We track Product so you don't have to. Top Podcasts summarised, the latest AI tools, plus research and news in a 5 min digest.

Hey Product Fans!

Welcome to this week’s 🌮 Product Tapas!

If you’ve been forwarded this or just stumbled upon it, you’re in for a treat. For the best reading experience, check out the web version and sign up for future editions here.

What’s on the menu this week? 🧑‍🍳

📰 Not Boring – Apple's iOS 19 looks set to be the biggest overhaul in years with a complete design refresh and a virtual health coach that might just eat other health apps' lunch. Meanwhile, OpenAI appears to be positioning ChatGPT as "the internet's front door" according to a leaked strategy doc, while A16z declares SEO dead and GEO (Generative Engine Optimization) the new kingmaker. Elon's cooking up a storm with both Tesla's Robotaxi service launching June 12 and XChat taking aim at WhatsApp's dominance. Oh, and Mary Meeker's back after a 5-year hiatus with a 340-page AI state of play!

⌚️ Productivity Tapas - This week's tools will supercharge your workflow – from bringing Claude directly into your spreadsheets to building AI-powered UI components with just a few clicks. Whether you need a complete AI office suite or automated investment research, this week’s Productivity Tapas has you covered.

🍔 Blog Bites - Built for Mars returns with 5 brilliant UX patterns you'll wish you'd thought of, including how Superhuman turns boring email setup into an engaging experience. Learn how DoorDash's Chief Growth Officer structures teams to move at lightning speed while staying customer-obsessed, and discover bol.com's systematic approach to killing ideas based on discovery insights before they waste development resources.

🎙️ Pod Shots

Five-time founder Ryan Carson reveals his 3-step AI coding workflow that's letting him build an entire company solo. Spoiler: it's not about "vibe coding" – it's about combining AI speed with the discipline of traditional product management. The solo founder advantage is real!

Plenty to get stuck into - off we go! 🚀

📰 Not boring

Product & Design

• Miro roll out prototyping in beta

• Slack launches ‘pinch to summarise’ inspired by the Arc browser feature

• John Rush tried 46 AI coding agents; here are his demos and notes

Apple

• iOS 19: All the rumoured changes Apple could be bringing to its new operating system

  • New name [this was in last week’s newsletter; now aligning with years]

  • Design overhaul

  • Dedicated gaming app

  • Smarter battery management

  • AI translation for messages

  • Virtual health coach - interesting is/how this further eats into other health apps

AI & Productivity

• ChatGPT kills meeting note takers…? Interesting to see if it can do better than Granola, but hopefully will reduce the 365,975 bots that seem to join calls

• Also interesting is this redacted OpenAI product strategy document from late last year that shows they want ChatGPT to become ‘the internet’s front door’

• A16z shows how SEO is losing its dominance to GEO (Generative Engine Optimisation) and how brands need to position to get cited by the likes of ChatGPT, Perplexity, Cluade etc

• Google released OpenEdge Gallery - an experimental app allowing you to download a host of open models to your phone to run locally

• Google’s NotebookLM now lets you share your notebook and AI podcasts — publicly

• And Alphabet CEO Sundar Pichai dismisses AI job fears, emphasizes expansion plans

• Mary Meeker used to do an annual ‘state of the internet’ every year but stopped 5-6 years ago. She’s back with a whopping 340 pages on the state of AI

Business Moves

• Tesla Targets June 12 Launch of Robotaxi Service in Austin

• Elon rolls out XChat to take on WhatsApp and Telegram; with encrypted audio/video calls, file-sharing etc.

• Salesforce buys Moonhub, a startup building AI tools for hiring

• After previously ‘opting for bots over humans’ Klarna CEO now says company will use humans to offer VIP customer service

Startup Insights

• Pattern matching 20 habits of Exceptional Startups. TL;DR:

  • Beat Expectations: Consistently exceed ambitious goals.

  • Top Talent & Fast Shipping: Obsessively recruit great people and ship high-quality products incredibly fast

  • Intense Culture: Avoid bureaucracy and inefficient spending, fostering a cult-like, intense environment with engaged leadership

  • Crisis to Momentum: Turn every crisis into an opportunity for growth and improvement

Get Your Free ChatGPT Productivity Bundle

Mindstream brings you 5 essential resources to master ChatGPT at work. This free bundle includes decision flowcharts, prompt templates, and our 2025 guide to AI productivity.

Our team of AI experts has packaged the most actionable ChatGPT hacks that are actually working for top marketers and founders. Save hours each week with these proven workflows.

It's completely free when you subscribe to our daily AI newsletter.

Subscribe to get instant access

⌚️ Productivity Tapas: Time-Saving Tools & GPTs

• Claude for Sheets: Use Claude in your Google Sheets ;)

• Magic Path: AI powered UI component builder and infinite canvas

• Context: AI powered office suite; documents, spreadsheets, presentations etc.

• EasyFin: AI-powered Investment Research Automation

Remember, as a Product Tapas Pro subscriber you can access the full time saving tools database for fast approaching 400 time-saving tools relevant for product managers and founders 🔥.

Check the link here to access.

🍔 Blog Bites - Essential Reads for Product Teams

Design: 5 Clever Design Patterns from Built for Mars

It's been ages since we've had a built for Mars feature in the newsletter, but they're always great. Below are some of Peter Ramsey's latest UX bites, you can check out his full set here:

1. Progress Visualisation (Superhuman): During setup, Superhuman shows a real-time count of emails being organised, leveraging both loss aversion and curiosity to keep users engaged during what could otherwise be a boring wait.

2. Contextual Awareness (Monzo): Monzo detects when you're using the app while on a phone call and dynamically indicates whether you're speaking with their customer service team, creating a seamless cross-channel experience.

3. Information Density (Strava): Strava's activity calendar uses variable-sized dots to represent different workout durations, allowing users to quickly visualise their exercise patterns without requiring additional space or interaction.

4. Adaptive CTAs (Amazon Prime): Amazon dynamically changes call-to-action labels for content you've previously watched, replacing "Watch" with more relevant options like "Resume" or "Watch Again," reducing friction in the user journey.

5. Priority Content (Airbnb): Airbnb prominently displays active bookings on their homepage with an animated calendar countdown, ensuring users can easily access their most important information without searching.

Leadership: How Brian Hale's Growth Mindset Fuelled DoorDash's Success

Bill Kerr recently interviewed Brian Hale, Chief Growth Officer at DoorDash, who sharesd insights from his journey scaling both Facebook and DoorDash through customer-centric growth strategies and cross-functional team structures. Read the full article here.

💡 "Making things feel easy is actually very hard. A lot of my time goes into working with teams to solve for that simplicity."

Product Strategy: How bol.com Kills Ideas Based on Discovery Insights

Tim Herbig continues his mini-series on bol.com's product practices, exploring how one of Europe's largest e-commerce companies adapts discovery processes to their unique context, resulting in both killed ideas and adapted solutions. Read the full article here.

💡 "Real Progress happens when you choose methods because they create value for you in your context, and you can use each domain to improve the others."

🎙️ Pod Shots - Bitesized Podcast Summaries

Remember, Product Tapas Pro subscribers get access to an ever growing database of all the top Podcast summaries we’ve ever created. 

Check it out here

🛠️ A 3-Step AI Coding Workflow That Actually Works: Lessons from a 5x Founder

Ryan Carson has built and sold three companies over 20 years, but his latest startup feels different. As a guest on Claire Vo's "How I AI" podcast, the five-time founder shared how he's building an entire company solo using a structured AI coding workflow that goes far beyond "vibe coding."

While most people treat AI like a magic wand—throwing vague requests at ChatGPT and hoping for the best—Ryan has developed a repeatable, three-step process that lets him ship complex features reliably. His approach combines the speed of AI with the discipline of traditional product management, proving that the future isn't about replacing human judgment, but augmenting it with better systems.

What it covers:

• The 3-file system that turns chaotic AI coding into a predictable, scalable workflow• AI-generated PRDs that actually work - no more vague requirements or endless back-and-forth

• One-task-at-a-time execution that prevents AI rabbit holes while maintaining human control

• The context paradox - why slowing down 5 minutes upfront saves hours of debugging later

• MCPs that supercharge your AI - connect directly to databases, browsers, and tools without switching apps

• The solo founder advantage - how to build entire companies with a fraction of the engineering resource.

 🎥Watch the full episode here

📆 Published: May 26th, 2025

🕒 Estimated Reading Time: 3 mins. Time saved: 30 mins🔥

🎯 The Context Problem: Why Most AI Coding Fails

The biggest mistake everyone makes? Rushing through context. We're impatient—we want to jump straight to the solution without giving the AI what it actually needs to solve our problem. Ryan's insight: slow down just a tiny bit on the setup, and everything speeds up dramatically.

Think of it like this: you wouldn't ask a junior developer to build a feature without explaining what you want, why you want it, and how it should work. Yet that's exactly what most people do with AI. They throw a vague request at ChatGPT or Cursor and wonder why the output is messy or incomplete.

Key Takeaways:

• Context is everything—the time you spend setting up the problem properly saves hours of debugging later

• Treat AI like a genius PhD student who can't connect simple, obvious things without guidance

• Most people fail because they're too impatient to properly explain what they want

📋 Step 1: Generate a Product Requirements Document (PRD)

Ryan starts every significant feature with a PRD—even for solo projects. He's created a custom Cursor rule that instructs the AI to write PRDs "suitable for a junior developer to understand and implement." This isn't just about documentation; it's about forcing clarity.

His PRD rule includes clarifying questions to ensure the feature is well-defined, structured sections covering functional requirements and design considerations, and a format that's easy for both humans and AI to parse. The magic happens when the AI asks clarifying questions—instead of answering every single one, Ryan often says "you pick" or "make your best judgment" for non-critical decisions.

Key Takeaways:

• Even solo builders need PRDs—they force you to think through edge cases before coding

• Create reusable Cursor rules that standardize how AI generates documentation

• Answer the important clarifying questions, but don't get bogged down in every detail

• Store PRDs in your project's tasks folder for easy reference and version control

✅ Step 2: Break It Down Into Tasks

Once the PRD is complete, Ryan uses another Cursor rule to generate a detailed task list. This isn't just a to-do list—it's a step-by-step implementation plan with subtasks, dependencies, and clear completion criteria.

The task generation rule creates numbered tasks with checkboxes, subtasks that break complex work into manageable chunks, and clear dependencies and sequencing. Ryan stores these in a simple markdown file in his project's tasks folder—no fancy project management tools, just a clear, version-controlled list that lives with the code.

Key Takeaways:

• Break big features into small, specific tasks to prevent AI from trying to do too much at once

• Use markdown with checkboxes—checking things off provides psychological momentum

• Keep task lists in your repo, not external tools, so they're version-controlled with your code

• Include subtasks and sub-subtasks for complex features to maintain granular control

🔄 Step 3: Execute One Task at a Time

Here's where Ryan's approach really shines. He has a third Cursor rule that manages task execution, with one critical constraint: complete one subtask at a time, then stop and wait for approval.

This human-in-the-loop approach prevents the AI from going down rabbit holes or making assumptions that compound into bigger problems. After each subtask, Ryan reviews the changes, commits to git if everything looks good, and gives the go-ahead for the next task.

Key Takeaways:

• Always work one subtask at a time—never let AI attempt multiple tasks simultaneously

• Build in mandatory stopping points for human review after each completed subtask

• Commit to git after completing parent tasks or when the app is in a stable state

• Small course corrections are easier than major rewrites—maintain control throughout

🔧 Advanced Tools: MCPs and Context Management

Beyond the core workflow, Ryan uses Model Context Protocol (MCP) servers to extend Cursor's capabilities. His most-used MCPs include Postgres MCP for querying databases directly, Browser Base MCP for controlling headless browsers, and Repo Prompt for precisely controlling context when working with large codebases.

These tools reduce the toil of switching between applications and give the AI access to real-time data about your system. Ryan also emphasizes the importance of context management, sometimes using tools like Repo Prompt when he needs precise control over what information the AI receives.

Key Takeaways:

• Invest in MCPs that eliminate context switching between tools and applications

• Use Postgres MCP to let AI query your database directly instead of writing SQL manually

• Browser Base MCP enables automated front-end testing and debugging workflows

• Repo Prompt gives you precise context control when Cursor's magic isn't enough

💡 Why This Matters for Product Builders

Ryan's workflow isn't just about coding—it's about sustainable product development. By following this structured approach, he's built "huge features with 10,000+ lines of code reliably" and can operate as a solo founder doing the work of entire teams.

For product managers, this workflow solves a common bottleneck: translating PRDs into actionable engineering tasks. For founders, this represents a fundamental shift in what's possible. As Ryan puts it: "I literally feel like I'm able to do all of it. Am I able to do it as well as a dedicated product manager? No. Am I able to think as deeply as a CTO? No. But I am able for sure to build this company."

Key Takeaways:

• AI doesn't just make coding faster—it makes entire categories of work accessible to non-specialists

• Even if you're not coding, having AI break down PRDs into specific tasks improves engineering handoffs

• Solo founders can now realistically handle product, engineering, and technical architecture

• The workflow scales from small features to complex, multi-thousand-line implementations

🚀 Getting Started: Start Small, Stay Consistent

Ryan's advice for getting started is simple: pick one model and get to know it well. He's currently defaulting to Gemini 2.5 Pro, but the specific model matters less than consistency. Learn what your chosen AI is good at, what it struggles with, and how to prompt it effectively.

Start with the PRD and task generation, even if you're not ready to automate the execution. Most importantly: get your hands dirty. "Nobody really knows how to do this stuff," Ryan says. "The only way you're really going to figure it out is by getting in here and getting your hands dirty and see what works."

Key Takeaways:

• Choose one AI model and learn its strengths and weaknesses deeply

• Start with PRD and task generation before attempting full workflow automation

• Begin with small projects to build confidence and refine your process

• Expect to iterate—nobody has this figured out yet, so experimentation is key

Ryan's approach proves that the future of product development isn't about replacing human judgment—it's about augmenting it with structured, repeatable processes that let you move faster while maintaining quality. Whether you're a solo founder or part of a larger team, these principles can help you build better products with less friction.

The vibe coding era is fun, but the structured AI workflow era is where the real productivity gains live.

🎥Watch the full episode here

📅Timestamps:

• (00:00) Introduction and Ryan’s recent AI projects

• (03:25) Demo: Creating a PRD with Cursor

• (05:00) Ryan’s open source links: https://github.com/snarktank/ai-dev-t... 

• (09:53) Quick recap and common mistakes to avoid

• (11:00) Demo: Generating a task list from the PRD

• (15:31) The importance of context when working with LLMs

• 18:07) Demo: Working through tasks systematically using Cursor

• (18:56) Change management

• (20:00) How task lists save time for product managers

• (21:50) Demo: Using MCPs for front-end testing

• (24:50) Specific MCPs and what to use them for

• (26:45) Demo: Using Repo Prompt to gain precise control over context

• (31:23) Music’s role in Ryan’s development stack

• (32:10) Lightning round and final thoughts

That’s a wrap.

As always, the journey doesn't end here!

Please share and let us know what you would like to see more or less of so we can continue to improve your Product Tapas. 🚀👋

Alastair 🍽️.