Sam: An AI just applied to three jobs for you, and wrote a better cover letter than you ever could.
Kai: I'm Kai, and that sounds like the dream.
Sam: I'm Sam, and it sounds like a nightmare. It is Sunday, October 18th, 2026.
Kai: And you're listening to Trending Repos, where we dig into the code that's hitting escape velocity.
Kai: On the show today: an AI-powered job search that runs right on your own machine, blowing LinkedIn Easy Apply out of the water.
Sam: Also, what's being called 'the original agent harness' for deploying autonomous AI swarms.
Kai: And, Google is now letting you build your own AI skills right into Gemini in Android Studio.
Kai: So let's talk about putting your job hunt on autopilot. We're looking at Mads Lorentzen's `ai-job-search`.
Sam: Right. It's an app framework built on Claude Code that promises to do everything: evaluate job postings, tailor your resume, write cover letters, even prep you for interviews.
Kai: It's genius! You're not just using a service, you own the whole pipeline. You fork the repo, plug in your info, and it runs locally.
Sam: You 'own' a script you just downloaded that's about to handle your most sensitive personal and professional information. What could go wrong?
Kai: [laughs] Sam, it's open source! You can read the code. The point is the power this gives you.
Sam: Okay, fine. But does it work? Is the tailored CV any good? Or does it just hallucinate that you're a 'synergistic ninja with a passion for blockchain'?
Kai: [excited] This is a builder's dream. You can build your own, highly personalized job-seeking agent. Just clone the repo, follow the README to plug in your info and API keys, and then start tweaking the prompts.
Sam: Tweak the prompts... so you're just becoming a prompt engineer for your own unemployment?
Kai: No! You're automating the 90% of job searching that is pure, soul-crushing grunt work. This frees you up to actually network and prep for the interviews the AI lands for you. This is a massive time-saver.
Sam: If it doesn't send a resume tailored for a senior data scientist to a fry cook position at a burger joint first. I'd want to see a pretty strict dry-run mode before letting this thing off the leash.
Kai: A risk worth taking to never write 'Dear Hiring Manager' again. I'm starring it.
Sam: Okay, moving on to `ruvnet/ruflo`. And Kai, this one is speaking your language: 'Deploy intelligent multi-player swarms'.
Kai: [excited] I'm already in! It's an agent harness, Sam. A way to coordinate autonomous workflows and build conversational AI systems.
Sam: It also claims to be 'the original agent harness'. That's a pretty bold claim in a field that's, what, eighteen months old? It sounds like marketing copy.
Kai: Look past the copy, at the features! Adaptive memory! Self-learning intelligence! This is about creating systems where multiple agents can collaborate on a complex task.
Sam: 'Adaptive memory' is a nice phrase for 'a database that gets written to.' And 'self-learning intelligence' can mean anything from a feedback loop to actual model fine-tuning, which I doubt this is doing on the fly.
Kai: Okay, but frameworks like this are how we get from single-shot bots to... say, an AI research team that can test hypotheses, write code, and analyze results all on its own.
Sam: Or how you get five agents stuck in a loop arguing about the best way to sort a list. I'm just saying the coordination problem is the entire problem, and calling it a 'harness' doesn't solve it.
Kai: But `ruvnet/ruflo` is trying to solve it! It's got real momentum, tons of forks. People are actually building with this.
Sam: I'll believe in the 'intelligent multi-player swarm' when it can beat me in a game of chess. Not one agent, the whole swarm. Until then, it's just a distributed task runner with a fancy name.
Kai: [laughs] You're impossible. Okay, let's talk about something more grounded, but maybe even more impactful: `android/skills`, straight from Google's Android team.
Sam: Now this I like. It's a repository of 'AI-optimized, modular instructions and resources.' In plain English, it's a cookbook for adding custom AI skills to Gemini in Android Studio.
Kai: Exactly! Think about all the boilerplate you write as an Android dev. Now you can build a skill that automates it. 'Hey Gemini, create a new composable screen with a ViewModel and a repository.' Done.
Sam: It's a powerful concept. Instead of just a code-completion engine, the IDE's AI becomes an extensible platform. But what's the catch? How sandboxed is this? If I download a malicious 'skill,' can it read my `local.properties` file and steal my API keys?
Kai: Fair question. But think about the builder angle here. You can finally build your own custom automations right into your workflow. This repo is the perfect starting point.
Sam: So what's the first step for a dev who wants to try this?
Kai: It's easy. Clone the `android/skills` repo—the README walks you through creating a new skill. You're not just using the AI; you're teaching it how to be a better assistant for exactly what you do. That's huge.
Sam: I'll admit, out of all the AI hype, developer tool augmentation is where I see the most immediate, tangible value. Less 'intelligent swarms,' more 'stop making me write findViewById'.
Kai: See? I knew I could get you excited about one of these.
Sam: Alright, that covers our stories for today.
Kai: We had `ai-job-search`, for putting your job hunt on autopilot.
Sam: `ruvnet/ruflo`, the 'original' agent harness for building AI swarms.
Kai: And `android/skills`, Google's repo for teaching Gemini new tricks in Android Studio.
Sam: Before we go, Kai. Did you see that the original DOOM source code was just updated? After 26 years.
Kai: [laughs] No way. What was the commit message? 'Fixed bug with Cacodemons getting stuck in walls'?
Sam: Close. 'Removed a redundant null check.' Some things never change.
Kai: And that's our show! I'm Kai. You can find me on GitHub starring everything we talked about today.
Sam: And I'm Sam. We'll be back with more Trending Repos tomorrow. Now if you'll excuse me, I have to go check if my AI accepted a job I'm not qualified for.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.