49: Using AI Agents in Software Development 2026: Current Uses and Future Possibilities

Brittany Ellich (00:02) Welcome to the Overcommitted Podcast, your weekly dose of real engineering conversations. Myself and my fellow co-hosts, Bethany and Erika met while working on a team at GitHub and realized we were all obsessed with getting better at what we do. We started this podcast to share what we’ve learned. We talk about everything from leveling up your technical skills to navigating your professional development, all with the goal of creating a wonderful community where engineers can learn and connect.

This is a very special episode of Overcommitted where Bethany and I got to meet up in person at the GitHub office before attending the Pragmatic Summit together. So the format is a bit different from normal, but we hope you will all still enjoy this episode. Thank you.

Brittany and Bethany (00:49) Brittany, hello. Hi. We’re here in person. We’re in the GitHub office in person. We’ll have to take a little bit of B-roll of the office after this so that we can show it off. We’re gonna look like total nerds.

I mean, yes. Luckily, there’s not a lot of people, it seems like, who show up at the office. So, yeah. Yeah. But it’s great to see you. It’s so good seeing you. We just literally saw each other like probably, I don’t know, 30 minutes ago for the first time in years, like in person. Yeah. The last time, yeah, it was like a year ago, I think. No, was a year and a half ago that. Two years ago, I was in Raleigh. I feel like that was last time.

Shoot, I forgot about Raleigh. briefly. Yes, that is the last time. It’s long time. Good seeing you again. Good seeing you too. We are here for the Pragmatic Summit. Yes, it’s going to be super cool. ⁓ I can’t wait to meet everyone and see what the event is like. This is the first year that they’re doing it. Yes, yeah. And I’m stealing your question because you asked me this. What are you most excited about for the Summit? ⁓

Question, there’s a lot of big names that are going. So I think it’ll be cool to talk to folks there. It’s also, the number of people that have names that I recognize and the number of attendees, the ratio is very interesting. So it’ll be interesting to see if they actually, I don’t know, I also go to events and there will be some big people talking and then I think they just get overwhelmed because everybody gets all fangirlish. And so I never actually see them walking around or whatever. That’s true, that’s true.

Yeah, I agree. think like from the pragmatic engineer, ⁓ like he’s tailored a big like, ⁓ not just on the technology, but on the meta of it and how to effectively guide your career and how to how to think through engineering problems or like large scale things. And so I think I’m excited to see it through that lens or see what the lens is that

folks are talking about and just get an idea for where the industry is heading because it’s a crazy time. Yeah, unsurprisingly, I did take a look at the agenda today and it’s almost all AI. Who could have guessed? Yeah, oh my gosh. What? Oh man, wild. Yeah. Who could have imagined this? So speaking of AI, while I’m

I was flying all six hours, no, seven hours of flying. was noodling over what we could talk about today. And you’ve been talking a lot about how you use agents. And I’ve been definitely leaning more into agents lately. So I thought we could just have a conversation around agent use in the year 2026, like what that looks like, where we see it going.

I’m hoping to steal some from you because you actually work on this stuff. So my problem is I work on the back end of it. So I don’t I don’t really get the cool like forefront of what people are actually doing with these things. It very much feels like I am I am like just trying to make sure that it’s available for people to do that. Thank you.

Thank you for doing that because you make it so that we can use those tools, which is awesome. Happy to help. Yeah, we can do just a state of where things are at. February 2026 might be completely different by March 2026. It probably will be, Yeah, it probably will be. Yeah, before we get into that, I feel like we just need to like take a moment to like appreciate the current moment. think that there has been a lot of talk in the last week.

week and a half about like Opus 4.5 came out and… 4.6 just came out. 4.6, okay. 4.6 just did just come out. So I don’t know why, because 4.5 came out in what, like November? Yeah, I should know this. think. But yes, it was towards the end of the year, November, December. But it feels like there’s been a shift in the last like two weeks, basically. There’s been a couple of big ⁓ like tweets and…

things going that people have been sharing. Like there’s that one Andre Caparthi thing where he was sharing on Twitter where he’s like, yeah, in the last two weeks, I’ve gone from like 20 % AI generated code to like an 80 % me. And now it’s like flip-flopped. like the AI story has changed a lot where like, mean, I’ve noticed it too in my own work and play where like the things that I’m building, I’m like, oh wow, like it’s capable of coding.

just doing the coding tasks just as well as I am. And that’s weird, exciting, terrifying, everything. No, I agree. It’s been weird since coming back in January to work. I feel like I have not really written much code by myself. I’ve been on call a few times, so I haven’t necessarily had time to really sit down and like…

code things, but what I’m coding is usually agent-driven at this point, which is just so weird. It’s such a night and day difference between where I feel I left last year and where I’m starting this year. I’m not sure if it’s the same with you or if it’s something where you’ve been going down this path for a while, I feel like, no. I mean, I’ve been playing with this stuff a lot and trying to figure out where it’s useful and where it’s not for a while. ⁓ And it does really feel like a…

of switch flipped and now things are just, you know, they’re just much more capable than they were. I think a lot of it you can probably attribute to, ⁓ you know, models getting better. There’s also, I feel like there’s been a lot more practices that have, you know, finally sort of come together around like managing context and things like that. And just a lot of ideas too. I think a lot of it is just how much

you have to share ideas with other people that are also experimenting. then doing that opens your mind. So like, I could use an agent for this, or I could do this. And I think all of these things have sort of come together to where we are at this weird time. Yeah. Absolutely. And yeah, there’s a ⁓ lot of feelings that go with it. So yeah, let’s talk about it. Yeah, definitely.

⁓ I guess maybe let’s take a pause and talk about what is even the current state of agents. What is available, what’s out there ⁓ at this time? Because I’m sure it’s going to change in a couple months anyways. ⁓ So maybe like kind of documenting, I mean, what’s available for coding? What’s available for life agents? I feel like we see commercials all the time about like,

Agents can do this and I’m like, I haven’t really seen agents really come into my life yet, but I’m very interested I know you’ve been look kind of looking into ⁓ Agents for helping you like I think you were mentioning meal planning and stuff. Yeah, I would love to yeah if you got into that my gosh. Yes. I have a whole demo of like I’ve been trying to figure out How do I make an app that can do everything? for me, ⁓ like I

I want to bring all of the disparate context and things that I’m doing and managing and looking at into a single view. And so I’ve built this command center dashboard that does that. And it’s been really fun to work on. And that’s just for life things, work, everything? Everything. Wow. Everything. So when I’m like, I want to branch off and solve this problem, how do I build a UI into it? I’m very.

I need a UI for things. just cannot get behind. Sounds like you, yes. Yes. can’t just live in a terminal all the time. I wish that I could because I think a lot of people have mentioned it’s actually a little bit faster, especially if you’re orchestrating multiple things and have lots of code bases you’re switching between. In the CLI, it’s a little bit faster. ⁓ But that is not me. I can’t do it. Yeah. we have this power now to just run something locally.

have it do everything. Like if you have like MCP available, APIs available, like for anything that’s GitHub specific, I’ve got both the Copilot CLI and the GitHub CLI that I can reach out to. And so I’m like, okay, if I want to keep track of, you know, what PRs are assigned to me and what PRs are like, are my coworkers, you know, are ready to review for my coworkers right now. And just bringing that into a single view. And it’s been really fun to play with. Yeah. It’s like the Copilot LLMs like.

your inference behind that or are you using other APIs there? ⁓ As much as possible I try to use Copilot because it’s free. I mean not free, but it’s you know it’s included at work. There’s benefits to I’m dogfooding okay it’s okay. ⁓ Yeah so I try to use that. I also have a Claude subscription ⁓ but I mean that just I don’t want to spend you know 175 bucks a month or whatever on the really fancy one so.

I’ve got the cheap one and I keep hitting limits. ⁓ Inference is pricey. is cool that there’s options for open source and running things locally, which I think is a very interesting aspect to things. ⁓ It just feels like they don’t work as well as the ones that are hosted, ⁓ which I’m sure is intentional in some ways. Maybe not. Maybe not. mean, the open source ones, just like the,

the power that you have behind it. What I have that I can run on my MacBook is very different from what can run on some nice GPUs. Very true, very true. But yeah, so for agents, guess a couple concrete things. So I use the coding agent for coding. Makes sense. I have experimented a little bit with custom agents. feel like most of what I… It feels like the industry is more revolving around using skills instead of building out custom agents.

⁓ And I know skills.sh came out and that’s like a big like ⁓ sharing of skills which is really cool to see. ⁓ For custom agents there’s a couple that I made like for work for like repetitive tasks that I had to do like hey we need to you know I work in billing so like we need to make a new SKU or something like that then like there’s a bunch of different things that have to be updated for that so I built a custom agent to do that.

And it’s been awesome because that makes my life so much easier to not have to remember how to do this thing every few months. That’s interesting. I guess, how would you even define the difference between a sub-agent and a skill? Because I don’t think I have that ⁓ in my head. Because I know there’s tools. ⁓ So that’s very much tool calls where you say,

hey, edit this file or reach out to the web and things like that. ⁓ But a skill’s more high level than that, it seems. But then an agent’s even higher level, I’m guessing. But I’m not sure if you’ve done more research into skills versus ⁓ subagents. So I think that defining a custom agent is sort of like defining. So there’s a couple of different things. So there’s repository instructions, which you’re familiar with, or like the agents.markdown file that’s

becoming really common. So that is like the context for that repository that needs to apply all the time. And then the custom agents are sort of like the base context that gets applied when that agent is doing something. Whereas the skill is more like a very specific tool call where you’re like, hey, I want to give my agents the ability to use this skill. And so when I ask for this specific thing, it’s basically like defining extra context around a tool call instead of saying, instead of just

like having the agent figure out how to use that tool. You’re just saying like, this is how to use this tool. Essentially. Okay. All useful. Yeah. But I think part of the thing is like, you have to have a specific thing that you want to do and like a problem to solve. And from there you can figure out like what the right way to do it is. But I don’t think there’s also a single right way, especially now in this world where we have so many things available. There’s just so many more ways to solve problems, which is really cool. It’s true. Yeah, there definitely is. And I feel like

people don’t use agents the same. And I mean, there’s different ⁓ tooling for agents, especially with coding. Like you have the competitors of Copilot CLI, course, ⁓ Cloud Code, Codex, and all that. And they all do things slightly different. But it is very much like a opinionated platform on how to use agents. ⁓ And so I’m not sure.

You primarily use Copilot, is that right? Yeah, yeah, so I use the Copilot coding agent and I guess the review agent. Yeah, which I never I didn’t realize it was like an agent until it was, you know, that’s true. But yeah, so I use the coding agent and I use that a lot. So most of the time when I’m using I use it for two primary purposes right now. So one is when there’s a thing that I know that I need to fix and I know exactly how to fix it. I don’t need to like do a lot of

like exploring or anything like that. ⁓ And so then I’ll say, hey, know, coding agent, go make this change in this file that looks like this. ⁓ And it almost always does that like really well and writes tests for it and stuff. And then I don’t have to write the tests and everything, which is really nice. And then the other use case I’ve been using for it recently is more like exploratory. Like, here’s this large problem we’re having, ⁓ you know.

try to fix it and see what happens. And sometimes I’ll use what it actually comes out with and sometimes I don’t. And guess I can give like an actually more concrete example. like, we need to figure out how to, know, if it’s a bug fix, it’s like, you know, this.

thing is not using the correct number, it needs to use a different one, like go fix that. So that’s like a concrete thing. Whereas the more exploratory is like, want to solve the problem of when to create budgets or something like that, when to create these states at a certain time. I tell it more. One of the things I’ve heard recommended recently is like, tell it what the outcome is that you want instead of like how you want it to solve the problem.

which I don’t think is something that we, it has been like, that the most models have been and agents have been capable of until recently. It’s like in the past when I would do that, I mean, it would just, you know, go off and figure something out and be totally wrong and useless. But now it like, it’s pretty good at reading through and like tracing through the different calls and stuff. And so, yeah, I feel like that’s where I’m using it. What about you?

⁓ And I mean, to your point, think there’s, ⁓ lot of things that we end up on is like, how do we encourage people to code the way we want them or use the standards we want them to use? And I think half of the solution there is making it easy for LLMs to do what we want them to do.

So providing enough documentation and structures for LLMs to scan through and say, this is a pattern in this code base. This is how I should solve this problem. Or, there’s a direction on how I should solve this problem. I should use that. ⁓ And just making sure to on a lot of the instructions and such to guide ⁓ LLMs to doing the architecture that we’re trying to enforce in our code base. I like that. Yeah.

So it’s very interesting how documentation is still very important, but it’s almost more important for LLMs now than for humans. I mean, anyways, the LLMs can translate the documentation for humans and humans can just ask a question and basically interact with the docs that are written for LLMs. So it does make sense to almost write those for in a way that is easier to parse for them. Yeah.

I’ve heard one recommendation specifically to that that I’ve heard recently is to use mermaid diagrams, like a lot when you’re writing those docs and have the LLM come up with like a mermaid diagram of like what the flow is. And apparently they’re really good at parsing those and understanding them, which makes sense because it’s marked down. And you can contain a lot of context and information in a very small amount of like actual text, which is kind of nice. Absolutely. No, I haven’t.

whenever I’m reviewing a PR, sometimes I’ll say co-pilot, diagram out what this PR is doing, where it’s changing, to get a visual, as a visual person, it just helps to like visualize what’s going on and then map that to what I’m actually reading and see if that’s what’s going on. So it makes sense to use that to also guide it and add some, add that context back to it. Yeah. That makes sense. Nice. Yeah.

⁓ But the way I use agents right now is I’ve been using ⁓ Cloud Code more heavily lately. ⁓ I think hooks are just so interesting. And I think it’s really cool that you can specify basically how you want to work with it via these hooks. Like you can say, hey, notify me if you have an issue. Notify me if this happens or if something goes wrong. Run.

the lint whenever you’re done, run tests whenever you’re done, and things like that. And I think that’s a really interesting way of doing things. And I love that GitHub is very much trying to plug into all the tools so you can really work the way you want to work. But I’m very interested in the fleet ⁓ command. We were talking about this before we started recording. I mean, running agents in parallel has

been a very hot topic lately, ⁓ like using Get WorkTrees and ⁓ letting them basically work in their own branches and their own directories. Gastown. Yeah, I guess it is Gastown. And so I think it’s interesting that there’s been a lot of tools solving it. I feel, and I don’t know Codex too well, to be honest, but ⁓ Cloud doesn’t.

Implement it directly or cloud code doesn’t implement it directly. You very much have to like create your own Get work trees and have it branch out. It’s like you have to do the manual effort there. There’s a tool called Conductor I believe that can parallelize agents, but I feel like github or Copilot is kind of the first to actually integrate it into the agent, which I find really interesting Because it can almost like fan out and then pull together ⁓

that output. So I think that’s really fascinating. Yeah. excited to explore that more. Same. I think the thing that I’m most nervous about, getting multiple agents going at once. There’s occasionally times when I’m in agent mode and have a few things going, and it’s usually like, need to think about two very different contexts to make sure they’re not overwriting things or competing for changes. ⁓

And I just don’t know that I’m capable of having more than like a max of two work streams going at a time. But I do, I mean, I think developer productivity is very fun. It’s kind of weird too, to be like, I to see how good I can do my work. But like, it’s fun because I’m also using it a lot for like side projects and things like that. And building all these things is like, it’s very fun, fun testing ground to see how I can like.

do better at just development in general. I mean, I think we’ve been talking about productivity since we started working together. It’s true. Yeah. I mean, I think that’s how like you, me, John and Erica really bonded was talking about productivity and tooling we use. So it’s just kind of the evolution of that and trying to improve and get better. And I think like I’ve been reading, so I’m reading five coding by Gene Kim and Steve.

Yegge I’m so sorry. It’s okay, he’s not gonna watch this. We’re good. But I did find it interesting that a big part of their thesis is that you’ve got to adopt these tools otherwise you will get left behind. Yeah.

people coding games, like if they’re not adopting game engines. mean, yes, you could code your own game engine, but it’s fine if that’s what you want to do, I guess. But you might get left behind if you’re trying to get to market or if you’re just trying to make a simple game. So I think that it’s very interesting that it’s felt a little forced. ⁓ Developers.

have to use AI, but on the other hand, if it is a helpful tool and it does help you go faster or help build things better, it’s just an interesting dichotomy there. Yeah. Yeah, and think you kind of touched on it a little bit too, how you and your team are using agentic tools to improve code quality. I think that that’s going to be a really interesting space in the next couple of months. think one of the really common

arguments against using any AI for anything, it’s people are like, but it’s not secure and like the quality sucks and blah, blah. But I’ve noticed that these tools are actually very good at producing high quality things and secure things when you ask them to. You just have to know like, like, okay, like what are the security concerns with this thing that we just wrote or whatever. And I’m finding myself like, we go through and solve a problem and get it working. ⁓ And then the next step is like, okay, like,

think about security and maintainability and things like that. ⁓ And I do know that Copilot does is starting, it does a pretty good job too of like finding these things too. Like there’s a lot of times where it’s like, this could be a SQL injection concern. And I’m like, I’m glad that you’re bringing that up because I didn’t think of that. So ⁓ I think that like the frontier of like building all of those quality things in is something that most of the models are going to be, well, not the models, but the harnesses, I guess, are going to be considering. ⁓

And I’m excited to see how that changes things because I think there’s just so many… The thing that I keep coming back to is throughout my career, there’s never been like a shortage of things to do. You know, like there’s just never enough time. And now we theoretically have, you know, infinite time. You know, not really, but like, yeah, like it’s like the the time is not the constraint anymore ⁓ necessarily because these things work pretty fast. honestly, they’re

I’m not necessarily the world’s best coder at all, but I feel like they’re at least as good to me at this point. I ⁓ do think that there’s still a need for engineers to be focused on that quality side of things too, and make sure that we’re ⁓ validating that the code is correct as much as possible. I mean, we can use…

AI to help with that and assist with that. But I do think there’s still something to be said for engineers using our knowledge to make sure that we’re designing and developing things the way that they should or that we’re envisioning. Yeah. It’s a weird world. Yeah, it’s true. I mean, we are building software for people at the end of the day, ⁓ for the most part. ⁓ Who knows what the future will look like.

⁓ And that means that you have to use it and know what is the right feature to put in. I think that’s going to become more important too. It’s so much easier to add new features now, but we have to add the right ones. You don’t want to just overwhelm users with every feature imaginable. You also want to create the right experience that helps people accomplish the task that they want to accomplish with that software. I think that…

developers are going to be moving more towards building product management skills. Product managers are definitely building more technical skills. I know the product managers on my team are pushing almost as much code now, it seems like, as we are. Yeah, definitely. Yeah, those two roles seem to be converging a little bit. Yeah, that is an interesting point that they are very much converging. And then we’re almost… ⁓

managers of all these agents and such. Yeah, it’s true. I think there is something to be said about having some deep technical expertise, but for day-to-day things when you’re not solving those like super niche bugs or anything then I mean you’re just kind of making sure you know how to test things and validate that they work. So I know you were talking about you had this kind of

thesis on tools being async versus sync. ⁓ I’d love if you could maybe touch on that and talk through maybe some examples on what would be async tool versus an async tool when we’re talking about agents or LLM usage. ⁓ Yeah, so I think that I’ve been thinking about this a lot. I’m working on a blog post for it too. We’ll see if it ever ends up finishing. ⁓ I’ve been thinking about this a lot, that there’s just so many things coming out.

that I need to put them all into context of how do you actually use these things? And so I see agents as a very async tool. ⁓ so it’s this thing that I delegate out and it does the thing for me and then comes back to me. ⁓ Not to be confused with agent mode ⁓ in VS Code or CLI or whatever, in which case like…

It’s still, I guess, an agent, question mark. The whole idea of an agent is very ⁓ wishy-washy at this point, and we’re still defining it. ⁓ But that’s more of a synchronous thing, where I’m physically, not physically, I’m in the code base, exploring things. ⁓ So I think that just thinking about things, like, when a new tool comes out, what is this for? Is this for asynchronous tasks or synchronous tasks or whatever?

that helps me sort of ground how I’m thinking about new tools. Yeah, definitely. That makes sense. And I do think that, I think thinking of a lot of these newer agent tools like Cloud Code, Codecs, and such as async makes a lot of sense. And a lot of folks are trying to move in that direction of saying, hey, only ping me if you need.

something but and you probably end up wasting a lot of time if you’re watching every step that it’s doing. It’s Because it’s going to go back and forth and it’s checking its work and I mean if you’re using I don’t know have you heard about Ralph Wiggum mode? I have yeah where it’s just like a continuous loop basically yeah. Exactly I’m sure that gets tiring trying to ⁓ watch it go through that and validate itself so ⁓ it’s nice to kind of step away.

parallelize what you can. ⁓ So that makes a lot of sense. Is there anywhere that you’re using ⁓ agents outside of coding? yeah. I never talked about that. Truthfully, no. I have mostly used it for coding. I do use it a lot for like, I mean, I use AI, I guess, in general for brainstorming, but I’m not using agents for that.

So for instance, I used AI heavily because we were doing like a D &D one-shot with some friends and it was helpful for brainstorming out my characters. good for D &D. ⁓ So good. I got chat GBT to make images of my characters. It was beautiful. I was like, wow, this is so easy now. ⁓ But it’s not like I’m using an agent for that though. It’s just chatting with an LLM. Things that I would…

love and maybe I’m spoiling the next section a little but I Google introduces so many so much AI into everything except the calendar why oh my gosh that’s the one place I want it I don’t want it in my email I don’t want it in my Google Docs I just want it in my calendar Google please yes we are begging you this is my plea put it in the calendar yeah

Yeah, no, that’s valid. dark mode in the calendar before we got AI in the calendar. What is happening? anyways, love like an agent to schedule things because I live and die by my calendar. I have multiple calendars. I know Claude can connect to your calendar, but it only connects to your primary Google calendar, not to any infinite sub calendars that you may have. ⁓ And as any, you know, productivity obsessed person, I’m sure you have multiple.

calendars to organize your life. absolutely. I have one for me and my husband. I have one for working out. I’ve got one for ⁓ just me. I’ve got the Charlotte Hornets now, a calendar just for that so I know when games are. And I would love if like an agent can pay attention maybe to my like texts and see.

we’re talking about scheduling things. Like maybe I can recommend a time rather than recommending what I should say next. No, I know what I’m going to tell my friends, but like suggest a time. Yeah. Yeah. I like that. Yeah. That would be cool. Yeah. Google, get on it. Yeah. ⁓

Yeah, maybe we can kind of wrap up though with more like where we see agents going. We didn’t really plan a technical fun mode, I mean, or fun segment, but maybe we can take this to the extreme and maybe some realistic where we see agents going or where we want agents to go and like maybe some wacky ones. Yeah, that sounds good. You can go first. ⁓ she’s… Okay. So calendars, of course.

I just want agents to kind of abstract that tedium. I mean, I guess one thing I recently used an alum for that I would love a specific agent for is I’ve been playing ⁓ Expedition 33, Clarab Scare. But I take a lot of breaks and I forget what the lore is in between.

And I would love, and I know my husband like also will come back to games after a while and I don’t want to restart the game every time I like come back to it. want, so I just want to lore dump based on where I’m at. having an agent to tell me, give me like, dump the lore and tell me what I was doing prior to saving would be amazing. That’s such a good idea. Thank you. my gosh. Yes. There’s so many games that I’ve just abandoned because I’m like, it’s been too long. I can’t. Yeah.

I can’t come back to this. Yeah, exactly. have to almost… want to start over. Yeah. I mean, my husband has restarted several times, like other games, just because he can’t remember where he’s at, especially like the RPGs and stuff that are so lore heavy. Yeah. Yeah. And I mean, like, to Kendall’s credit, I don’t give Kendall very much credit, but I will give it credit for… It now has a chat where you can…

ask questions for where you’re at in the book and it won’t give you spoilers. Which was amazing. I was reading Three Body Problem and then the sequel to Dark Forest and there it gets crazy. I needed so like I would need so much I’m like wait who is this character again? Are they?

Is this something that they’ve answered or are they building up to it? Did I miss this answer for why this character did a thing or is it just like, need to wait and things like that? It was so nice to just be able to ask that and say, okay, no, we haven’t answered That’s cool, yeah. It was amazing, yeah. was a great addition and I think more story-driven things should have that because AI just is so good about…

searching a known body of work, a known body of context. And I don’t think we’re leaning into that as much as we should. Yeah. Yeah. That’s my soapbox. I love that. Yeah, somewhat related to that, I actually talked to somebody at work recently who was using, they like save all of their transcripts. And that gave me some ideas around like building like, I don’t know, like

coaching agents for conversations or something like that for work conversations. Maybe I don’t want AI to tell me how to do things. But also, there’s a lot of things that I think that are skills that you build as you’re building your career that aren’t super obvious. Where it’s like, okay, you could have done a better job mentoring somebody by doing this thing or finding opportunities within those conversations to do that. So I feel like.

searching those and maybe building up a body of those transcripts over time would help with those sorts of coaching agent conversations. Yeah, I love that. I also just want things to help with the overall mental load of being a parent and being a working parent. I bet. There’s just so many things I have to remember. Like, I’ve to remember the doctor’s appointments and the dentist’s appointments and…

you know, figure out like, wanted to get them to different places, like just coming here to this, my husband was out of town. And so, you know, figuring out like, oh, one of the kids got an ear infection yesterday. So I had to take her to the doctor and I had to take all of them to the doctor and then go get the medication and everything and like make sure that my parents had that. And like, there’s just so many little tiny things that I remember that I would like to offload somewhere. So I’m really excited about the future of that. I’ve been…

using Zapier quite a bit, which is just like such a great tool for building agents because like it’s literally, mean, it did agents before agents existed. This is not sponsored by them. but it could be. It could be. are easily bought. But you know, like just it was it was like the original connector app to connect between different things. And now

like the next logical step is like, okay, like just make that automated and make, you know, add LLMs to the loop and add these things. So I’ve been using that a lot for specifically for the podcast for things like, you know, researching guests and making sure that the schedule is set up. But I’m just excited about all of the different options there ⁓ and ways I can just make my life just slightly easier. Definitely. Yeah. That does it.

That sounds really interesting, especially, I mean, you were mentioning with childcare. If you could kind of like shift your brain to like your parents and say, like if you take notes with an alum and you’re like, okay, if you have any questions, you can ask this first and it probably has the answer and give it like access to that context that you personally have. ⁓ So if they’re in a pinch, they could use that. That’s really interesting.

Or, man, what if there was a thing that just was built into my, ⁓ I’ve got that tablet in my house, something built into that that’s in the same, where I view everything ⁓ that would manage medications. Just for like, my kid is on 10 days of antibiotics. Let’s make sure that we write down and check off when they occurred.

Because it’s always something that happens like very it’s very infrequent typically like my dog needs you know three weeks of medication or Whatever like there’s just so many living beings in my house where I am responsible for giving them medication And it would be nice if that would just be like something that would easily slot in when I needed it and disappear when I don’t ⁓ There’s just so many things the future is so cool the future is really cool, but I’m sure we’ll talk about this in the future and ⁓

update with what’s happening. This has been a great conversation. It’s so good to actually talk to you. I Wild. I know. And I guess stay tuned for more. Hopefully we’ll get to chat with some people tomorrow. Yes. And yeah, stay tuned. Cool. Yeah. Thank you for joining us.

Brittany Ellich (38:58) Thank you so much for tuning in to Overcommitted. If you like what you hear, please do follow, subscribe, or do whatever it is you’re supposed to do on the podcast up of your choice. Leave us a rating. ⁓ It helps a ton. Check us out on Blue Sky and share with your friends. Until next week, goodbye.