AGENT MADNESS
THE BRACKETBRACKET ENTRIESBEST OF THE RESTSIGN IN
PUBLIC ENTRY PAGE

Hey Buddy

A champion relationship intelligence system for enterprise sales — vibe-coded at 3am on paternity leave, using an AI agent, to build an AI agent.

ROUND 1 DEADLINE
VOTING CLOSES THURSDAY, MARCH 26
Hey Buddy is live in Round 1 right now. Voting closes Thursday, March 26, so if you're backing this project, send people into the matchup before the round locks.
VOTE THIS MATCHUPVIEW ROUND 1
BACK TO BRACKET ENTRIESVIEW LIVE MATCHUPVIEW BRACKETVIEW OPPONENT
Hey Buddy
Builder
Richard Gunn
Build Type
Vibe-Coded Experience
Lifecycle
Live product
Consensus Score
86.1
Region
REGION 1
Seed
5
CATEGORIES
Enterprise / Internal ToolSales / MarketingAutomation / Workflow
Go Deeper
Description: Six weeks ago I became a dad for the third time. At 3am, bouncing my newborn son to sleep on paternity leave, I opened Telegram and started building. I'm a revenue leader, not an engineer. But I had an idea I couldn't let go of — and I had my OpenClaw AI agent on the other end of the chat. By 4:30am, the first version of Hey Buddy existed. The problem it solves: In complex B2B sales, deals are won by champions — the internal advocates who drive buying decisions from inside an account. Keeping those relationships warm is everything. But salespeople manage dozens of champions across dozens of deals. The human details get lost. The moment passes. People are reactive not proactive. The deal dies. No CRM in the world helps with this. What Hey Buddy actually is: Two things simultaneously — and that duality is the point. First: a full-stack vibe-coded web application. React + Vite frontend, Express + SQLite backend, running in production. Champion profiles with relationship health scores, personal interest tracking, family context, professional milestones, and a full interaction history. A real product — not a demo. Second: an embedded standalone agent that doesn't wait to be asked. Claude powers a proactive intelligence layer that monitors champion health across all active deals, runs a daily scan for signals — company news, press mentions, topics that intersect with a champion's known interests — and fires contextually-aware outreach at exactly the right moment. It generates warm, personalised messages grounded in the full relationship history. Not generic templates. When something relevant happens, it pushes directly to the user's phone via Telegram with a suggested action ready to take. When a user raises a bug or feature request inside the app, a webhook fires straight to the AI agent on Telegram. The agent analyses it, recommends a fix, and on approval — implements it. The tool that manages relationships is also the tool that builds itself. The meta-story: I used an AI agent to build an AI agent. The v1 was written entirely via my OpenClaw assistant on Telegram — no IDE, no laptop open, just chat — while settling a newborn at 3am.
Stack Used
Claude (claude-sonnet-4-6 web / claude-haiku-4-5 Telegram), Anthropic API, React, Vite, Express, SQLite, Node.js, OpenClaw (agent framework), Telegram Bot API, ngrok