Think "Spec Kit" finely tuned for startup apps... video below!

What is Spec-Driven Dev?: Spec-driven AI development is a method where you write a detailed specification (PRD/Requirements/etc - the "what" and "why" of your software) upfront, and then AI coding tools use that spec as the source of truth to generate, test, and validate the actual code in more manageable pieces —so you're steering the AI with clearer intent. Note - @Doug Montgomery was an advocate for getting me to test these techniques, so special shout out required! What is AI-First Dev?: This is just terminology I like to use in my own little world. It's what I'd call "Vibe Coding" with much more intention, management and realistic expectations/understanding of the capabilities of current AI. Why?: Remember the hype of "context engineering"? I like to think of Context with AI being how much information the model can manage before it degrades in quality. And, as it happens, this is the biggest limiting factor when doing full (or nearly full) AI-coding. Spec-Driven Dev is just one of many techniques to manage context, or "context engineering." How does it help with context? Picture you have a destination and the choice between a maze of roads/highways/gas-stations and car to get there (Vibe Coding) or a train on tracks with stations (Spec Driven Dev). Through various techniques based on Software-Engineering principles, in Spec-Driven Dev we are trying to keep the LLM on tracks and limit the length it has to travel before resetting its context - think reducing distractions and increasing focus. Problems?: I actually have found the platforms I've tested (such as GitHub's Spec Kit) to be over-engineered (too advanced and bulky) for the scale of apps I've had to build for clients and myself. Imagine our previous destination example - there was a better chance I'd get to the destination with the car because the train was going so slow. Hope?: A simpler Spec-Driven Dev workflow has been serving me amazingly for the past couple of months of testing. Further, I've found it to be effective, not only in more my comfort zone (backend), it has yielded very impressive results in frontend designs as well! They have some pre-built lighter weight solutions out there I believe.

Made a video walking through the AI-assisted development workflow I use for production voice AI projects. Quick context: I used this on a Vapi replacement project at a $4 billion unicorn. Within a couple of weeks, their engineering team had adopted the workflow without being asked — and they wanted me to present it division-wide. The video covers: - How I coordinate specialised AI agents for Pipecat, Vapi, and Twilio research - Building features that would normally take hours of documentation reading - Live implementation of JSON-based assistant configuration (swap providers via config file, no code changes) https://youtu.be/AG68VC_mOGY Happy to answer questions.

I don’t want to create a long list of rules. We all know how to be respectful with one another. The only guideline I want to make very clear is the focus of this community. I started this space because there wasn’t a dedicated place for open-source Voice AI, and that’s what our posts, discussions, calls, and shared resources should revolve around. There are already plenty of communities centered on proprietary Voice AI platforms like Vapi, Retell, Bland, Synthflow, etc. To make sure everyone gets the most value here, I kindly ask that you keep your contributions focused on open-source Voice AI and topics closely related to it. Additionally, please avoid posting general AI content that isn’t related to Voice AI — for example, tutorials on how to automate your social media channels. There are other spaces better suited for those conversations. Thank you everybody for your contributions so far!

I’m excited to share that another member of the LiveKit team has joined the party! Please join me in welcoming @Yepher Yepher from LiveKit’s Developer Relations team!