What one person can actually ship with AI in 2025 (live session + meta twist)

Hey everyone 👋

Just dropped something I've been working on and thought this community would appreciate it.

QUICK CONTEXT: I'm a solo founder with 20 years in software and 4 years deep in generative AI. Currently building multiple products without a team - just me and AI tools.

WHAT THIS IS: A 56-minute live, unscripted session showing exactly how I work. No polished demos. Just the real, messy process of shipping software solo.

WHAT YOU'LL SEE:

→ Building a voice AI desktop app (local models, text-to-speech, voice morphing)

→ Claude Code deep dive - skills, parallel agents, sub-agents

→ Creating a landing page from scratch in real-time

→ Research and competitive intelligence workflows

→ My actual development philosophy after years of doing this

THE META TWIST: I also edited this 97-minute raw recording with AI. Built a pipeline in ~10 minutes that:

→ Transcribed everything with word-level timestamps

→ Auto-detected content structure

→ Identified and removed sensitive content (I accidentally showed API keys 😅)

→ Created two versions with professional transitions

→ Zero manual frame editing

The 30-second time-lapse attached shows the entire AI editing process.

WHY I'M SHARING THIS: A lot of people ask "does AI-assisted development actually work in practice?" This is what it looks like - not theory, but the actual workflow.

FULL SESSION: https://www.youtube.com/watch?v=fAYkcwo45Ww

What am I working on in the video: https://vois.so

If you're building automation workflows for clients, this shows what's possible when you go deep on one tool (Claude Code) rather than shallow on many.

Happy to answer any questions about the workflow, tools, or approach.

0:30

0 comments