking system Is it oftentimes when working with AI tooling a lot of the ways you improve the results you get are

Finding ways of constraining it because otherwise it's just got two wide a set of possibilities And so anything you can give it to sort of constrain its output or to tell it This is the lens through which I want to interpret things Pruning branches it might go down Sort of giving it negative space that it can't search through All of these are techniques that tend to give better results This is this is nothing new for an LLM Like these are all piece of advice that you would give for yourself coworkers and other teammates as well Yeah, I think that's it kind of important to highlight is that Creativity happens in a box and and all of these different things like these are all having context and those commit messages Like you're saying these are all useful tools Not only for the LLM if you're only just now thinking of it because of an LLM great I'm glad you're here at the party anyway, but they're also going to help me if we're on the team So yeah, I feel like a lot of the things I've been pointing LLM's that have been Yeah, I guess I've been doing co-generation all that but a lot of these are like little side tasks that are like pretty scoped but often tend to be sort of Time-wasters for me things were like oh, it would be useful for me in my thinking process if I could have this artifact or if I could have

This information and like do I want to go down that side quest? So things like hey, help me find a couple of commit shots and you know, I'll be honest There are some days where when I was doing everything by hand I know there's a commit shot somewhere that relates to the work that I was doing and it would probably be useful for someone I'm too lazy to go find it and maybe I'll just like Give a rough description of past work and move on. You know, you can do the homework yourself It's an exercise left for the reader. Oh, hey, that's right So having some automatic tooling there helps me to sort of change some of the trade-offs There was like oh instead of it being like I'm gonna take 20 minutes and like break my flow to go dig around and find a commit shot I can just point on made a tool take 30 seconds write a paragraph say hey This is the committee I'm looking for find it put it in I can keep moving I think that actually highlights one of the reasons that I That I've been hesitant to use LMS in my work is that I love a good side quest like the way my brain works I need different challenges and little dopamine hits and like Go do this and then come back to this and go do that and then come back to that and you know approaching it differently

So like for me moments like that don't feel like a distraction or like a tedious You know think this log through before I can get back to what I was really doing Those are those are part of the fun for me. Mm-hmm And like the get history thing I think of as get stonking or being a get detective Which is you know, there's only little stories in my brain that make me enjoy my day But it's yeah, it's like a fun little exploratory jot for me So now I'm just picturing the the hard-boiled detective and the trench coat walking through the rainy streets of a get commit history to uncover what happened here More metaphors Of all the commit shots and all the world you had to walk into mine I feel like my metaphor for this is code archaeology On the topic of archaeology. I watched the original indiana jones and the raiders of the rut lost ark for the first time During this this season. I'll just slid out of their chair and died What did you think? First of all, indiana jones is a horrible archaeologist. This is I mean, it's true

He's more a treasure hunter than he is an archaeologist. Yeah, yeah, he's I don't think he really is respected by academic peers It's a little bit like i think it's heimlich schleeman the 19th century guy who discovered troy He was searching for it and you know the way he like ancient cities are They're like mounds because the you keep for building layers on top of each other And he was sure that like homeric troy was like further down And so like now what he's what happens is that archaeologist will like layer by layer like capturing all the information all that No, he just like dynamite his way down to the bottom Not interested in anything that happened before this or after this blow it all up. Right. Where's my hector and Achilles?

I'm sure we didn't lose any fascinating history that way He's a little bit supposed to be I think before he in the case speak indiana jones right the a little bit more in that tradition before archaeology may be Crystallizes as a more serious science. Sorry, but the superhero movies were like you saved Gotham But now Gotham needs 10 billion dollars in infrastructure rebuilding just because you say it belongs in a museum doesn't mean you're doing science That's good On that note shall be on that note the british museum we'd like to have a word with you. No, sorry Enjoy your get archaeology my friends Maybe we stick to the hard-boiled detective for this one Because that trope is no problematic No, no, no whatsoever and let's not dive into that at all. Let's just let's wrap up Guess we're taking a quick break and then on to season three see all them

Frequently asked questions

How long is this episode of The Bike Shed?

This episode is 37 minutes long.

When was this The Bike Shed episode published?

This episode was published on March 17, 2026.

Is there a transcript available for this episode?

Yes, a full transcript is available for this episode. You can read the complete transcript on the episode page.

Can I download this The Bike Shed episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.