$ cat manifesto.txt

Everyone talks about Claude. I show the receipts.

I run a marketing agency and build software with Claude every day. This site documents the actual implementations: the prompts, the process, the failures, and what shipped.

jeff@humbear — claude code

$ claude "audit humbearmedia.com and write up what's broken"

✓ crawled site · pulled Semrush data · wrote audit report

$ claude "build the career planning app we scoped yesterday"

✓ shipped to production — careerthatlasts.com

$ claude "turn this morning's call into ClickUp tasks"

✓ transcript pulled · 6 tasks created · summary in Slack

$ claude "build godofclaude.com while I'm out"

▮ you are looking at the result

$ why this site exists

The internet is full of Claude content. Lists of use cases nobody has tried. Prompt packs from people who do not run a business. Tool roundups written to farm affiliate clicks. Almost none of it shows actual implementation.

This site is the opposite. I run a marketing and automation agency, and Claude is wired into how it operates. Everything published here comes from real work on real projects, with the exact process attached.

That includes the failures. Half the value of a build log is the part where the first approach did not work. If a workflow fell apart in practice, I say so, because that is the information I wish other people published.

You will find three things here. Build Logs of real projects. Playbooks with the exact processes I use. The Stack of tools actually in rotation. Nothing theoretical, nothing I have not run myself.

There is no course, no community upsell, no gated PDF. I am not selling you a system. I am documenting mine, in public, because this is the resource I could not find when I needed it.

$ ls ./builds --sort=proof

Build logs

Real projects, documented honestly: what I asked Claude to do, what it produced, what shipped, and what didn't work.

careerthatlasts.com● live

Career That Lasts: shipping a live AI career app with Claude Code

I shipped a live AI career-planning app by pairing with Claude Code. Here is the actual commit log.

Claude CodeClaude API (@anthropic-ai/sdk)Next.js 16 (App Router)React 19

read the build log →

humbearmedia.com● live

Auditing my own agency's website with Claude

Claude audited my agency's site and found our three money pages were hidden from Google. Fixed same day.

Claude CodeConnected Chrome session (logged-in wp-admin)WordPress with ElementorYoast SEO

read the build log →

internal▪ internal tool

Humbear Ops and AI Console

The internal ops dashboard I built with Claude Code instead of paying for another analytics SaaS.

Claude CodeNext.js 14 (App Router)React 18TypeScript

read the build log →

internal▪ internal tool

iMessage Triage

An always-on agent that answers my texts in my own voice — but only while a Focus mode is on.

Claude (Sonnet 4.6) — the live decision and voice engineClaude Code — built the dispatcherPython 3launchd (run at load + keep alive)

read the build log →

easyvob.com● live

EasyVOB: Claude as the compliance watchdog, not the verification engine

A live HIPAA-focused VOB service. Claude's verified role: compliance watchdog, not the verification loop.

WordPressElementor (with Pro Elements)Lexend theme (child theme)Hostinger (LiteSpeed, PHP 8.2)

read the build log →

highintentlead.com● live

HighIntentLead: the one-file lead-gen site

One static HTML file selling identity-resolved lead gen. Live, fast, and honest about the missing receipt.

Claude CodeHand-coded static HTML and CSS, single file, no JavaScriptGoogle Fonts (Inter)Cloudflare

read the build log →

$ cat ./playbooks/*.md

Playbooks

The exact processes, written so you can run them yourself. No gatekeeping, no course upsell.

01

My exact Claude Code setup

Claude Code is only as useful as what it can reach. This is the exact setup I run on Windows 11: one install, two CLIs, eleven connectors, and persistent memory, in the order that gets you to a working system fastest. Every connector listed here is one I actually have wired in, with what each one does for me.

An afternoon for install, auth, and connectors. Memory and the first real project take the week that follows.9 steps
02

Shipping an MVP with Claude Code

This is the exact process I used to build Career That Lasts, an AI career-planning app that is live at careerthatlasts.com. Git shows 18 commits: 17 in one overnight session and one follow-up two nights later. The short version: cut scope to demo mode, scaffold with the official generator, run a tight loop with Claude Code, deploy to Vercel immediately, and keep monetization behind a feature flag so shipping never waits on a business model.

One long overnight session for the core (first commit 2:09am, last commit that morning 7:54am), plus a follow-up evening. Call it a weekend with sleep.8 steps
03

How I audit a website with Claude (the exact process)

This is the exact process I used to audit a live WordPress lead-gen site with Claude: inventory from the sitemap, crawl every URL, build a per-page SEO table, test tracking two ways, submit a real test lead, probe security from the outside, then fix the safe stuff the same day with verification. No tool did this for me. Claude did the fetching, parsing, and verifying; I made the judgment calls.

A full day for the audit itself. The fixes that need owner decisions stretch out over the following weeks.10 steps
04

Wiring Claude into your actual business tools (MCP)

Most MCP content stops at 'look, Claude can connect to things.' This is what I actually do with it: six workflow patterns from my real setup, where Claude reads and proposes across ClickUp, Gmail, Slack, Notion, Calendar, Fireflies, and Semrush, and I stay the one who hits send. Every pattern includes the prompt shape, what the connector does under the hood, and exactly where the human checkpoint sits.

An afternoon to connect and test the basics. Give it a week of real use before you trust Claude with write operations.8 steps
05

Putting Claude on a schedule

Recurring agents that do real work while you are away from the keyboard. This playbook covers how to set up scheduled tasks in Claude Code for three patterns I use and recommend, a morning digest, a weekly SEO check, and a content-draft generator, plus the rule I use for when Make.com is the better tool. No theoretical use-case lists. Just the setup, the guardrails, and the failure modes.

An afternoon for your first scheduled agent, then under an hour for each one after that9 steps

$ follow --the-build

Watch it happen, don't take my word for it

Every build on this site gets documented as it happens — the prompts, the dead ends, the results. No course at the end of this funnel. There is no funnel.

follow on x →