Fragmented - Android Developer Podcast

You already know how LLMs work from our popular 20-minute explainer. Now we take it to images. What does Michelangelo have to do with stable diffusion? More than you'd think. Walk away knowing how image generation actually works — and what it has in common with the text models you already understand. Full shownotes at fragmentedpodcast.com. Show Notes Episode 303 - How LLMs work in 20 minutes - text generation VAE -Variational Autoencoder RGB Color model - wikipedia Word2Vec technique - wikipedia Efficient Estimation of Word Representation - original Word2Vec paper by Mikolov et al. High-Resolution Image Synthesis with Latent Diffusion Models - Rombach et al. (2022) — the paper behind Stable Diffusion Image Training data LAION-5B - 5 billion image-text pairs scraped from the web, used to train many image generation models WebLI - Google's internal image-text dataset Michelangelo Get in touch We'd love to hear from you. Email is the best way to reach us or you can c

The hard part of AI coding isn't generating code — it's controlling quality, safety, and drift. Kaushik and Iury break down harness engineering: the five pillars for shaping an agent's environment and what it looks like when teams build custom harnesses from scratch. Full shownotes at fragmentedpodcast.com. Show Notes Why it matters Harness Engineering - OpenAI's post on building their Codex codebase (~1M lines of code, 1,500 PRs merged, zero manually written) Shaping the harness The Feed's Lost and Found - Iury's newsletter consolidating harness engineering themes Agent legibility Closed feedback loops Persistent memory Entropy control Blast radius controls Building the harness Minions: Stripe's one-shot, end-to-end coding agents - Stripe forked Goose to build custom agents for their codebase Goose - open-source coding agent from Block Superpowers by Jesse Vincent - skills that enforce a proper software engineering process Open Code - open-source coding agent you can fork and cu

AGENTS.md is becoming the common language for AI coding tools, but keeping repo rules, personal rules, and tool-specific files in sync is still messy. In this episode, Kaushik and Iury break down the sync problem, compare their own setups, and unpack what the latest AGENTS.md research actually says. Full shownotes at fragmentedpodcast.com. Show Notes The sync problem AGENTS.md - Official spec Custom instructions with AGENTS.md - Open AI Keep your AGENTS.md in sync - Kaushik's post Rulesync - What Iury uses Tweet by Ryan Carson and Claude frustrations Other links Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents? Harness engineering - Check the section about using AGENTS.md as a table of contents OpenCode Get in touch We'd love to hear from you. Email is the best way to reach us or you can check our contact page for other ways. We want to hear all the feedback: what's working, what's not, topics you'd like to hear more on. Contact us Newsletter Youtube

Subagents are becoming a core primitive for serious AI-assisted development. In this episode, Kaushik and Iury disambiguate "agent" terminology, unpack plan mode vs subagents, and explain how parallel, scoped workers improve research quality without polluting the main thread.Full shownotes at fragmentedpodcast.com.Show NotesResources & DocumentationOfficial DocumentationAgents, Modes, Subagents: official harness docsClaude Code SubagentsGemini CLI SubagentsOpencode SubagentsCursor SubagentsAntigravity Agent ModesAOE ScoutingResearch Papers & ArticlesIntroducing Claude Opus 4.5Deep-Research Agents PaperPost: GPT-5 System Card by AlexXuSelf-Driving Codebases Blog -multi-agent systems making 1,000 commits/hourGet in touchWe'd love to hear from you. Email is thebest way to reach us or you can check our contact page for otherways.We want to hear all the feedback: what's working, what's not, topics you'd liketo hear more on.Contact usNewsletterYoutubeWebsiteCo-hosts:Kaushik GopalIury Souza[!fyi] We transiti

Agent Skills look simple, but they are one of the most powerful building blocksin modern AI coding workflows. In this episode, Kaushik and Iury break down whento use skills, how progressive disclosure works, and how skills compare withcommands, instructions, and MCPs.Full shownotes at fragmentedpodcast.com.Show NotesMain ReferencesProgressive Disclosure -how skills are loaded into contextAgent Skills Open SpecificationAAIF (Agentic AI Foundation) -Linux Foundation initiative for AI interoperabilityNeedle in a Haystack Problem - original"Lost in the Middle" paperAgent-Invokable vs User-Invokable -merging skills and commands in Claude CodeCreating SkillsSkill Creator -Anthropic's skill for creating new agent skillsClaude Code frontmatter referencesee model: * & context: forkUsing other SkillsAnthropic Skills GitHub Repository -official collection of Claude skills and examplesClawdhub - Clawdbot's skill hub. All versions arearchived hereSKILLS.sh - Vercel's skills hubWarnings before installing random skills[

Ever get asked "how do LLMs work?" at a party and freeze? We walk through the full pipeline: tokenization, embeddings, inference — so you understand it well enough to explain it. Walk away with a mental model that you can use for your next dinner party._Full shownotes at fragmentedpodcast.com.Show NotesWords -> Tokens:OpenAI Tokenizer visualizer -Visualize how text becomes tokensTokens -> Embeddings:RGB Color model - wikipediaWord2Vec technique - wikipediaEfficient Estimation of Word Representation -original Word2Vec paper by Mikolov et al.Embeddings -> Inference:Word embeddingTemperature, Top-k, Top-p sampingGet in touchWe'd love to hear from you. Email is thebest way to reach us or you can check our contact page for otherways.We want to hear all the feedback: what's working, what's not, topics you'd liketo hear more on. We want to make the show better for you so let us know!Contact usNewsletterYoutubeWebsiteCo-hosts:Kaushik GopalIury Souza[!fyi] We transitioned from Android development to AI starting withEp

MCPs are everywhere, but are they worth the token cost? We break down what Model Context Protocol actually is, how it differs from just using CLIs, the tradeoffs you should know about, and when MCPs actually make sense for your workflow.Full shownotes at fragmentedpodcast.com/episodes/302.Show NotesMCP - Model Context ProtocolRemote MCP server example - GleanAAIF -Agentic AI Foundation setup by Linux foundationGithub MCPGithub gh CLIPlaywright MCPContext7 MCPAnthropic's announcement onAdvanced Tool UseTipsIury: use ast-grep to structurallysearch code fasterKG: use agent-browser by Vercel to give browsingpower to your agentGet in touchWe'd love to hear from you. Email is thebest way to reach us or you can check our contact page for otherways.We want to hear all the feedback: what's working, what's not, topics you'd liketo hear more on. We want to make the show better for you so let us know!Contact usNewsletterYoutubeWebsiteCo-hosts:Kaushik GopalIury SouzaWe transitioned from Android development to AI starting

Most folks reference "AI coding" like it's one thing. It's really not. In this foundational episode Kaushik & Iury walk through (at least) four paradigms — from super autocomplete to agent orchestration — each with different workflows, expectations, and mental models.What do most developers follow today? Where is the frontier? What's coming in the future?Listen to the episode and find out!Full shownotes at fragmentedpodcast.com.Show NotesGen 1: Super autocompleteIntellisense - regular autocompleteGithub CopilotCursor TabGen 2: Chat Oriented ProgrammingCursor IDEFirebenderGen 3: AgentNvidia's definition of an AgentReAct PromptingChain of Thought was a prompting hackDeepSeekDeepSeek - R1 paperTUI tools (or Harnesses):Claude CodeOpen CodeCodex CliGemini CliIDE style toolsCursor AgentCopilot (MS)Junie - IntellijAntigravity - GoogleHeadless Tools:Jules - GoogleClaude Code on the WebCodex WebGen 4: Agent OrchestrationGit worktreesTipsIury: Transfer between agents using your own compact commandKG: Ask the agent

Fragmented is changing. New direction, new cohost. Kaushik explains the pivotfrom Android to AI development and introduces Iury Souza.From vibe coding to software engineering — one episode at a time.Full shownotes at fragmentedpodcast.com.Contact usNewsletterWebsiteContact usYoutubeCo-hosts:Kaushik GopalIury Souza

Join us as we talk with Vinay Gaba, Android GDE and leading voice in Android development, about the future of the field. Vinay shares insights from interviews with top Android devs on their three-year predictions, and offers his own perspective. We cover AI's impact, evolving development roles, and crucial future skills.You can find the full shownotes over at fragmentedpodcast.com.Show NotesVinay's personal websiteDispatch - Jetpack Compose newsletterIssue #11 - Future of AndroidDevCursor AI IDEJetbrains AI assistantJetbrains Junie - the coding agentPessimists avoid risk, Optimists change the world.Contact usfragmentedpodcast.comContact usYoutubeKaushik:kau.sh (links to everything)ThreadsBlueskyYoutube@kaushikgopal

In this episode, we dive into the power of rapid prototyping for Android developers using Kotlin. We explore how this crucial skill can impress stakeholders, accelerate your workflow, and help you stay ahead in today's fast-paced tech landscape. We'll cover use cases across scripting, web development (with Ktor & HTMX), mobile apps (Jetpack Compose), and even touch upon how AI is changing the game!You can find the full shownotes over at fragmentedpodcast.com.Show NotesSimple scriptingState of Kotlin Scripting 2024 - Jetbrains blogKotlin mini-app repo playground-ktAdvent of CodeWeb sitesHugo static site generatorCloudflare PagesSlashdot effectFragmented Podcast Website powered by Henry (Kaushik's custom theme)Web appsKtorHTMXAjaxCSS TransitionsServer Sent EventsYoutube Video showing Ktor + HTMXMobile appsEpisode 252 talking about playground-androidFlutterKotlin Multiplatform KMPReact NativeLLM Based appsStreamlit.Snowflake acquires StreamlitGoogle AI StudioKotlin AIVibe coding an AirBnb cloneContact usfrag

In this episode, we dive into the programming paradigm — Data Oriented Programming (DOP) and why making data the star can simplify your code. Learn how well-modeled data reduces defensive logic, prevents invalid states, and keeps your apps stable. We’ll also contrast DOP with Object Oriented Programming (OOP) and Functional Programming (FP), sharing practical examples, tips, and resource links to deepen your understanding.The full shownotes with illustrations are on fragmentedpodcast.com.Show Notes[announcement] Fragmented has an email newsletter.Programming ParadigmsObject Oriented ProgrammingFunctional ProgrammingData Oriented Programming (DOP)UUIDRFC 4122Kotlin now includes UUID in the standard lib even for kotlin multiplatform.ResourcesBrian Goetz's seminal article on DOPTies Van de Ven Advanced Kotlin Dev Day in 2022Data Oriented Programming in Java by Chris Kiehl (by Manning Publications)Devoxx talk by Nicolai Parlog on youtubeContact usfragmentedpodcast.comContact usYoutubeKaushik:kau.sh (links to ever

In this episode, discover how Dan Rusu’s pods4k Immutable Arrays library can deliver 2–8× speed boosts and 5× lower memory usage in Kotlin/Android apps. We first revisit the fundamentals of autoboxing/unboxing and immutability to understand their impact on performance. Then we hear from Dan himself on his library, motiviations for building it, how the benchmarks were calculated and much much more. Our grand finale episode for 2024. Hope you enjoy it!Full Shownotes at https://fragmentedpodcast.com/episodes/254.Show NotesImmutable Arrays on githubImmutability episode #66 on Immutability with Ryan HarterJMH - Java Microbenchmark HarnessImmutable Arrays Benchmarks page for Immutable Arrays (pods4k)Dan's post - Kotlin avoids entire categories of Java defectsK2 compilerDan RusuWebsitepods4k github discussionsContact usfragmentedpodcast.comContact usYoutubeKaushik:kau.sh (links to everything)ThreadsBlueskyYoutube@kaushikgopalDisclaimer: Links shared might be affiliate links. They help support the production of Fragm

Kaushik looks at a new logging library from Square called logcat. He starts by seeing how the popular Timber library does it along with the benefits. He then interviews Pierre-Yves Ricau (Piwai) of Square, the creator of logcat, to explore its origins and advantages.You can find the full shownotes over at fragmentedpodcast.com.Show NotesTimberlogcatMotivations in READMECompiler plugin issuePlayground android app demonstrating logcatbitdrift.ioDesign of everyday thingsPrevious episode #191 on loggingContact Piwai:WebsiteBlueskyContactYou can find us on a few places:fragmentedpodcast.comContact usYoutubeKaushik:kau.sh (links to everything)ThreadsBlueskyYoutube@kaushikgopalDisclaimer: Links shared might be affiliate links. They help support the production of Fragmented. Thank you for your support.

In this episode of Fragmented, Kaushik dives into the importance of creating your own starter template to streamline app development and minimize decision fatigue. He shares insights from his own starter template - Playground Android.Looking to the future, JetBrains has an exciting tool called Amper that might make all of this much easier. Kaushik chats with JetBrains’ Márton Braun about Amper, an exciting new tool that could revolutionize Kotlin & Android project setups.Tune in to learn how to go from idea to code with less friction!You can find the full shownotes over at fragmentedpodcast.com.Show NotesPlayground AndroidThe Architecture Templates (blog post)Inspiration (Other starter templates):Bloco's starter template what i referenced a lotDonn's project - Jumpstart Android if you care about quickly getting a Rails app integrated with Android in a hybrid experienceNow in Android too much functionality for my tasteJake's u2020 for some good times nostalgiaPlayground Androidgradle version catalog - BOM

In this episode, Kaushik explores the evolution of dependency injection (DI) in Android development. Dagger has been the de-facto solution for DI in Android but there might be a new king in DI-town. He also chats with friend of the show and dependency injection expert Ralf Wondratschek for a final gut check.Shownotes: https://fragmentedpodcast.com/episodes/251

We're back from the hiatus with our SemiQuicentennial episode! With the momentous 250 comes some big announcements and a shift in the way we do things.Listen to find out the details!Shownotes: https://fragmentedpodcast.com/episodes/250

In this episode of our podcast, we explore the diverse landscape of Java versions within the Android ecosystem. Our guest is Michael Bailey, a seasoned Java expert who has been a frequent presence on our show since the early days of our podcast. We kick off with a solid foundation, discussing the differences between JDK and JRE, as well as the distinctions between the available Java JDKs. We also guide listeners through Android Studio settings, exploring how to select a suitable JDK, its utilization, and how it relates to JAVA_VERSION on one's home path/terminal.As we dig deeper, we start to unpack some of the crucial Android app settings. From compileOptions to sourceCompatibility/targetCompatibility, we shed light on why these versions are important. We also demystify the compileSdk vs minSdk vs targetSdk, and how they interconnect. Drawing from Kaushiks's recent experience in building a new app, we provide real-life examples that can better clarify these topics for our listeners.We conclude the episode by

In this edition of Fragmented, we're thrilled to host Ishan Khanna, a software engineer at Tinder who possesses great enthusiasm for feature flags and A/B testing. Donn discusses why he invited Ishan on the show, highlighting Ishan's passion for feature flagging and A/B testing. The conversation kicks off with an insightful story from Ishan about feature flagging at Booking.com, leading to a discussion on the difference between A/B Testing and Feature Flags, when and why to introduce feature flagging, and how to measure its effectiveness. The show also focuses on the benefits and risks of feature flagging, along with ways to manage potential complexities in the codebase.We then delve deeper into the topic of feature flagging, covering how to get started, what to look for in a tool, and the role of testing. Discussion points include the best practices for rollout percentages, considerations for multi-platform implementation, and the specifics of targeting in feature flagging. The conversation wraps up with an

In this episode, we talk to, Siggi Jonsson. Siggi helps guide us through the complex and often confusing world of Android app modularization. Our conversation will begin by shedding light on the importance of modularization, what triggers the need for breaking projects into multiple modules, and how modularization relates to team size, feature teams, and code ownership. This discussion is aimed at demystifying how and why developers often choose to modularize their applications.As we move forward, we'll dive into more advanced aspects of modularization, such as the challenges of scaling and the potential missteps in over-modularization. Drawing on real-world experiences, we'll examine the red flags that signal a need for revising your modularization strategy. This part of our discussion will also highlight some specific tools and techniques that can help manage and visualize your project's modularization, and Siggi will share his own experiences and insights into how these tools have assisted in his modulariz