A guide to 'AI coding without going broke' gets schooled by people who go broke far slower

The guide, and its headline number

The post lays out three ways to do AI-assisted coding affordably: self-host open models on hardware you buy (no per-token cost, but steep upfront and weaker models), rent those same open models through a provider like OpenRouter (the recommendation for most people), or bundle frontier subscription plans for their generous included tokens. Its economic hook is a specific arbitrage — that roughly $400 a month of plans buys about $2,800 of API usage at list prices, "a real bargain right up until you hit the ceiling" — alongside the claim that a solo developer can ship what a team of twenty would in a month for around a thousand dollars.

The thread's verdict on the writing was unkind and close to unanimous, captured in one line:

"How is this even an article? The advice is just 'pay for max'." — a reader on the thinness of the recommendation

The people who actually do this spend cents

The post's premise is that home AI coding is a budgeting problem. Several commenters who run exactly these setups reported costs that make the framing look quaint. One challenged the math while agreeing subsidies are the real story:

"it never stops being a 'real bargain' at any point. And it's more like $200/mo for $4000+/mo in tokens… There's no sense in running local models… as long as VCs… are willing to pay your bill." — a startup co-founder disputing the post's own numbers

Another reframed what "at home" even meant, from a machine with no GPU at all:

"Oh, so this is not a post about AI coding at home. It's about vibe coding at home… I'm posting this from a home computer with 64 GB of RAM and no GPU. I do lots of AI coding while spending very little money. I run Gemma… and Qwen 3 coder with Ollama." — an economist and professor running local models

And the low-water mark for spend came with receipts:

"This month I've spent only 15 cents using DeepSeek API and my own coding agent. Three apps delivered to clients…" — a developer billing his work to a DeepSeek agent

So where does the money actually go?

Put the voices together and the real 2026 picture inverts the post. Self-hosting rarely pays off purely on cost, because frontier subscriptions are so heavily VC-subsidized that $20–$200 a month buys token allowances no local rig can match on price. "Going broke" isn't a hardware decision; it's what happens when an agent is left to grind unmanaged through a long task at list-price API rates. The honest guide isn't "buy a GPU" or "pay for max." It's: rent subsidized tokens while they're cheap, keep your sessions short, and don't mistake a thin blog post's arbitrage for a strategy.