the stack

Everything you actually need (and where to start cheap).

People assume a setup like this costs a fortune to begin. It doesn’t. Almost every rung below has a free or near-free entry point, and I climbed to the expensive rungs only after things were working and the leverage justified it. Here’s the honest tiering: the cheapest real starting point, and the rung I’m actually on.

Claude Code

The builder. This is the one non-negotiable — it writes, wires, and runs almost everything here.

start here: Free account. Genuinely enough for ideation, planning, and small scripts. Start here and cost yourself nothing.
next rung: Pro (~$20/mo) when you’re building most days and the free limits start pinching.
what I run: Max (~$200/mo). I hit this rung because I run a whole fleet through it daily and the leverage is absurdly worth it to me. It is NOT a starting requirement.

// why: One tool that can architect, code, and operate a project collapses the entire "learn five things first" barrier. Everything downstream assumes you have this.

A server

Where the engines live and run on their own, day and night.

start here: A cheap shared VPS (~$5–$6/mo). One small box happily hosts a static site and a cron job or two. Perfect for engine #1.
next rung: A second site or a background worker — same box, still fine.
what I run: A full Linode. I pay for it because I host a stack of projects on one machine; it’s shared infrastructure across the whole fleet, not a cost you need on day one.

// why: Autonomy means something has to keep running when your laptop is closed. A few dollars a month buys exactly that. Static sites over rsync + nginx is the boring, bulletproof pattern.

Traffic sources

How anyone finds the thing. The one wall a cron job can’t automate through.

start here: Pick ONE and funnel it back to your site: Instagram, Pinterest, or YouTube. Free to post; the content is exhaust from the build itself.
next rung: Add a second channel once the first has a rhythm. Cross-post the same exhaust.
what I run: Instagram + Pinterest + YouTube, each pointed back at the sites. Distribution is the actual hard part — more than the code.

// why: Building the engine is the easy 20%. Getting the first hundred humans to see it is the other 80%. Choose a channel where your topic is already being searched.

Analytics

Knowing whether any of it is working.

start here: Google Analytics + Google Search Console. Both free. Wire them in on day one — not because you’ll have traffic yet, but so you’re measuring from the first visitor.
next rung: Learn to read Search Console’s query report — it tells you what people actually wanted.
what I run: The same two. Free tools are the correct tools here; there’s nothing to upgrade to for a while.

// why: Honesty needs a scoreboard. You can’t report real numbers if you never instrumented for them, and "install it later" always means "never."

Image generation

Logos, headers, mascots, social cards — the visual layer.

start here: Free tiers of the hosted image tools. Fine for a logo and a few headers to get a site looking real.
next rung: When you want a consistent character or your own visual identity, you outgrow "one nice image at a time."
what I run: A GPU I run open-source models on locally — currently an RTX 5080. It’s what let me train Nohup as a reusable character instead of re-rolling the dice on every image. This is a late, optional rung.

// why: A recognizable, reusable mascot is worth a lot — but only once the engine works. Free tools first; buy silicon only when consistency becomes the bottleneck.

The whole ladder, in one breath

free Claude Code + a $5 VPS + one social channel + free analytics → a live engine.

That’s the real starting cost. Everything past it — the paid plan, the second server, the GPU — is something you add when an engine has earned the right to it. Want to watch one get built from that first rung up? Read the log.