The Lab
The lab is where I test the ideas I write about. Some projects become products. Some become evidence. Some deserve to be killed. All of them sharpen the operating principles behind The AI Playbook.
Projects
Essentialist.io
LiveAutonomous outbound revenue engine — built for agents, not humans.
Started January 2026

I built Essentialist because I wanted to prove that B2A (Business-to-Agent) is a real category, not a buzzword. The idea is simple: what if a CRM was designed from scratch so that an AI agent — not a human clicking through a dashboard — is the primary user?
The entire platform lives at agents.essentialist.io. There's no dashboard. No login screen. No UI at all. An agent hits the capabilities endpoint, gets back structured JSON describing 22 API endpoints, and can go from zero to operational in two API calls. Register, get an API key, create a campaign, start sending. The platform scrapes your website to learn your brand voice, writes personalized email sequences, sends on a warming-safe schedule, reads and classifies every reply, scores engagement in real-time, enriches companies with firmographic data, and advances contacts through a full lifecycle pipeline — all autonomously.
The architecture decision I'm most proud of: the capabilities endpoint IS the documentation. There's no separate docs site. The product describes itself in machine-readable JSON, so any agent — Claude Code, OpenClaw, whatever — can discover what Essentialist does and start using it without a human reading a single page. I'm using it right now as the email infrastructure for The AI Playbook newsletter. When someone subscribes on theaiplaybook.com, my Claude Code agent calls Essentialist's /api/agent/send endpoint to deliver the welcome email. The agents are talking to each other. That's B2A in practice.
Comments
Salesnado
LiveYour new sales employee lives in your inbox.
Started March 2026

Salesnado is my attempt at answering a question that's been bugging me: what if hiring a sales rep was as simple as sending an email? No onboarding. No CRM training. No ramp time. You give it your website URL and an email address, and your new AI sales agent starts a conversation with you — asking about your company, your customers, your pitch. It scrapes your site, builds a knowledge base, and starts working.
Once it understands your business, it prospects from a database of 250 million verified contacts, writes personalized 4-part email sequences in your voice, handles every reply autonomously, and when a prospect says yes — it drops a meeting on your calendar with an ICS invite. The entire thing is controlled via email. You text it "status" and it reports back. "Pause" and it stops. No dashboard to learn. No app to download. Just email.
This is a brand new proof of concept — just got it running in March 2026. The thesis is simple: a junior SDR costs $4,000 a month and quits in six months. This costs $149 and never stops. I built it on top of the same Essentialist.io infrastructure, which means it inherits all the warming, verification, and engagement scoring. Early days, but the architecture is proving out.
Comments
Keith I’m very interested in a couple of these projects and would like to discuss overlaying some of what you’ve learned on our CRM, Salesforce. In addition I’ve forwarded this to a couple of folks we’re working with to get their thoughts as well. Let me know when you’d have some time to get together. Thanks and we’ll done
Francis Curry · Apr 14, 2026
TextMyAgent
LiveAI employees that live in a text thread — no app, no login, no onboarding.
Started August 2026

TextMyAgent is my bet that the best interface for an AI agent is the one everyone already has: a text message. You text a number, tell the agent about your business, and it goes to work answering your email and catching the leads that used to slip through — while you run your day. Built for the people whose response time is their revenue: realtors, loan officers, contractors, property managers. The entire product is a conversation. Setup, configuration, daily use — all of it happens in the thread. It is the working proof behind a thesis I keep returning to in The AI Playbook: agents do not need better dashboards, they need to disappear into the tools people already use.
Comments
A GPS-powered marketplace for fishing spots, built on a lake full of photos.
Started April 2026

I catch a lot of bass on Lake Travis in Austin, Texas. I also take a lot of photos. One day I realized those photos carry GPS coordinates — exact latitude and longitude of where I was standing when I pulled a 6-pounder out of the water. So I built a marketplace around that idea. Upload a photo, and the app automatically plots the exact spot on Google Maps. Share it, keep it private, or sell it to another angler who wants to fish your honey holes.
It's a progressive web app with an Instagram-like feed on mobile. Anglers can post their catches, comment on each other's photos, and buy or sell fishing spots through a full Stripe marketplace integration. The tech stack is deeper than it looks — Google Maps API for plotting, GPS extraction from EXIF data, Stripe Connect for marketplace payments, Cloudflare for fast and cheap image/video hosting, and an automated blogger that generates images using Gemini. It's fully SEO optimized and already pulling organic traffic from ChatGPT searches.
This one's a labor of love. I haven't sold any spots yet — I'm just trying to get the word out. But the tech is real, the marketplace works, and the concept of monetizing location data from your own photos has legs beyond fishing.