Build On Demand with AI

Creating an AI-Powered Voice Memo Plugin for Obsidian

For as long as I can remember, I’ve been a builder. Back in 2007, I started a company focused on Web 2.0—yes, the era of shiny new AJAX apps and social media breakthroughs. Creating software for me probably feels like painting a picture or fixing a classic car to others: there is a sense of fulfillment in solving a specific problem and seeing a finished product take shape before my eyes.

From Inspiration to Action

Lately, I’ve been experiencing a huge shift in how I build. With AI, it’s as though I can do in hours what used to take days, weeks, or even months. Tools like Claude (within Windsurf or Cline) have made it possible to “speak” a solution into existence, one step at a time. I’ll describe what I need, and these AI tools help me flesh out the design, generate documentation, set up repositories on GitHub, and even guide me through the submission processes—whether that’s the iOS App Store, Chrome Web Store, or in this case, a pull request to the Obsidian community plugin repo.

I like to think of myself as the conductor of an orchestra. The instruments—coding languages, frameworks, architectures—are all within my reach. With the help of AI, I can call upon any one of them with just my words. It’s exhilarating.

Why Obsidian?

I’ve built my share of Next.js apps, iOS apps, and Chrome extensions. But Obsidian captured my attention because it’s where all my ideas, notes, and tasks live. It’s a knowledge worker’s dream: fast, markdown-based, offline-friendly, and infinitely extensible. But it’s the extensibility that truly intrigued me.

I had a personal need: quickly record ideas or reminders via voice and convert them into structured, actionable notes. With everything else I juggle, it was time to build an Obsidian AI Voice Memo Plugin—and do it in a way that was as frictionless as possible.

GitHub Link: https://github.com/BioInfo/Obsidian-AIMemo

My Workflow: A Lean, AI-Powered Process

I keep things simple. First, I draft a Product Requirements Document (PRD) with OpenAI, outlining the concept and the “why” behind it. Then, I generate a few core documents—like roadmaps, session logs, and design considerations. After that, I feed everything into a coding agent (like Cline), where I iterate rapidly. The AI helps me follow the roadmap, frequently update my docs, and manage my GitHub repo.

Before I know it, boom—I have a finished product (or at least a working beta). This entire process often costs me just a handful of dollars in AI tokens, and the turnaround time is ridiculously fast. It’s still mind-blowing to me that I can build a custom Obsidian plugin in a matter of hours.

The Plugin: AI Voice Memo for Obsidian

Here’s a quick look at what it does:

• Voice Recording: Assign a keyboard shortcut to record audio on the fly (both on desktop and mobile).

• Transcription: Use an AI model (like Whisper) to transcribe your recordings in near real-time.

• Summaries & Tasks: Generate a concise summary of each memo. The plugin also flags action items automatically, placing them into a task list.

• Organization: Store everything in a structured way—organized by date in the AI-Memo folder within your vault.

• Customization: Let users tweak transcription settings, summary length, and action item extraction.

What started as a need to capture voice notes quickly is rapidly evolving into a more comprehensive note management tool. Each user’s needs might differ—but that’s where the plugin’s configurability comes in.

Challenges and Triumphs

Ironically, the trickiest part in building apps like these (iOS, Chrome, or Obsidian) often lies in the final acceptance or publishing step. But it’s not really a blocker when your primary goal is to solve your own problems. If I can run the plugin locally, then it’s already serving my needs—even if the official publishing process takes a while.

That’s the beauty of building on demand. The friction of “getting published” is a formality; what really matters is harnessing AI to solve my acute problem right now.

Build On Demand (BoD): A New Era of Software

We’re witnessing a revolution in how software gets made. I call it Build on Demand (BoD)—a future where you can create exactly the tool you need at precisely the time you need it. Soon, software might feel disposable: you build a solution, use it, and either enhance it or move on.

Sure, some tools will remain essential for the long haul. But the ability to spin up a custom plugin or app for a specific use case—and do it in a matter of hours—feels like a game-changer.

Next Steps

I plan to keep refining this plugin, likely with help from the community (I’ve released it under an MIT license). Future improvements might include more robust error handling, better offline support, or advanced transcription refinements.

And after this? Probably another build-on-demand project. Because that’s the best part of our new AI-infused world: I can dream up a new tool or solution, and then just build it—often in a single day.

Take it Home…

If there’s one takeaway from my journey, it’s that building exactly what you need has never been easier. With AI at our fingertips, we’re only limited by our imaginations. Whether it’s a voice memo plugin for Obsidian or a custom Chrome extension, the time from idea to working product has shrunk dramatically.

Go ahead—envision the tool you wish you had. Chances are, in the span of an afternoon and a handful of AI prompts, you can make it real. The future of software is here, and it’s yours to build on demand.