Modern Web

In this episode of Modern Web, Danny Thompson chats with MelkeyDev, a Machine Learning Infrastructure Engineer at Twitch, about AI’s real-world applications, developer productivity, and the future of careers in Go. They cover everything from the rise of tiny AI-driven teams competing with large enterprises to how system prompts may matter more than model choice. Melkey shares his thoughts on cost-effective LLMs, production pitfalls, and the cognitive downsides of over-relying on AI. The conversation also explores backend development with Go, what makes it great for fast-moving teams, and how new developers can get started.Keypoints from this episode:- AI’s real value lies in business use cases. Melkey emphasizes that AI isn’t just a productivity tool; it enables small teams to build faster, cheaper, and more effectively than ever before.- System prompts are underrated. When it comes to LLM performance, prompt engineering often matters more than the model itself, especially for UI generation and agent design.- Cognitive cost of AI reliance. Referencing recent research, Melkey warns that overusing AI tools can reduce your ability to retain knowledge and perform certain tasks independently.- Go remains a strong backend choice. Despite being “boring,” Go continues to power developer velocity and scalable infrastructure, making it a smart language for backend-focused engineers.Follow MelkeyDev on Twitter: https://x.com/MelkeyDevSponsored by This Dot Labs: thisdot.co

In this episode of the Modern Web Podcast, hosts Rob Ocel and Danny Thompson sit down with Andre Landgraf, Senior Developer Advocate at Neon (now part of Databricks), to explore the evolving role of AI agents in developer workflows. They discuss how more Neon databases are being spun up by agents than humans, what that means for developer and agent experience (DX vs AX), and how tools like MCP and step functions are enabling scalable agent orchestration. The conversation also touches on agent security concerns, real-time vs. async UX, and how developers can build resilient, human-in-the-loop AI systems today. Plus, Andre shares practical insights from building his own personal CRM agent and experimenting with tools like Cortex and Ingest.Keypoints from this episode:- Agents now outpace humans in provisioning databases on Neon, thanks to agent-friendly APIs, early MCP support, and seamless integration with platforms like Replit and v0.dev.- Developer experience (DX) principles directly inform agent experience (AX), tools designed for simplicity and clarity often translate well to agent interactions, but agents still need unique guardrails like resumability and fine-grained permissions.- Agent orchestration is the next big frontier, with tools like LangBase, Ingest, and step functions offering patterns for chaining tasks, running agents in parallel, and retrying failed steps—enabling more resilient and scalable AI systems.- Async UX patterns are crucial for agent-powered apps, especially as LLMs become slower and more complex. Real-time feedback, task progress indicators, and human-in-the-loop controls will define effective agent interactions.Chapters00:00 Why apps don’t talk to each other 01:44 Meet Andre Landgraf from Neon 02:39 Agents now outnumber humans on Neon 05:03 DX vs AX: Building for agents 08:58 Security and authorization for agents 13:06 What’s missing for real adoption 17:06 Building a personal CRM with agents 20:04 MCP as the universal app interface 23:32 Agent orchestration and async UX 26:46 Step functions and background tasks 30:04 Are agents ready for real-time UX? 33:19 Human-in-the-loop patterns 35:59 Where to find Andre Follow Andre Landgraf on Social Media:Twitter: https://x.com/AndreLandgraf94Linkedin: https://www.linkedin.com/in/andre-landgraf/Sponsored by This Dot Labs: thisdotlabs.com

On this episode of the Modern Web Podcast, Rob Ocel and Danny Thompson talk with Brian Morrison, Senior Developer Educator at Clerk. They cover the state of authentication today, what makes Clerk stand out for small teams and indie builders, and how thoughtful developer experience design can make or break adoption.Brian shares why bundling tools like auth, billing, and user management is becoming more common, how Clerk handles real-world concerns like bot protection and social login, and why starting with a great developer experience matters more than ever.The conversation also explores the role of AI in software development and content creation, where it helps, where it hurts, and how to use it responsibly without losing quality or trust.Keypoints for this Episode:Modern auth is about experience, not just security. Clerk simplifies user management, social login, bot protection, and subscription billing with developer-friendly APIs and polished default UIs.Bundled platforms are making a comeback. Developers are shifting from handpicking tools to using tightly integrated services that reduce setup time and complexity.Developer education needs more care and creativity. Brian emphasizes the importance of visual storytelling, thoughtful structure, and anticipating confusion to help devs learn faster and retain more.AI is a productivity multiplier, not a replacement. The group discusses how AI can accelerate development and content creation when used with oversight, but warn against using it to blindly build entire apps.Follow Brian Morrison on Social MediaTwitter: https://x.com/brianmmdevLinkedin: https://www.linkedin.com/in/brianmmdev/Sponsored by This Dot: thisdotlabs.com

Tejas Kumar, host of The Contagious Code podcast and Developer Relations Engineer at DataStax, dives into the revolutionary Model Context Protocol (MCP). He explains how MCP enables seamless communication between AI apps and servers, essentially making AI interactions more intuitive. The conversation highlights real-world use cases like AI managing emails or booking flights. They also tackle the challenge of AI hallucinations and discuss the balance between user convenience and data privacy, envisioning a voice-driven future for online transactions.

In this episode of the Modern Web Podcast, Rob Ocel, Danny Thompson, and Adam Rackis sit down with Ahmad Awais, CEO and founder of LangBase, to talk about agents, context, and the future of AI-assisted software development. Ahmad shares the origin story of Chai.new, an agent that builds agents, and why he believes context, not code, is the true value layer in the AI era. The group unpacks how "vibe coding" is reshaping who can build software, why Chai isn’t just another AI assistant, and how agents might evolve into personalized, production-grade tools for everyone, technical or not. Plus: Tailwind analogies, Stanford lectures, sports nutrition agents, and a CLI that went viral in a hospital.Key points from this episode:- Ahmad Awais explains that AI agents aren't magic; they're just a new paradigm for writing software. What makes them powerful is their ability to act autonomously with relevant context, not just generate text.- Chai.new helps developers (and non-developers) create purpose-built agents without needing deep ML expertise. It abstracts complex concepts like memory, retrieval, and orchestration into an approachable interface.- Ahmad emphasizes that the real opportunity lies in agents tailored to individual users and use cases. Personal agents with custom context outperform generic ones, much like small teams beat massive frameworks for specific problems.- Chai and LangBase aim to bring AI development to the millions of engineers who aren't AI researchers. With tools like Chai, you don’t need a PhD to build powerful, production-ready AI agents.Follow Ahmad Awais on Social MediaTwitter: https://x.com/MrAhmadAwaisLinkedin: https://www.linkedin.com/in/mrahmadawais/Sponsored by This Dot: thisdot.co

In this episode of the Modern Web Podcast, Danny Thompson sits down with Reed Harmeyer, CTO of Skylight Social, and Brandon Mathis, React Native engineer at This Dot Labs. They unpack the technical and strategic decisions behind Skylight’s meteoric growth: why they built on the AT Protocol, how they tackled video discovery and scaling challenges, and how a fast-tracked in-app video editor gave them an edge.Keypoints from this episode:Skylight Social was built on the AT Protocol, allowing users to retain followers across platforms like Blue Sky and enabling creators to publish interoperable content in a decentralized social network.The team used React Native with Expo to achieve rapid development and cross-platform performance—launching a high-quality, TikTok-like video experience in just days.An in-app video editor was prioritized to reduce friction for creators, built using a native SDK wrapped with Expo Modules, enabling features like clip rearranging, overlays, voiceovers, and AI-generated captions.User behavior data—specifically watch time—drives content recommendations, not just likes or follows, helping Skylight offer a personalized experience while navigating scaling challenges from hypergrowth.Follow Reed Harmeyer on Social MediaBluesky: https://bsky.app/profile/reedharmeyer.bsky.socialLinkedin: https://www.linkedin.com/in/reed-harmeyer/

In this episode of the Modern Web Podcast, Rob Ocel and Danny Thompson sit down with Julián Duque, Principal Developer Advocate at Heroku, to talk about Heroku’s evolution into an AI Platform-as-a-Service. Julián breaks down Heroku’s new Managed Inference and Agents (MIA) platform, how they’re supporting Claude, Cohere, and Stable Diffusion, and what makes their developer experience stand out.They also get into Model Context Protocols (MCPs)—what they are, why they matter, and how they’re quickly becoming the USB-C for AI. From internal tooling to agentic infrastructure and secure AI deployments, this episode explores how MCPs, trusted environments, and better AI dev tools are reshaping how we build modern software.Key Points from this episode:- Heroku is evolving into an AI Platform-as-a-Service with its new MIA (Managed Inference and Agents) platform, supporting models like Claude, Cohere, and Stable Diffusion while maintaining a strong developer experience.- MCPs (Model Context Protocols) are becoming a key standard for extending AI capabilities—offering a structured, secure way for LLMs to access tools, run code, and interact with resources.- Heroku's AI agents can perform advanced operations like scaling dynos, analyzing logs, and self-healing failed deployments using grounded MCP integrations tied to the Heroku CLI.- Despite rapid adoption, MCPs still have rough edges—developer experience, tooling, and security protocols are actively improving, and a centralized registry for MCPs is seen as a missing piece.Chapters0:00 – What is MCP and why it matters3:00 – Heroku’s pivot to AI Platform-as-a-Service6:45 – Agentic apps, model hosting, and tool execution10:50 – Why REST isn’t ideal for LLMs14:10 – Developer experience challenges with MCP18:00 – Hosting secure MCPs on Heroku23:00 – Real-world use cases: scaling, healing, recommendations30:00 – Common scaling challenges and hallucination risks34:30 – Testing, security, and architecture tips36:00 – Where to start and final advice on using AI tools effectivelyFollow Julián Duque on Social MediaTwitter/X: https://x.com/julian_duqueLinkedin: https://www.linkedin.com/in/juliandavidduque/Sponsored by This Dot: thisdotlabs.com

In this episode of the Modern Web Podcast, Rob Ocel and Danny Thompson talk with Hannes Rudolph, Community Manager at RooCode, to explore how this fast-moving, community-driven code editor is rethinking what AI-assisted development looks like. Hannes breaks down Roo’s agentic coding model, explains how their “boomerang tasks” tackle LLM context limits, and shares lessons from working with contributors across experience levels.Keypoints from this episode:- RooCode's "boomerang" architecture breaks complex coding tasks into structured, recursive subtasks, helping AI agents stay focused while avoiding context bloat and hallucination chains.- Developers can build their own orchestrator and agent modes in Roo, tailoring persona and instructions to fit specific workflows—crucial for long-term productivity.- Unlike many tools, RooCode shows developers exactly how much each LLM call costs in real time, empowering teams to manage both quality and budget.- RooCode is deeply community-driven, with user-submitted PRs frequently reshaping priorities. The team emphasizes transparency, collaboration, and accessibility for contributors at all levels.Follow Hannes Rudolph on Linkedin: https://www.linkedin.com/in/hannes-rudolph-64738b3b/Sponsored by This Dot: thisdotlabs.com

In this episode of the Modern Web Podcast, Rob Ocel is joined by Danny Thompson, Adam Rackis, and special guest Coston Perkins for a lively discussion on the evolving role of AI in software development. The group swaps thoughts on everything from the rise of AI agents like RooCode and Claude, to what makes tools like Vercel’s v0 surprisingly powerful for frontend work. They debate Tailwind’s dominance as the styling output of choice for AI tools, unpack the implications of Shopify’s AI-mandate memo, and tackle the big question: will AI reshape team structures or just amplify developer productivity?Keypoints from this episode:- AI agents in everyday development – The hosts discuss how tools like RooCode, Claude, and Cursor are reshaping daily coding workflows, enabling everything from automated documentation to feature planning and refactoring.- Vercel's v0 is changing perceptions – Originally seen as a landing page generator, v0 is now appreciated for its live, code-focused interface, showing promise for serious frontend development with real-time editing and deployment.- Tailwind’s dominance in AI output – The conversation dives into why Tailwind has become the styling default for AI-generated components, and whether that’s a productivity boost or a future limitation.- AI’s impact on hiring and team structure – The group debates whether AI will reduce developer headcount or empower mid-level devs to produce senior-level output—suggesting AI may reshape team dynamics more than replace them.Follow Coston Perkins on Linkedin: https://www.linkedin.com/in/costonperkins/Sponsored by This Dot: thisdot.co

In this episode of the Modern Web Podcast, host Danny Thompson and co-host Adam Rackis chat with James Perkins, CEO of Unkey, an open-source API management platform built for scale, security, and developer simplicity. James shares the challenges of building globally distributed infrastructure, and why his team ditched serverless and TypeScript in favor of Go and servers.They talk candidly about the realities of API management at scale, how Unkey balances open source transparency with enterprise-grade performance, and what it takes to build developer trust—both as a brand and as a product.Keypoints from this episodeUnkey simplifies API management by acting as middleware for authentication, rate limiting, and security—without requiring deep backend expertise. It's designed for developers to go from idea to production with minimal setup.Go over serverless – James and his team initially explored TypeScript and serverless architecture but ultimately returned to Go and servers for better performance, scalability, and developer experience at scale.Open-source transparency is core to Unkey’s philosophy – The entire codebase is public, and the team maintains a radically open company culture, where even investor updates and customer support emails are shared internally.Customer obsession drives every decision – Regardless of whether a user is paying $0 or $2,000/month, Unkey responds quickly, prioritizes community support, and encourages a culture of ownership and responsiveness across the team.Chapters00:00 – Intro + Why Unkey exists02:00 – James' background and API pain points03:50 – What Unkey actually does05:45 – Engineering challenges + scaling architecture07:30 – Tech stack changes: Go, TypeScript, Serverless08:45 – Unkey as middleware: auth, rate limiting, analytics10:40 – Future vision: making APIs as easy as deploying on Vercel11:45 – Why Go instead of Node or TypeScript13:30 – Go vs TypeScript: hiring, dependencies, developer experience15:00 – Why API management is hard at scale17:15 – Case study: Fireworks and Google Apigee performance issues19:00 – The complexity of modern API platforms20:00 – Sponsor break: This Dot Labs20:35 – Will Unkey expand into app hosting?22:00 – Unkey's focus on doing one thing really well23:45 – Content strategy: personal brand vs corporate marketing26:20 – Customer obsession: internal culture and open company model30:30 – Open source dynamics and being fully transparent33:45 – Advice for developer-entrepreneurs36:24 – Wrap up + where to find the speakersFollow James Perkins on Social MediaTwitter/X: https://x.com/james_r_perkinsBlue Sky: https://bsky.app/profile/jamesperkins.devUnkey: https://www.unkey.com/Sponsored by This Dot: thisdot.co