Kaushik Gopal, Iury Souza
269 episodes listed below
PRESS PLAY
Fragmented is an AI developer podcast for engineers who want to go beyond vibe coding and ship real software. We cover AI-assisted development the way working engineers actually use it: prompting strategies, code review, testing, debugging, workflows, and building production-grade software with AI tools. No hype. No "I shipped a SaaS in a weekend" stories. Just tactics that work. Hosted by Kaushik Gopal and Iury Souza — software engineers using AI daily to build and ship real products. From vibe coding to software engineering — one episode at a time. Our goal: help you use AI to become a better engineer, not be r...
Based on 6 episodes we listened to · September 2026
Fragmented connects AI-assisted coding and Android engineering through concrete questions about tools, feedback loops, performance, and maintainable software.
Fragmented follows software engineering down to the decisions that determine whether an idea survives contact with real code. Kaushik Gopal and Iury Souza move between interviews and host-led conversations, asking how tools behave, where their costs hide, and what developers…
Even an array's capacity check gets cross-examined; your Kotlin shortcuts may want legal counsel.
Press play if you want
Short reviews from the PlayNext desk, based on the episodes we processed.
A clear, grounded case for sub-agents that keeps the hype under control.
The piece treats sub-agents as a working pattern, not a slogan. Kaushik and Yuri make the distinction cleanly: the main agent keeps the reasoning thread, while smaller agents spread out to gather context and return useful findings.
A clear map of AI coding's progression, from autocomplete to agents and thinking models.
The episode treats AI coding as a sequence of shifts rather than a single breakthrough. Kaushik and Yuri move from generation one, the super autocomplete era tied to GitHub Copilot, into agent loops and thinking models, which gives the conversation a clean sense of movement.
A grounded look at how on-device AI, privacy, and hardware could reshape Android work.
Vinay Gaba's Android outlook puts the focus where it probably belongs: on-device AI, privacy, and the hardware shifts that may follow. The conversation keeps returning to a simple question.
A practical case for shipping prototypes fast, with Kotlin and HTMX as the working example.
Kaushik makes a clear case for rapid prototyping as an engineering advantage, especially when momentum matters more than polish. The piece lands best when it moves from a broad defense of speed into specific stack choices like Ktor and HTMX, and when it sets the scrappy…
A clear case for modeling software around explicit outcomes, with a brief detour into neighboring tools.
The episode treats data-oriented programming as a way to make complexity manageable by spelling out the outcomes a system can take. The API example is the strongest part: success, failure, timeout, and interrupted are all given a place in a sealed class, so the control flow is…
A clear reminder that logging stops being boring the moment a real Android device fails.
Logging looks dull until an Android app breaks on the one device you cannot easily reproduce. This episode treats that problem plainly, starting with Timber and then showing what `println` turns into on Android, how it lands in logcat, and where production logging starts to get…
A practical look at keeping agent instructions aligned across tools.
Keeping agent instructions aligned is presented as a real maintenance problem, not a stunt, and Fragmented treats it that way. Kaushik and Yuri move through the global-versus-project split, dotfiles habits, a separate agents repo, and a sync workflow that cuts down on repeat…
A practical look at the setup that makes AI coding agents useful.
The episode treats harness engineering as the real story and keeps the focus on what makes AI coding agents actually useful. The strongest thread is simple: can the agent see what changed, respond to it, and keep going?
Long-form interviews on AI coding agents, infrastructure, and security, plus a recurring SED News roundup.
Founders and engineers on AI coding agents like Claude Code, and supply chain security.
Stack Overflow's Ryan Donovan interviews founders and engineers on AI agents, protocols, and dev tools.
Michael Kennedy interviews Python maintainers on packaging, security, web tools, and AI.
Michael Kennedy and Brian Okken run down the week's Python tooling, typing, and packaging news.
Matt Stratton interviews DevOps practitioners on security, platform engineering, and trusting AI-written code.
Fragmented - AI Developer Podcast episode 304 focuses on agent skills and opens by contrasting them with agents.md, cloud.md, MCPs, and slash commands.
Fragmented episode 306 focuses on keeping agent instructions in sync across tools like Cloud Code, Codex, Gemini CLI, and Cursor, with Kaushik and Yuri contrasting global user instructions with project-level agents.md files.
Kaushik opens Fragmented with a claim that Dan Rusu’s Immutable Arrays can deliver a 2× to 8× speed boost and about 5× lower memory use in Kotlin.
Kaushik and Yuri take on the question developers keep hearing in social settings: how do LLMs work?
Kaushik and Yuri pick up where episode 303 on how LLMs work left off and move into image generation.
Kaushik and Yuri unpack harness engineering as the next step in how people work with AI developer tools.
Hosts Kaushik and Yuri interview Mitchell Hashimoto about his work at HashiCorp and his new terminal app Ghostty, focusing on AI-assisted development.
Hosts discuss Andrew Karpathy's quote on maximizing agent autonomy and contrast it with the vanity metric of token maxing.
This episode of Fragmented dives into sub-agents, explaining them as isolated execution units with fresh context windows, specific system prompts, and scoped tool access.
In this episode of Fragmented, hosts Kaushik and Iury discuss MCPs (Model Context Protocol), explaining them as standardized protocols that integrate external tools and data sources into AI models.
Fragmented podcast hosts Kaushik and Yuri map three generations of AI coding paradigms: Generation 1 is probabilistic autocomplete (GitHub Copilot) replacing rule-based IntelliSense - trained on vast codebases enabling function-level generation from comments.
The host announces a pivot from Android development to AI-assisted software engineering, addressing the gap in practical AI content for working developers.
Kaushik talks with Vinay Gaba about where Android development could be heading over the next three years.
Kaushik frames rapid prototyping as an essential skill for Android developers and argues that Kotlin can support it across scripting, websites, web apps, and mobile apps.
Kaushik opens with a note about the newsletter before moving into data-oriented programming.
Kaushik talks with a guest about logging as the unsung hero of Android development.
Kaushik discusses the overwhelming number of technology choices Android developers face when starting a new app and advocates for creating a starter template to overcome decision paralysis.
This episode discusses the evolution of dependency injection in Android development, comparing Dagger with newer solutions like Kotlin Inject and Kotlin Inject Anvil.
An interview with Android developer Joachim Borges about his career and the evolution of Android development.
Interview with Ralph Wondertek, creator of Anvil, discussing how the library simplifies Dagger-based dependency injection in Android development.
An interview with Valera about the state of Android testing, discussing his background and current practices.
The hosts discuss a controversial tweet by tech journalist Ben Thompson about Android app quality and developer laziness.
An AI developer podcast discussing practical tactics for using AI tools in software engineering.