I Burned 100M Tokens on pSEO. Here's Everything.

01/10

Why this exists

I share everything because nobody else will

Most pSEO content is theory. "Ship quality content." "Build topical authority." "Be patient." None of it includes the actual numbers, the actual GSC screenshots, or the actual moment Google deletes 1,368 pages in a single day.

I run 5 Claude Max subscriptions ($100/month each) from a laptop in Da Nang, Vietnam. I push roughly 100 million tokens per month through multi-agent pipelines using tmux session management. I have built, deployed, crashed, and partially recovered 4 websites in 6 months. The dataset is unique because it includes the same operator making the same category of mistakes across concurrent domains. Pattern recognition across the portfolio is what produces the rules below.

What Lifetime members actually receive

// PRODUCTION FILES — NOT A COURSE

.md

Sprint documentsThe exact format that scores 10/10 on QA gates

.md

CLAUDE.md filesAgent instruction files for each sprint type

.sh

QA gate scriptsBash validators, binary PASS/FAIL, zero self-reporting

.md

Agent governanceApproval flows, irreversibility, drift detection

.md

Deploy checklistGooglebot-aware scheduling, rollback triggers

.md

Content pipelineDataForSEO → published page in under 4 hours

.md

Extension Engine31 playbooks for Chrome extension growth

.md

Satellite protocolLaunch, measure, scale or kill in 30 days

.conf

tmux fleetSession configs for 5/10/25 agent orchestration

.json

DataForSEO configsAPI templates, polling, keyword clustering

// EVERY FILE UPDATES WHEN I LEARN SOMETHING NEW · YOU GET EVERY FUTURE ADDITION

02/10

The Portfolio · four domains, four datasets

Four domains, four experiments, four datasets

Every rule in this document is verified against GSC data from all four sites. Not one site. Four. The cross-domain pattern is the evidence.

belikenative.com Primary

Clicks · 90d4,933

Daily avg~55

StrategyGradual

ResultSteady growth

claudecodeguides.com Frozen

Peak91 / day

Now0 / day

Indexed17 / 3,558

ResultDomain killed

zovo.one Live

Clicks · 21d176

Indexed8,502

HighlightTool +344%

Not indexed173,436

securetoolsguide.com Satellite

Indexed7

Not indexed1,010

404 errors67

5xx errors2

GSC Evidence · All Four Properties

Google Search Console overview for belikenative.com showing 4,933 total clicks with steady 40-90 clicks per day — belikenative.com — 4,933 clicks. Steady growth. No velocity spikes, no penalties. The control group.

Google Search Console overview for claudecodeguides.com showing 806 total clicks with traffic spike to 91 clicks/day then crash to zero — claudecodeguides.com — 806 clicks. Spike to 91/day, then flatline. 3,558 not indexed, 17 remain.

Google Search Console overview for zovo.one showing 176 clicks, 8,502 indexed pages, and css-gradient-generator up 344% — zovo.one — 176 clicks, 8,502 indexed. CSS gradient generator up 344%. Tools outperform articles.

Google Search Console page indexing for securetoolsguide.com showing 1.01k not indexed vs 7 indexed pages — securetoolsguide.com — 1.01k not indexed, 7 indexed. 67 pages returning 404. Same velocity mistake.

03/10

The Crash · 91 → 0 in 48 hours

How I killed 91 clicks/day in forty-eight hours

claudecodeguides.com went from 91 daily clicks to zero. From 2,510 indexed pages to 17. This is the exact timeline, verified against GSC and git logs.

March 24 — April 14

Organic growth: 1 → 91 clicks/day

~2,800 articles indexed naturally. 9,023 peak daily impressions. Zero manual intervention. Everything worked.

April 15—17

Five sprints in one day

2,614 CTA injections, copy rewrites, homepage redesign, lead magnet. Impressions begin declining: 7,609 → 6,659.

April 18—20

13,000+ file changes in 48 hours

8,452 files across Sprints S19–S22. Multiple build failures. Stale content served during Googlebot crawls.

April 21

1,368 pages de-indexed in a single day

Indexed: 2,510 → 1,142. Active quality action. Not gradual. Not algorithmic drift.

April 22 — present

Flatline. Zero impressions/day. 17 pages indexed.

1,535 crawled and rejected. 985 not crawled. Domain effectively dead. Frozen for recovery observation.

Indexation collapse · each square ≈ 22 articles

3,575 of 3,760 articles — erased.

17 indexed 3,558 not indexed · ratio 1 : 209

0.45% survive

2,510

Peak Indexed

Indexed Now

3,558

Not Indexed

404 Errors

Canonical Issues

Noindex Excluded

Exhibit B · GSC Page Indexing · Verified

Google Search Console page indexing for claudecodeguides.com showing 3.56k not indexed vs 17 indexed pages with canonical issues and 404 errors — claudecodeguides.com page indexing — 3.56k not indexed, 17 indexed. Canonical issues (11), 404s (9), noindex excluded (15). Google actively removing pages.

04/10

The Three Velocity Rules

The numbers that decide whether Google kills your site

Every domain penalty I have observed, across all four sites and dozens of satellite domains, traces back to violating one or more of these.

01
Never ship more than 50 pages per sprint on a domain under DR 30.

              CCG shipped 3,760 pages in 30 days on DR 5. STG pushed 1,010 pages with 7 indexed. BLN shipped ~200 pages over 4 months and has 4,933 clicks. The velocity threshold scales with domain authority. New domain: 10–20/week. Established DR 20+: 50/week. Authority DR 50+: no practical limit.
              
                CCG → 3,760 pages / 30 days  =  killed

                STG → 1,010 pages bulk  =  7 indexed (0.7%)

                BLN → ~200 pages / 4 months  =  4,933 clicks, no penalty
              
02
Never change more than 200 files in a single deploy.

              Sprint S21: 3,647 files. Sprint KW-3: 3,553 files. Google interprets massive simultaneous changes as site-wide bot rebuild. Normal healthy sites change 5–50 files per deploy. If your commit touches 200+, split into batches with 48 hours between each push.
            
03
Forty-eight hours minimum between deploys.

              CCG had 10 commits in 48 hours. Each deploy triggers a re-crawl. Rapid re-crawls on changing content equals "content churn." Deploy once, wait 48 hours, verify indexation, then deploy again. Sprint 14 (10/10) shipped once. Sprint S21 (the domain killer) was one of 6 deploys in 2 days.

Signal	What I Did · CCG	What Works · BLN
Pages / month	3,760 in 30 days	~50 / month, gradual
Files / deploy	3,647	Under 50
Deploy cadence	10 commits in 48h	1 per 48 hours
Domain age	30 days old	8+ months
Measurement	0 checkpoints / 14 sprints	7-day check per sprint

05/10

Content Quality · the moat

Information Gain is real. Templates are the enemy.

04
Template articles are a liability, not an asset.

              "Claude Code for [X] Workflow Guide" × 200 with the same 6-section structure, same word count, same CTA placement. Google scores whether your page adds information not available elsewhere. Template + keyword swap = zero Information Gain. Ship 50 excellent pages, not 500 template pages.zovo.one proof → CSS Gradient Generator (one interactive tool) got 344% more impressions. One tool outperformed hundreds of CCG template articles.
05
Interactive tools survive. Articles do not.

              AI Overviews killed informational CTR in 2026. Articles are replaceable by AI summaries. Tools that require user input are not. Every content site needs pages that pass: "Would this be useful if Google disappeared tomorrow?" Articles drive traffic. Tools convert. Both required. Tools are the moat.
            
06
Topical dilution is a measurable penalty trigger.

              CCG had 588 non-Claude articles (19% of pages). STG mixed security and general tools. Both saw indexation collapse. BLN maintains strict topical focus on writing tools and has steady growth. The data is unambiguous: dilution kills indexation.
                CCG → 19% off-topic = 99.5% de-indexed

                STG → Broad mixing = 0.7% indexed

                BLN → Strict focus = 55 clicks/day, no issues
              

06/10

Validated Patterns · what survives

What actually works, across all four domains

✓
Gradual growth on established domains works every time.
BLN: 4,933 clicks over 90 days. Steady ~55/day. No spikes, no crashes. Content shipped gradually, measured at each step, topically focused. This is the control group. CCG and STG are the experiments that failed.
✓
$99 converts from cold traffic without Google.
Three customers during a 96% crash. Two paid $99 from payment links, not search. K0ntrast SARL is a French B2B company. If people pay when everything is broken, the product works. Fix distribution, not the product.
✓
Authentic first-person copy outperforms generic marketing.
"I run 5 Claude Max subscriptions from a laptop in Da Nang" converts better than "What if Claude Code got it right the first time?" $400K Upwork, 50K extension users, 100% JSS are facts no competitor can replicate.
✓
Machine-verified QA gates are the only gates.
Sprint 14 (10/10): validated with grep, wc, curl. Binary PASS/FAIL. P1 first attempt (6/10): self-reported, WARN category. Gates are PASS or FAIL. Never WARN. If the report cannot produce raw numbers, the validator does not exist.

Exhibit C · Stripe Revenue · During 96% Crash

Stripe dashboard showing 254 total payments, 184 succeeded, including $99 lifetime deals and recurring subscriptions — 254 payments, 184 succeeded. $99 lifetime deals converting from cold traffic. Revenue continued despite the CCG crash — proof that the product works independent of Google.

07/10

Agent Governance · 25 agents, 1 operator

Running 25 agents without killing your domain

07
Max 5 agents per sprint. Max 1 sprint deploying at a time.
15 agents across 3 PMs deploying simultaneously caused git conflicts, competing builds, and rapid-fire commits. One sprint deploys. Others wait. Measure. Then next.
08
Never batch-approve irreversible actions.
Bulk title rewrites on 2,333 articles. Bulk CTA injection on 3,370 articles. Each irreversible at scale. Irreversible actions require: code review, per-batch confirmation, explicit sign-off. Time pressure never justifies compressing verification.
09
"Shipped" means verified live.
Sprint 8 declared "9/10 complete" with edge functions never deployed. 731 uncommitted changes sat on disk 24+ hours. "Shipped" = curl -sI returns 200 AND content matches AND functionality works.

08/10

Pre-Deploy Checklist · run this every time

Run this before every content push

MUST

Is this commit under 200 files? If not, split into batches with 48h between each.

MUST

Has it been 48+ hours since the last deploy? No exceptions.

MUST

Did I measure the previous sprint? Pull GSC data. No sprint ships without data from the last one.

MUST

Does the build pass with zero errors? Grep for errors. Fix all before pushing.

MUST

Under 50 new pages this week? Count total across ALL commits.

SHLD

Does every page add unique information? Template + keyword swap = do not ship.

SHLD

5 random URLs return 200 within 5 minutes? Check rendered content, not just status.

NICE

Would this page be useful without Google? If not, make it interactive — or kill it.

09/10

The Framework · apply to every project

Apply this to every future pSEO project

I
Phase 1 · Foundation — Week 1–2
10 pages max. Homepage + 5 content + 3 tools + 1 hub. Payment, email, attribution live from day one. Do not expand until 10 pages are indexed and getting impressions.
II
Phase 2 · Validation — Week 3–6
20 pages/week. Measure after each batch. If indexation drops below 80%, stop and diagnose. Total: ~100 pages.
III
Phase 3 · Scaling — Week 7+
Only enter if Phase 2 metrics are healthy. 50/week under DR 30. Every page must pass the Information Gain test.
IV
Continuous · Distribution — from day 1
CCG had zero non-Google distribution. If 100% of traffic comes from one source, you have 0% the day it cuts you off. Post first, sprint second.

10/10

The Playbooks · the execution layer

The execution layer nobody else publishes

The rules above are the what. The playbooks below are the how — exact .md files, agent configs, bash scripts, and tmux layouts I use in production every day.

P1
Multi-Agent Fleet Configuration · 5 Claude Max Subs · tmux Orchestration · Task Routing
The exact tmux layout and session naming convention. Which agent types get which subscription. How to detect agent drift. PM hierarchy. Cost per agent-hour. Why 5 subs is the sweet spot.
P2
Sprint Document Template · The .md Format That Produces 10/10 Sprints
Sprint 14 format. Section by section with fields, constraints, verification commands. The 4-field difference between 10/10 and 6/10.
P3
CLAUDE.md Agent Instruction Files · Per-Sprint-Type Configuration
Files that control agent behavior per sprint category. Content sprints, CRO sprints, deployment sprints, QA sprints — each gets different instructions.

— Behind the gate —

Ten production playbooks

Fleet configs · sprint .md templates · CLAUDE.md files · QA scripts · deployment automation · satellite protocol · pSEO pipeline · Extension Engine (31 playbooks) · domain migration · revenue attribution.

Unlock all → $99 Lifetime

The Offer · One Payment · Forever

Every playbook. Every future update. Every .md file I use in production.

This is not a course. These are the actual operational documents from running 100M+ tokens/month on multi-agent pSEO. You get the same files I use.

✓

10 Production PlaybooksFleet config, sprint templates, QA gates, deployment, satellites, content pipeline

✓

CLAUDE.md FilesAgent instruction files for every sprint type, copy-paste ready

✓

QA Gate Scripts (.sh)Bash validators with binary PASS/FAIL — the scripts that caught regressions

✓

Extension Engine — 31 playbooksThe system behind 16 extensions, 3,400+ users, 4.6 stars

✓

DataForSEO ConfigsAPI templates, polling scripts, keyword clustering, long-tail workarounds

✓

Every Future AdditionNew playbooks, new sprint formats, new tools. Lifetime means lifetime.

Get Lifetime Access → $99

ONE PAYMENT · NO SUBSCRIPTION
INCLUDES EVERY FUTURE PLAYBOOK