Building AI‑First
Products.
I ship the tools I wish existed — for content, for developers, for the work of shipping itself.
BlackOps Center, VitalWall, and VoiceCommit are live experiments in what AI‑native software looks like in production — built and operated with the same patterns I apply to everything.
Front-end architect · 30+ years building at scale · shipping AI-first products
psst — try ⌘ A for Asteroids, ⌘ ⇧ L for Lunar Lander, or ⌘ ⇧ X for Missile Command
Get the Inside Track
Weekly insights from 30 years of building software — now focused on AI workflows and the systems we're building at BlackOps Center.
AI Development Tips
Real workflows that save hours
30 Years of Dev
Lessons from a long career shipping software
Founder Notes
Behind the scenes building BlackOps Center
No spam, ever. Unsubscribe with one click.
Applied AI in Production
These products are how I test and refine AI-driven development in real systems. Each one is built and operated with the same AI-driven patterns I apply to everything I build.
BlackOps Center
Agent-native publishing, driven over MCP
The agent-native content operating system. Exposes its whole surface over MCP, so the agent already holding your context can research, draft, publish, and read the analytics back.
VitalWall
Used by 500+ websites
Real-time website analytics and visitor tracking platform that boosts engagement through intelligent social proof.
VoiceCommit
Zero to GitHub in 60 seconds
Voice-first developer tool that turns spoken ideas into GitHub Issues, PRs, and updates. An experiment in removing friction between thinking and shipping.
Latest Articles
My Claude Code mod lets Jev choose the agent and model for every call
Jev picks the model for every subagent call in about a third of a second. On lookups it cut the cost by 51% and every fact still checked out.
Every Claude Code subagent inherits your model, so a file count runs on Opus. I built a mod that asks Jev to pick the cheapest model that can do each job. Lookups got 51% cheaper.
I never remember to send my newsletter, so BlackOps sends it now
If I miss my own deadline, BlackOps writes the issue, tells you it's covering for me, and sends it.
My last newsletter went out August 19. So I built Newsletter Autopilot: if I miss my deadline, BlackOps writes the issue from what I published and tells you it's covering for me.
I asked Claude to read 30 of my Claude Code sessions and build the mods I needed
The exact prompt, the five mods Claude built from my own sessions, and what broke when they ran for real.
I asked Claude to read my last 30 Claude Code sessions and suggest mods for how I work. It came back with ten ideas and built all ten as five mods. Here's the prompt, what each mod does, and the bugs a real run caught.
Jev now makes the decisions in BlackOps that I used to hand to an LLM
A decision model that answers with numbers instead of prose. It scores my reply targets, checks voice on every publish, filters my feeds and reads replies to brain tasks.
Jev returns typed decisions, not prose. In BlackOps it scores reply targets, checks brand voice on publish, filters feeds and reads task replies.
Every Version of My Obsidian Daily Note Since 2021
A template, a patched calendar tool, two iOS Shortcuts, a script, an overnight filer. Every rebuild took one chore off me. The newest one I just talk to.
I've rebuilt my Obsidian daily note over and over since 2021 and wrote up most of the rebuilds as they happened. This is the whole line, what each version fixed, what it left for me, and where it landed.
I Gave Claude Code the Voice of Rocky From Project Hail Mary
When a deploy is actually live or a stubborn bug finally dies, Claude Code says one short line out loud in Rocky's voice. Everything on screen stays normal English.
I started with Rocky reading every reply aloud. It was too much. Now Claude Code cheers once, in Rocky's voice, when something big lands.