Claude Code podcast mentions

Quote

Before you write your first like line of code, right? You should have like a test framework. What are you gonna test? How are you gonna do it? And that's just like first principles engineering of like, okay, I I want to make sure I've covered everything. can I even when you get to hardware right like it's like I have some i squared c lines I I I want to simulate Each of that so that the driver is not there. Now it's just like ninety-nine percent of the time is your test bridge. Even with Claude Code and so I when I start with a plan, I expect Explicitly ask it for like, okay, can we write down more tests? Can we take, you know, like for this ISA And use that as a test set. Every one of them should pass in every case. for every commit that you have to make. Right but if it if like you know the ability for a human to read takes a lower priority. Yes, of course at some point we'll be like, okay, you know, um the robots are taking over and We need to go check the code. That's fine. You know, we'll we'll deal with that at that time. But um But right now f uh in my personal space I treat all code like ephemeral, right? Every c every line of code.

Quote

tools and your own preference. So there's pretty rich, I would say there's like six or seven tabs in the user settings where you can Session returns. So super rich features here. And I would say the whole like multiply player multi agent. So the multi agent I've used a lot. So I've I'll put like twelve Agents in parallel, a mix of codex and claude code, mostly some Gemini. You know, um but so multi agent works awesomely well. It's super great. Uh So single player agor is fantastic. That's part of the reason why I was able to build it. But in such little time I built like an app with a lot of product surface very quickly. works extremely well the multiplayer stuff we have yet to see how it's gonna work out. You know, what happens when you in Invite a QA person to come see the live app that's essentially your dev. environment and let them you know prompt against it. So that's pretty interesting. What's gonna emerge here? Another interesting shift that I can anticipate. Participate when you have that agentic session being long running and multi user where it's not just one person prompting, giving feedback. Steering the agent until you decide that it's done and then shutting it down.

Quote

product called Prethink, which gives them essentially a pre-computed architectural context and understanding of their code base before they even start. And we started to see Some of the results. It's four times faster reasoning, fifty five percent fewer tool calls, and then thirty percent fewer tokens used. Then the other piece is how do we give it a little bit of a little bit of a them that visibility. It's another product called Changelog, which gives them the awareness of what's already in flight across the portfolio. So they're not duplicating work or sequencing it. wrong and how we're giving these agents all these tools is through MCP. So think about it as your Claude Code, your copilot, pilot or whichever agent you're using or that the customer is using. But going back to what we were saying, that deterministic harness, we're giving the tools to your agent, whichever agent you're using you're using through MCP to search better, get better context, and get awareness so that you can essentially execute deterministically. Again, super Super, super cool and completely geeking out. Every time you say deterministic, the skepticism, the AI skepticism in me as an engineer. really starts to kind of melt away there. So with the tooling you're building for agents, what does the future look like? So about what you offer, how you're gonna Where are you going to take this to the next…

Quote

Um I mean maltbot would be another example of a genetic um systems that have uh I mean agentic is also like a I would say almost like uh not well defined term because uh people use the term differently. But um for this uh podcast maybe we can think of agentic as something that runs Uh and I think um yeah that is something we will see more of uh Recently Claude Code and uh GPT five Five point three codecs, the app, they added a lot of these tasks where you can even schedule something. And I think uh we will see. see more of that. It's just like the beginning. It will be more like plug-ins and I mean it's still the same LLM. It's just like how we use the LLM and um how to Where companies though that uh like you know open. May I Claude, they are more like okay, let's um build these tools so we can actually do more and more impressive bigger things with these LLMs. And I think maybe by the The end of the year we will have uh systems that can reliably um book a

Quote

actually get it from multiple email sources for myself because I have multiple different emails that I use which is very cool. Um but that's that's the same thing. looking at final outputs. So if I go over to My Claude here. Um Claude co-work and And Claude uh and Claude Code both have a really cool schedule. type stuff. And here's another example of me putting the same prompt where I ask Every Friday ask employees to send what projects they worked on so that uh during the week so that they can track there is by project and then it says from whom right similar thing and And it creates a scheduled task, as you can see, that has all this in here. that I can turn on and off, right? Um it's an active automation that starts Next Friday. And if I press run now, it'll actually run and send it now. If I go back here to Um what I just made. Give me a second. Oh yeah. did was answer the question, right? As to who the people were. It created the tasks and then

Quote

So this was a very subtle. What you apply multiagents to matters tremendously and how you use them. like the take that apply multi agents to problems that are easily paralyzable that are read only. For example, context gathering for deep research, and do like the final quote unquote right. in this case report writing at the end. I think this is trickier for coding I did find it interesting that Claude Code now allows for sub agents. I still think I actually kind of agree with Walden's take, it can be very tricky in the case of coding if sub I think that's a well. explained uh contrast and comparison. Not much to add there. I think um it's interesting that they have different use cases and different architectures evolved. I don't know if that's a permanent thing that that might Yes. We should probably talk about the same thing. Yeah. uh it has been in the air for now many years, obviously well before L One thing I found pretty interesting is

Quote

Skills and connections were now tied to your anthropic account, not to your and terrible in this one use case. So first thing I did is slap the control into the settings.json on to make sure it doesn't do that anymore, just for data isolation reasons. But yeah, ha n that is not something most people are going to do. They're not going to build a box that does this. Claude Code gets very greedy with the RAM. So this is a couple hundred month bucks a month for the EC2 instance. to sit there running uh four or five of these things. And there needs to be a better solution. I'd love to have something like that on the laptop since I'm mostly around when I want this stuff doing Yeah. Last weekend I did open claw in a box is a uh a repo I put together. this and you wanna have like constraints and you wanna have like network shut down on certain interactions. So So the use case I use is on Sundays, I go through my email, it labels a bunch of stuff for me. only it times out at two hours. So if it can't get it done in two hours, then it's probably this it's it's too long. Or I need less email. So like I can figure that out. I'll go call the government and tell them to sit stop setting up. E email. Or whatever the uh I'm thinking of the Seinfeld, uh when when uh Kramer said no more. mail. Not sure if that's possible yet. Yeah, I have a uh in the next room I have a

Quote

DailyBrief.ai. It's gonna have links to pretty much everything we got cooking from free education programs to our AI operators community to the new newsletter that is where you can find All the links that I talk about in the show, to of course our companion experience page. We did one of those for our How to Use Opus 47 and the new Codex app episode on Friday. Today's episode combines a number of the big themes that we've been talking about this year. 2026 kicked off, with everyone coming back from the holiday break, really understanding the first time. and grappling with just how big an advance we had gotten between the harnesses like Claude Code and Codex, and like the models like Opus 4.5 and the GPT-52-5-3 series. Ever since then it has just been a race and it has been very clear that there is a massive difference and a growing difference between the people who are not going to be able to do that. people who are best at using AI and using it most fully and those who are not. Now clearly a huge amount of that work is running through code, even when the ultimate output is not building Software turns out that when you give AI and agents the ability to write code, it in fact unlocks a huge amount of other capabilities that are relevant for knowledge work way outside of just software engineering. When OpenAI announced their new Codex app, they said that a full 50% of the usage…

Quote

Really arguing about the harness for a while, then harness being less and less important, you know. So I think in this day and age no one can truly tell what's the even the rights there's probably no rights or best, but what has been different reasons why someone would choose maybe a factory? versus codecs versus something else that you've seen in your user base. Well look one is policy, right? If you're at a company, they might only have a few tools that you're allowed to use. That's one driving factor. factor. You know, another is just comfort and familiarity. If you started working with Claude Code or with Codex. You know, these things, for lack of a better term, have some personality. There's like a style. You get used to it. So like if I am communicating with Claude and I switch to codecs which I I do use codec sometimes, it has a different feel to it. Um so whatever you get comfortable with, I think. there's a tendency maybe to stay there. It's why I stay with Claude Code, for example, is I'm just very comfortable with it. There are folks who will model jump, like they're looking at the benchmarks and they're switching immediately. Like, you know, right now I think five four has the has the edge and it will change. Right. But overall I do feel like you know, whereas before they were making huge leaps on top of one another. and it was worth my switching back and forth. No…

Quote

Uh has AI c like changed that product thinking at all or like how you view uh what the product can and should do. Definitely. Definitely. Uh we haven't settled on any solution to that yet. Uh like there's no AI you know sparkles icon in the in the product yet. But we have tried to do things that will help developers. Like I'm a big fan of Claude. I use Claude Code like all the time, every day. And you know, having uh a markdown export of an error, for example, so that it has the backtrace and it has the context like that's why we have a button in our UI to give you that to your clipboard right so you can just paste it right into Claude. So definitely it has impacted like some of the design decisions we've made and some of the workflows that we're thinking about. And it has definitely impacted our wish list for what we want to do future. Like I would just this morning I was thinking like what's the ideal experience? with AI and honey badger. Is it something like I I log in and I'm like, hey, what What's going wrong with my app today? And then that's the question that I type into a box. Right, do do we want to have like a chat interface and then then it goes through and crawls my data and surfaces like The most frequent or the most user impacting, or you know, these are kind of big hairy questions and this is why we don't have that sparkles button yet because we haven't quite fi…

Quote

And wouldn't you know it? My personal leaderboard for ninety nine percent of the days over the last couple of years, saving me. Countless hours. But as you've probably heard, Claude is the AI for minds that don't stop. It's the collaborator that actually understands your entire workflow and thinks. With you. And with Claude Code, I'm now taking ready. Support to a whole new level. Claude has coded up its own tools to export, store, and index the last five years of my digital history. From the podcast and from sources including Gmail, Slack, and iMessage. And the result is that I can now ask Claude to draft just about anything. Anything for me. Research and write outlines of questions. Based on those, I asked it to draft a dozen personal And to promote the show, I asked it to draft a thread in my style for the first time. Featuring prominent tweets from the six guests that booked a slot. Drafts. Not because they're bad, but because it's important to me to be able to fully stand behind everything I I publish. But still, this process, which took just a couple of prompts once I had the initial setup completed. easily saved me a full day's worth of tedious information gathering work.

Quote

a custom rectangle selector, things went much better. Is this much better than I could do by hand? And sure is. End quote. Okay, who's next? Ken Thompson? Twenty twenty six is the year of self-hosting. Here's Jordan Fulgum. Quote I've wanted to self host a home for years, but I've always bounced off it. Too much time spent. That changed recently because CLI. Agents like Claude Code make self-hosting on a cheapo home server dramatically easier. and actually fun. This is the first time I would recommend it to normie slash software. Software literate people who never really wanted to sign up to become a sysadmin and stress about uptime. This resonates with me. to personally edit every config, schedule system updates, manage security settings, etc. etcetera, I'm excited about self-hosting again. Could AI agents usher in a golden age Formatting JSON for compact human Quote Most JSON libraries give you a choice between two formatting options Options. Minified JSON is very efficient but difficult for a person to read. Most beautiful