How AI is Rewriting Our Development Cycles
What Caught My Eye This Week - #11
Each week, I'll bring you the most relevant and insightful tech stories, saving you time and keeping you informed.
For years, our development process has followed a familiar path, no matter how big or small the project: The Product Requirements Document (PRD), flowing into detailed Figma mockups, then broken down into engineering designs and tasks.
It's a structured and linear process.
But what if that's changing because of AI?
Thanks to concepts like "vibe coding," the future process might shift towards: "vibe coding prototyping to replace the PRD/Figma part."
AI may not just be about smarter features like agents; it also might mean we have smarter ways to build.
In the past 5-10 years, one reason we needed a PRD was that even creating a lightweight prototype was a black hole of time consumption. Now, you can go straight to a lightweight prototype with a similar amount of information and in a similar amount of time as writing a PRD.
Is this a game-changer? Why?
Clearer Discussions: A working prototype, even a basic one, resonates far more powerfully with stakeholders than a dense document. It shifts the conversation from debating how something might work or look to focusing on how to measure whether this feature works or not. Especially in big corporations, most projects are killed before execution. A working prototype can help bypass parts of the traditional review process.
Rapid Feedback Loop = More Possibilities: You can explore multiple options in the same amount of time.
Deeper Understanding: Interacting with a prototype can lead to better PRDs, if they are still deemed necessary. This improves the understanding of additional requirements and design considerations you might not uncover until you're further down the product development cycle.
The Evolving Workflow: What Does It Look Like?
Previously: Rough Idea -> Formal PRD -> Figma Design -> Engineering Design / Tasks -> Development -> Experiment -> Launch
A New and More Iterative Flow: Rough Idea -> Lightweight Prototype (possibly several rounds of prototyping) -> (Formal PRD, may not be needed) -> Final Design in Figma -> Engineering Design / Tasks -> Development -> Final Experiment -> Launch
Still Something to Be Aware Of
The prototype may not consist of typical production-ready code.
The goal for aesthetics is often to "get close enough for prototyping purposes" or to "remove cognitive dissonance so that stakeholders don’t get distracted with feedback like ‘It’s the wrong color!’"
The main emphasis is on speed and idea validation.
Key Takeaways for the AI Age of Product Development:
Prototype as early as you can.
Shift focus to objectives: Use prototypes to facilitate discussions around goals and metrics, not just features.
Embrace iteration: Explore multiple options rapidly before deep investment.
PRDs evolve, not disappear: PRDs may become more concise thanks to the initial prototype.
Human insights remain crucial: Product Managers, designers, and engineers are still essential for production-ready execution.
What I read this week…
Google Gemini has the worst LLM API (venki.dev)
Design for 3D-Printing (blog.rahix.de)
New Products…
Stirling-Tools / Stirling-PDF
#1 Locally hosted web application that allows you to perform various operations on PDF files
https://github.com/Stirling-Tools/Stirling-PDF
aipotheosis-labs / aci
ACI.dev is the open source platform that connects your AI agents to 600+ tool integrations with multi-tenant auth, granular permissions, and access through direct function calling or a unified MCP server.
https://github.com/aipotheosis-labs/aci
Dash
Dash is your AI agent for modern work. It connects to your apps, remembers your context, and gets things done — so you stop wasting time switching, searching, and starting over. Full control, one conversation.
UI Builder + Mockup tool
UI Builder is a Chrome extension for quick, sketch-style wireframes directly in your browser. Drag buttons, inputs, headers onto any page or blank canvas. Perfect for fast feedback, live demos, and client calls.