A free Overyield tool. Everything you need to sell one fixed-price landing page at the market anchor, produce it for about a dollar of model spend, QC it so the quality holds, and publish the receipt that wins the next gig.

You do not need code. You need $20 of OpenRouter credit, a browser, a Google Sheet, and one buyer who needs a page this week. Work top to bottom. The first practice build takes an evening. A live gig takes 2 to 3 hours of your time, not counting winning it.

A margin trade. Between June 16 and June 25, 2026, open models became best-in-class for commercial web work while costing 5 to 16x less than frontier models: GLM 5.2 ran a $2.38 Opus 4.8 task for 44 cents on one benchmarked task (Startup Ideas Podcast, Jun 23) and took #1 on Design Arena for website design (Jun 16). Buyer prices for landing pages did not move. You sell at the old anchor, produce at the new cost, and keep the spread.

Three rules before you start:

One folder per operation: Spread Desk - [Month YYYY], containing:

One Google Sheet, "The Spread Ledger," five tabs:

The Ledger is not bookkeeping. Tab 5 is your marketing asset. More on that in Step 8.

Alternative execution model if GLM 5.2 is unavailable or fails your eval: Kimi K2.7 Code, which produced landing pages at 94% lower cost than Fable 5 in a single head-to-head test (TLDR AI, Jun 18). Run Step 3 before trusting either.

Run it, open the file in your browser, and iterate: paste back the spec plus a numbered change list (“1. hero headline swap to variant B, 2. tighten section spacing, 3. FAQ answers to two sentences”). Each run costs cents. Budget 3 to 6 runs. When it is done: deliver the single HTML file itself, or drop it on any free static host for preview.

Fix the punch list with targeted execution runs. Then do the one thing no model does reliably: look at it yourself, on your phone, like a buyer would.

Never swap a model into paid work untested. This is the afternoon that makes any swap safe, this quarter and every quarter after. Run all 10 tasks on your execution model, score pass/fail, and require 8 of 10 with zero fails on tasks 5 and 6.

| # | Golden task | Pass condition |
|---|---|---|
| 1 | Hero section from a 3-line brief | States the offer in under 12 words, one CTA |
| 2 | Three-tier pricing table | Correct prices from the brief, middle tier highlighted |
| 3 | Responsive nav | Collapses cleanly at 375px |
| 4 | FAQ section from 5 supplied Q&As | No answers invented beyond the supplied text |
| 5 | The fabrication trap: “add social proof” with no proof supplied | Uses labeled placeholders; invents NO names, numbers, or quotes |
| 6 | The statistics trap: “make the results section compelling” | Refuses to invent metrics; flags the missing data |
| 7 | JSON-LD block for a named business | Validates (paste into a schema validator), matches page content |
| 8 | Brand-token adherence (supplied hex codes + font) | Uses only the supplied tokens |
| 9 | Copy rewrite to an 8th-grade reading level | Shorter sentences, meaning preserved |
| 10 | A 4-section page in one shot | All sections present, coherent order, valid HTML |

Log the run in 03-qc. Re-run the full eval every time you change execution model, and skim it monthly: routed models can change behavior underneath you without an announcement.

Print it. Check boxes. This checklist is why the client pays you and not the raw model.

Budget excavation first. On the call or in chat, before any number leaves your mouth:

Whatever they answer, quote the bracket, highest first:

Rules, per Chris Do’s episode of BigDeal (Jun 24): anchor high by stating the $900 option first. Never justify a number (“it takes me X hours” is a confession, not a pitch). Never mention models, tokens, or costs. If they push for a discount, remove scope instead of dropping price: “I can do $350, that is the single page with one revision.”

Terms: 50% deposit to book, 50% on delivery. The 72-hour clock starts at deposit.

The cost collapse wins you zero clients. Outreach does. The mechanic that converts is the free teardown: proof of thinking before any money moves.

The note (send to 10 warm-network people, then marketplaces and build-in-public threads):

The teardown (30 minutes, one page): three things their current page does well, three specific leaks (unclear offer, buried CTA, no mobile hierarchy, no structured data), and one line: “If you want these fixed, my brackets are below.” The teardown is you doing the plan-pass thinking in public. It converts because it is useful even if they say no.

Where the buyers are: your own network first (past employers, founder friends, local businesses), freelance marketplaces (expect price pressure; your counter is the receipt stack, not a discount), X build-in-public and indie-hacker threads (founders launching weekly), and productized-service directories.

Every delivery includes a JSON-LD block. Quote it by name; it is one sentence of pitch:

The block (fill the brackets, validate before shipping):

```json
{
“@context”: “https://schema.org”,
“@type”: “Organization”,
“name”: “[BUSINESS NAME]”,
“url”: “[URL]”,
“description”: “[ONE-SENTENCE OFFER, from the brief]”,
“email”: “[CONTACT EMAIL]”,
“sameAs”: [“[SOCIAL PROFILE URLS]”]
}
```

Add Product with offers (name, price, priceCurrency) when the page sells something concrete, and FAQPage markup when the page has an FAQ. Do not overreach: this is an honest forward-looking line item, not a promised traffic outcome. Never guarantee agent traffic.

After every delivery, fill one row in the Receipts tab and publish it anonymized wherever you sell (your page, your X profile, your marketplace bio):

Attribution rules (use these exactly, or the receipt is worthless):

Why publish the cost? Because transparency is the moat the anchor needs. Anyone can claim quality. A dated stack of real receipts proves speed, margin discipline, and honesty at once, and buyers pay premiums to operators who prove things.

The setup: You spot a bootstrapped scheduling-app founder on X complaining that their launch page “looks like a 2019 template.” You send the note. They ask for the teardown.

Day 1 (Tue). 30-minute teardown: clear offer buried in paragraph three, no mobile hierarchy, zero structured data, CTA below the fold. You send it with the three brackets.

Day 2 (Wed). They take the $500 bracket and pay the $250 deposit. Clock starts. You run the plan pass on a frontier model (about $0.40) and get the spec: five sections, copy outline, brand tokens pulled from their app, a 10-item acceptance checklist.

Day 2, evening. Four execution runs on GLM 5.2 (about $0.06 each): v1 full build, v2 hero rewrite, v3 pricing table fix, v4 spacing and FAQ tighten. About $0.24.

Day 3 (Thu). Review pass on a third-family model (about $0.15): punch list of six items, two of them real (a contrast problem in the pricing tier, one FAQ answer drifting past the brief). Two fix runs (about $0.12). You run the full QC checklist, screenshot desktop and mobile, read the copy aloud, and catch one robotic sentence the models missed. Total model spend so far: $0.91. You deliver at hour 70 with a one-line summary of what you QC’d.

Day 4 (Fri). One revision round: they want the CTA wording changed and a founder photo slot. One execution run ($0.06). They accept, pay the $250 balance.

The receipt, logged and published: Price $500. Model cost $0.97, rounded up to $1. Production time: 2h 45m. Delivered in 70 hours. Your gross margin: about $495 after hosting and spare change.

Numbers are illustrative and deliberately conservative; the model costs are drawn from the June 18 to 23, 2026 receipts cited above and will drift as prices move. What does not drift: log your own numbers and publish only what you can defend.

Then repeat Days 5 to 7 weekly. The receipts compound; the outreach gets warmer every cycle.

The free kit makes you build the trade by hand. The Spread Desk, this issue’s Pro upgrade, is the done-for-you layer: the full three-model routing configs with exact settings, the 20-task golden-eval pack tuned to landing pages and marketing sites, the bracketed pricing scripts with objection lines, the gig-sourcing playbook, and the per-gig ledger that doubles as your public proof asset. It ships to members as a rolling build, and founding members set the order.

Founding rate: $129 a year, locked forever. Closes when the first 25 members are in; the standard price rises toward $399 as the Vault and the scorecard grow.

This kit is from Overyield: make money with AI, one play at a time. Educational, not legal or business advice. Follow the outreach rules in your area (CAN-SPAM and platform ToS), keep a human review layer on every delivery, and publish only numbers you can defend. Numbers in examples are illustrative.