Hello and welcome to another episode of the bike shed a podcast from your friends at thought about developing great software I'm aunties later. I'm Jimmy and I'm Jimmy thickman and together we're here to share a bit of what we've learned along the way So Tess what's new in your world? So how I live in Albuquerque, New Mexico and yesterday I attended blast off rails which First I know locally Travis decided to create his own regional rail cuffs because a city rails and rails cuffs and last year and he was like well, maybe I could and It was cool because when he spoke in his intro He talked about how in the last year yeah like last off rails and Albuquerque and there was a blue Ridge Ruby I think in like North Carolina and then there was a RBQ. It was a fun name Yeah, so just a lot of and then SF Ruby so just a lot of new regional Ruby cons. It's really cool and yeah, like the first like talk was about how a lot of startup founders are choosing rails right now and
How rails is in a good position to be like a good like agentic framework Because of its of its conventions and about how you might plug those into the you can plug those in to the agents Which I definitely have thoughts about but it was really cool that there was like a new like local conference And I hope I have a well-known session next year. We'll see yeah That's awesome. There's been so many of those regional conferences springing up as things kind of get some distance from The thing that happened that sent us all into our homes for a couple of years And it's really great to see the community bouncing back like that That's an interesting take with rails being kind of good for startups because of a gen agentic reasons because that was kind of always rails is story anyway, right that is for startups because of you know speed to MVP and different things like that so it's interesting to see that kind of story for rails being like built on with this sort of
New universe that we're in now. You said you had some thoughts on that are there any that are easily encapsulatable right now? Not not too much. I think it's a I I do agree that like rails definitely could be Whether like definitely like and like hesitant to see like you know Action or action agents like whatever railsy name I'm like I kind of don't want to see that added to like rails like like the base rails, but yeah, yeah Yeah, yeah, it does it does seem like that falls a little bit outside the kind of 80 20 threshold that rails to to be right Like not a really is gonna necessarily need that so Yeah, seeing that in core rails would be yeah I did that determines me of action table Yeah, that's true. They've been floating those you've been precedence, you know that don't follow that role so that's true So I do with the world my world to say Scotland that's true. Well, okay. Yeah, we're gonna or open that can you know
I could have said something about work But it's actually more exciting that right now Scotland is in the World Cup and the World Cup is happening in the US, Canada, Mexico And I'm not a huge soccer fan Necessarily, but I'm really enjoying the community and the excitement that is in the country right now for this event That hasn't happened since 1998 and so we're going to a lot of like watch events that are just sort of out and about and getting to meet folks and share in the excitement and the energy and that has been That's been really fantastic. Whether it's a sport that I've followed or not just the energy around town is kind of electric. Oh I guess that just reminded me just a fun fun thing I've seen online is a folks who are visiting the United States for the World Cup and they're they're like European and I saw a Skeet that was Somebody who's German posting a picture of a buckies and was like this is a gas station in the United States
And he was just like posting and it's like oh my god the Europeans are discovering buckies This is so fun. Yeah, yeah to see some folks from around the world discover the like American kind of We do things big here Over the top, you know And ranch dressing this is this is the thing that I hope comes out of it that there's been it then posts of Europeans discovering ranch dressing Cuz we don't have that over here and so I think enough people come back with the love of ranch dressing We might get some ranch dressing over here and I would be very happy with that Little off the beaten track, but How about you Jimmy what's new in your world? Oh I recently visited Norway not work related, but that was a beautiful beautiful country over there more related to AI I was thinking about so I've been using it for a bunch of stuff You know coding and then personal stuff like which vacuum am I gonna buy help me decide or like trip planning?
And I've just had clod like panda to the max effort for months and I realized that might not be the web That's way to do it two things because it's sending 20 minutes to like decide on a vacuum to buy and it's just overthinking so many things like small code changes CSS tweaks I just noticed to yesterday that yesterday they have ultra code which is even above max and when you turn it on I like make sure whole terminal like a rainbow effect temporarily which is Just even more of your overstimulating us that we need But yeah, so I've realized okay Maybe we don't need to use like the max effort for everything and just like using high which is I think the default or even like a lesser Model so like don't just you know go if they will come out I'm going to use that for everything with ultra effort You know these tools are getting so advanced I think it's easy to go down the rabbit hole like overthink things You know like 20 times and it's going to be a better output But it might not always be a better output that might just slow you down so yeah, there's like a I don't know It's like a push my pole. We're kind of figuring out how to visualize these tools Yeah, I feel like that's sort of a phase or a learning
Plateau or something that I've had to go through my own journey as a developer is When to overthink things and when to actually just use kind of what's right in front of you, right? I think that's a step towards becoming a senior developer is knowing when it is appropriate to explore some of those rabbit holes and And really kind of dive in and put that that max effort or that ultra effort onto something and then when that's actually detrimental Right and it's using too much time and too much of your your great matter resources and stuff So I wonder if there's kind of a similar arc that we'll go through with the working at these new tools and things Yeah, that has come up quite a bit in many of my conversations at work It's come up in conversations here already and I think that is partially some of the reasons why I thought bought has started a new initiative of a task force With some of our members that is the reason why the two of you have joined me here because you're part of that task force And wanted to I wanted to hear a little bit about what's going on on the inside and I think Folks who listen to how stories about how we work here at thought about that
I would also be interested to know what kind of conversations are happening in the company around AI and so you're both on the AI ethics Taxforce, is that's the name that's been settled on great? Yes Tell me a little bit about how that got started and what you're trying to do. Yeah, I guess Kind of towards the beginning. So it's June right now in like the beginning of the year there was a lot of internal chatter About oh does like thought but have like a stance on like a company stance on AI usage and like like what does that mean? Because there's so many different lenses you take with that like, you know, we have thought waters like it's not about like You know being like oh, we want everybody all of our consultants to use these tools or It's not that like well, it's just another tool in the toolkit similar how to like thought about never mandated that everybody used VIM So, you know, it's just a similar tool to that. There's also conversations about our clients
You know if our clients are using AI in their development workflow and their expectations But also the kinds of products that people are coming to with like coming to us with like they're like, you know Solo founders who want thought bots help and they've you know vibe vibe coded this entire app and they're like, okay Let's help me take a market thought bot and you know, there's so many different like perspectives And I from my perspective the task force is like, okay Maybe we can like get together and kind of hammer out like does, you know, what our stop-offs talk about this Yeah, I think it became clear that AI is like omnipresent. It's like coming at us in all directions from clients from Pressure to the end in the industry just to use it everywhere github dev workflows So I think that kind of really rooted the need for some entity to Like what test was saying figure out how are we going to use AI thought bot and like ethically workflow wise Just like what makes sense? What are we comfortable with?
Do you think some of those conversations spread from what we're seeing from other places in the industry of some companies either Mandating AI use or token maxing and leaderboards and all of these different things that We're kind of seeing pressure to Participate in these ecosystems or use these tools whether or not they seem to be the right fit with those the kind of conversations that we're driving This like that's what kind of sparked in my memory as you were as we're talking about that Yeah, I think so I think there was also part of it was just the feeling that it was all moving so fast and like everything was just like oh like these could these conversations be different a month from now and some feeling between the different Developers at Bob bot because we have a whole spectrum of like folks were like really leaning into it folks are really leaning out of it And I don't have anything to do with it There's kind of a whole spectrum and kind of on both sides of the spectrum folks were feeling like they were alone And we didn't know how everybody felt like oh, maybe people felt like unsupported or felt like alone or felt like oh
No, if I don't use AI I'm gonna be left behind by the company or some folks There was a lot of like oh, I only see negative talk internally and like am I a bad person for like using these tools? You know and I think you kind of all wanted to come together and just to like talk about it. Yeah, that's really great I love that the kind of starting point came from like communication and understanding and that's all beyond the same page and let's kind of share What these different perspectives and where these different camps and viewpoints and everything where we're all starting from because we yeah Exactly. We might be overlapping. We might be missing each other more or less than we think we are and without actually establishing that up front It's it's hard to have a conversation if you don't agree on what's real in the first place for sure Yeah, I feel like it would be easy to just like further silo like people are really into it and the people that are you know I was comfortable with AI but I feel like the thought-bought way that we would see that and then be like no We're gonna tackle this at on yeah, absolutely I think that's something that's a strength of thought-bought right is we talk about this the internally the thought-bought hive mind and for there to have not been some
Conversation really that was like what is the thought-bought stance? How do we as a collective feel about this for there to have not been an answer there did feel like there was a bit of a gap Right and so I'm glad that the initiative has been taken to start to fill that in a little bit What are some of the ways that the task force is going about answering some of these questions or addressing this head on like you say? Yeah, we've asked the company how we feel about AI like what has it been like working with clients and you know dealing with their expectations on AI we've put out a survey that we're currently like feeling the responses on so the goal to is to put out some external stuff like Blog articles on our learnings and kind of how we're addressing some of the concerns We also have like our internal handbook where we plan to put some of this like concrete the concrete takeaways and then thought-bought also runs consulting training for everyone I thought about to join and the idea is to put some of these like to me or some of the real-life scenarios You know, we've had our situations We've had with clients and the consulting training is tricky talking with clients that are some are technical some are not technical
So there's a fine line and you want to be respectful and be able to respond to you know, there are expectations on AI Are there any of your personal experiences that the two of you've had with client you're talking about clients and expectations that have sort of led to your Wanting to participate in this tax force? Yeah, well, at least for for me, it's kind of almost like the opposite because I Originally joined the task force because I'm very much like not into Using these tools and all the clients I've been on there was no like Expectation that that I use them and so I wanted to help shape get my perspective and help shape I've got my opinion about this and just in the last month I've been on a new clients and there the first client I've been on that's like pedal to metal like all of our tickets all of our P.R.s all of our communications are filtered through Claude and this has been new experience for me I've like like literally this this client is the first time I've installed Claude on my on my work laptop and just like okay
What's it's skill? Okay? I need to have still loaded so that way it doesn't like completely go off the rails and there's some kind of like semblance of normalcy and It's definitely been a hard adjustment. So it's kind of like like I'm glad I'm on the task force now I have personal experiences I can lead to this but yeah, that's kind of my journey so far Yeah, I guess for me I've been using it like to assist coding development since it sort of first existed a little bit You know like first phase was like send a little chat GPT little code snippet and ask for help and then we just progressed to know Where it's like full-on agents warms doing the things but I was just interested in the task force sort of around I just want to be a little like better talk to clients about AI and like set their expectations solidly in reality while being nice about it You know there have been a few projects where we've had to like readjust recalibrate expectations With you know some non-technical clients some technical clients where they see chat GPT
Which is a very built out thought out product And they think that's a thing that you can just extend and put into the world You know it's a lot more involved chat GPT is a product that has a bunch of different pieces strapped onto it It's not just a model So it's just kind of recalibrating and like helping these people understand like what is an L and what is a model? And like how do you extend functionality of a model? There's a lot of like elegance maybe that goes into explaining that well to the clients that aren't as technical as we are You know because we're working in the stuff all day Yeah, I guess that that kind of like your experience Jimmy is like there's like too kind of almost like client expectations There's like client expectations around the kinds of products that like can be built So like adding a like L M or chatbot to their to their product You know like like oh we're a health care company We want to answer questions and so build this on to here and like sending expectations around that And then there's like the other side of it of like client expectations around the code development workflow and
Like folks who are maybe like product people and then they're like hey I want to write this issue and it like spits out the technical requirements into like a GitHub issue and it's like Okay, well, yeah, it's like there's like what kinds of products are able to be built but also expectations around I am on a client and their Development workflow is integrated with AI like do I need to get on that bandwagon? And even if the product that we're developing is not AI but they use it in their workflow Yeah, I think there's been some tension with with our Developers and like client developers on how much AI use is going on like sometimes We're very thorough in our PR reviews. That's just a thought thing I think it's great and with AI for a lot of folks just in trust AI to just do the thing do it correctly and not require really any Of the why it were a good PR description, which I think really is something that I as a developer would see and I push back on I want to know why I want there to be a trace that go back and reference Yeah, I think there's a larger conversation going on I hope or around the industry of if we're going to be using these sort of tools
Like how do we not lose all of the lessons that we've learned over the last decades of software development about how to do this Well, and like the things that are baked into what thoughtbot is that we would be doing with or without You know, agent coding How do we make sure that we're not losing that as we're kind of going into this new world and like where does this new world fall short of those things that we hold as values to building great products and Navigating those not only with clients, but also amongst each other too. Yeah, I think there's a lot of kind of emotion Actually, that's tied up in these conversations a lot maybe a little bit of like tribalism my side likes these things Your side doesn't and our lizard brains fall very easily into those sort of traps Are there any of those conversations that are you've anticipated in the work or?
Experienced with the work in the task force or are trying to maybe get on the front of like how has that emotional aspect to this conversation entered into the work I remember for our like survey that we sent out Like my first like kind of draft that I wrote up was like sent out and then I got some feedback saying like hey, you know, you're saying that you want all kind of like all kinds of experiences using AI But the way I had frames of the questions were all like the questions were framed in a negative way You know, I didn't like recognize that but it's something from the outside being like oh, you know You're not gonna get somebody saying like a neutral or positive experience when you're framing it as like so what were the problems?
You had like what kind of arguments you have to clients? Yeah, yeah, this sounds great right everyone's just kind of mad about it Okay, so yeah, that wasn't my intention, but definitely it's hard to write Neutral, I mean, that's why it's a whole job to write neutral kind of like polling questions Or you'd have polls that are like slanted in one way or the other just based on a couple of words and So I was like, yeah, like that's like a nuanced thing, you know to create even just like polls or surveys or yeah So has there been anything in this survey or in conversations in the task force so far?
That's surprised you something that was either unexpected from the get-go or that has maybe Changed a little bit of your perspective of your mindset so far I almost want to say this might not be surprising But it's sort of surprising that we have I've read through some of the results of our survey and some feedback is from like perspective clients perhaps using AI to even like write PRDs, which is so surprising to me Because I feel like that's something I want to know like how it's written if I want to build a product and pay, you know To build this product it's surprising. Maybe how much it's just being trusted Something blindly by by parties that you know truly want to build a thing. There's another aspect where AI is kind of a yes man I think that has led to a lot of a fan sick Exactly I think that's led to a lot of people really thinking their ideas unique and special and has a secret sauce Whereas before AI you'd ask your friends and family and they'd be like, I don't know about that, you know Interesting yeah, yeah that kind of reminds me of I heard this story of there was like a
Professor who like just like pulled out a pencil and was like this this pencil name is Tom like hi Tom and just like pretending Tom was like a real thing and then at the end of the lesson the professor snaps the pencil in half and Everybody in the class gassed. How could you do that to Tom? And he was like and he's like remember this is a pencil and it's just like it was an example of how also as humans we imprint and are like oh this thing is like talking to me even though like LLMs and models are you know statistical Datasets and it's a fancy autocomplete. It's not a person But because it uses language it uses the language that we use we imprint upon it as humans and I think it's a hard folks to like not do that And so that's why people like implicitly like trust the outputs or more are more willing to trust the outputs even if it might be
Made up or hallucinated because it's not a real sentience thing. It's autocomplete I was just thinking off of Tess's point like you know We're there like push things forward build new products that don't exist yet and elements are trained on things that do exist and Great through with like AGI You're just kind of like spinning the wheel looking at a different way. Right. Yeah, that's why I thought that is like partially That's like our special we help clients build new products that don't exist So it's almost counterintuitive in a way to use these these elements to like try to build out some new unique idea Yeah, yeah Yeah, I think in the way that these tools are marketed and spoke about by the company who's Most likely to benefit from their adoption that are encouraging these these ways of thinking and encouraging this trust like that It is your PhD level friend or expert talking to you and Being very disingenuous with their capabilities and that's that's something I think that we as thought by we as developers are gonna come up against
often of people kind of personifying these things and It's very easy to do because partially because of the way that we're wired but partially because of the way that they're being presented to people who are Non-technical folks right where not only not aware because they're not kind of versed in the underlying technology But also like are not are not interested in that right? You don't need to know how your car works to use the roads and and everything like that You don't need to I mean that's part of one of the beauties of rails right? It's like you can get something off the ground as a fairly early career developer and you don't have to know all of the underlying things But you can still make good use of this tool and so yeah, that's a that's a conversation that I think we're gonna be having Are having and are going to be having a lot more and so I'm really happy to hear that That's part of what the the test force is going to be kind of investigating and like having suggestions learning sharing about these conversations to To make our all our thought butters more views and better evaluate those conversations with clients and prospective clients
I was gonna say like I know a lot of a lot of folks are also trying to be like Thought about is always kind of try to put into place like best practices and to be you know, like thought, you know thought we'd read things and I think I definitely see value in folks here trying to you know create skills that are and like There's one that's like so credit review that thought bought made and it's like the intention is you know To like rubber duck or like work back and forth with it You know to not let it do everything and to like use it as a tool But not to like take over and I think that thought about isn't a in a position given the history of it to be like Thought we'd read in that in that respect which would hope you could lead to like better outcomes on like the technical side and I know I just keep thinking of like uber in the last like six months where they blew through the entire years tokens budget within like by April like within the first quarter that the
They used at the entire token budget for the year because of like token maxing and all those things And I wonder if there will be like a pullback once folks realize right now extreme token usage is very it's being subsidized Because it's like any product that gets into a fight where it's like uber itself where they went into a market and they undercut You know taxi companies by offering cheap rides and then once folks got used to cheap rides and put taxis out of business then they raise the prices because now they actually have to turn a profit because they have investors and once the inch education of tokens and AI comes to fruition Maybe there will be you know more conversations about okay How do we get the most benefit with the least amount of tokens like those will be the new conversations?
Yeah, I feel like we're living in some massive like feedback loop right now We're like building a lens as to like where this fits best like in the product development The life cycle like where do we use AI them as efficiently? Where's it maybe not the best tool for the job? Yeah, well, I can definitely say that I am very interested in hearing more about the conversations that have come out of the the tax force and looking forward to reading the blog posts that you were mentioned and I think Hopefully now we'll have a lot of listener ears ready to turn into reader eyes That's weird But we're also going to be looking forward to more of what's to come and yeah our kind of collective Conversation on this omnipresent part of our lives now and so thank you so much for joining me today I'm sure we will be amongst to the hosts of the bike shed continuing this conversation ourselves and Being and involving a lot of the work that the task force is doing and maybe we'll have Yourself's or someone else from from the group come and join us again in the future, but let's wrap it here
Thank you so much for joining me. Thank you