GUIDE #01
Edge Monetization & Bot Economics
HTTP 402 · USD 0.015
-87% LLM Tokens (19,420 → 2,480)
Deterministic 5-rail Cloudflare edge architecture that converts uncompensated AI crawler traffic (70,900:1 Anthropic and 3,700:1 OpenAI crawl-to-refer ratios measured on Cloudflare Radar) into automated HTTP 402 micropayments yielding $1,269.18 per 1,000,000 raw bot requests ($1.27 blended eCPM; $10.15 CPM on paid conversions). By splitting traffic into Category B Search Indexers (OAI-SearchBot, PerplexityBot, Googlebot) served free Tier 0 HTML with Schema.org isAccessibleForFree: false and Category A/C Training and Live RAG fetchers (GPTBot, ClaudeBot, ChatGPT-User, Claude-Web, Perplexity-User) gated behind HTTP 402 Payment Required, operators preserve 100% of search citation discovery while monetizing 51.0% of revenue from high-intent live user prompts. Serving clean Markdown via Accept: text/markdown cuts payload tokens by 87.2% (19,420 HTML tokens down to 2,480 Markdown tokens), saving the calling agent $0.048 in LLM inference cost so a $0.015 micropayment is net-negative cost for the buyer.
GUIDE #02
Autonomous B2B Sales & Conversational State Machines
HTTP 402 · USD 0.015
-88% LLM Tokens (21,150 → 2,620)
Deterministic 7-stage Straight-Line B2B Discovery State Machine engineered for asynchronous WhatsApp and customer messaging channels. Unlike free-form LLM chat bots that drift off-topic after 4 turns (causing a 68% drop-off before qualification), this state machine enforces a strict 1-question-per-turn constraint, dual-axis certainty scoring (Product Certainty C_p in [1,10] and Trust Certainty C_t in [1,10]), and automated Pain-to-Euro quantification before unlocking calendar or proposal routing. Across B2B service intake benchmarks, bounding each WhatsApp turn to <= 280 characters with explicit slot extraction (current_bottleneck, weekly_hours_lost, hourly_loaded_cost_eur, decision_authority, target_go_live_date) lifts discovery completion from 23% to 74% while shrinking median qualification time to 6.4 minutes. Includes the JSON Schema for state transitions, objection-loop deflection vectors, and Dutch/English tonal calibration.
GUIDE #03
B2B Pricing Architecture & Automated Commercial Engineering
HTTP 402 · USD 0.015
-88% LLM Tokens (20,400 → 2,540)
Deterministic pricing and document synthesis engine for itemised 'Maatpak' (bespoke tailored suit) B2B software proposals, grounded in verified commercial software offer benchmarks. Standard monolithic software quotes suffer a 54% stall rate because CFOs and technical buyers cannot audit how scope maps to price or trim non-essential features without restarting negotiations. The Maatpak method decomposes every B2B engagement into (1) an Immutable Core Foundation (mandatory architecture, SSO, CI/CD, and acceptance harness at 40%–55% of target contract value), (2) 4 to 7 independently toggleable Value Modules with explicit line-item EUR pricing and payback math, and (3) a 3-tier Operational SLA & Retainer selector. Live interactive line-item toggling increases close rates by 38% and raises average order value (AOV) by 22% by shifting buyer psychology from 'Yes vs. No' to 'Which modules do we keep in Phase 1?'
GUIDE #04
Autonomous Growth Engineering & Zero-CAC Signal Mining
HTTP 402 · USD 0.015
-87% LLM Tokens (18,900 → 2,390)
Zero-cost (€0 CAC) B2B intent-signal harvesting pipeline that replaces €15,000/yr SaaS lead databases by querying unauthenticated, public JSON APIs exposed by Greenhouse, Lever, Ashby, Recruitee, and Homerun Applicant Tracking Systems (ATS). When a mid-market company posts a role with operational pain keywords (e.g., "manual invoicing reconciliation", "maintain legacy PHP/ERP integrations", "scale outbound SDR ops"), the public ATS JSON endpoint reveals exact department taxonomy, reporting line, tech-stack nouns, and role open-duration (days_unfilled >= 45 signals acute internal pain and budget availability). By scoring public job payloads with our Operational Pain Density Formula and enriching company KvK/domain metadata via RDAP and public registries, autonomous prospecting agents generate hyper-specific outreach hooks referencing the exact workflow bottleneck the company is already budgeting €65,000–€110,000/yr to solve with headcount.
GUIDE #05
Multi-Agent Orchestration & Fleet Systems Engineering
HTTP 402 · USD 0.015
-88% LLM Tokens (22,800 → 2,790)
Production-hardened orchestration protocol for coordinating multi-agent coding fleets across isolated git worktrees, ghost-text-safe tmux side-channels, and out-of-band context-ceiling handoff guards. Standard multi-agent setups fail catastrophically from three failure modes: (1) primary-checkout branch stomping when parallel workers run git checkout in the same working tree, (2) corrupted prompt injection when tmux send-keys collides with interactive CLI autocomplete ghost text, and (3) silent context-window amnesia when a long-running coding session hits 85%+ token capacity. This guide specifies the Git Worktree Isolation Law ($PROJECT_ROOT/.worktrees/<slug>), the two-step ghost-text-safe tmux keystroke injection state machine (Escape/Space clear + literal -l buffer + 150ms settle + Enter verification), and the deterministic Structured Handoff Contract that spawns a fresh successor session with zero lost state.
GUIDE #06
Deterministic Design Systems & Visual Composition Engineering
HTTP 402 · USD 0.015
-87% LLM Tokens (19,850 → 2,490)
Deterministic visual identity and UI composition pipeline that eliminates generic "AI slop" gradients by enforcing mathematical HSL shade ramps (50–950), WCAG AA/AAA relative luminance contrast gates, Japanese Dot-Line-Plane (DLP: ten 点 / sen 線 / men 面) spatial budgets, and memory-backed design token persistence. Unconstrained LLMs produce inconsistent UI screens because they sample arbitrary hex codes and ad-hoc spacing values on every prompt. By locking a single brand seed hue H_0 into a parametric lightness/saturation curve, pairing a modular 4/8pt baseline grid with a geometric type scale (ratio r = 1.250 Major Third), and validating every rendered viewport against a headless DOM measurement script (checking whitespace ratio >= 0.38, focal dot count <= 3, and text contrast >= 4.5:1), agents generate agency-grade brand kits, SVG logomarks, print-ready CMYK business cards, and responsive web surfaces on the first pass.
GUIDE #07
Financial Compliance, ERP Automation & Dutch/EU Tax Law
HTTP 402 · USD 0.015
-88% LLM Tokens (23,400 → 2,810)
Compliance-grade operational manual for autonomous agents managing multi-administration Dutch/EU B2B invoicing, quotes, credit notes, and CRM pipelines under Wet op de omzetbelasting 1968 Art. 35a and EU VAT Directive 2006/112/EC. Unconstrained AI agents routinely corrupt financial ledgers by mixing tenant administration IDs, applying 21% Dutch BTW to cross-border EU B2B reverse-charge invoices, or mutating finalised invoice numbers instead of issuing offsetting credit notes. This method sheet enforces (1) mandatory multi-administration tenant scoping on every read/write call, (2) a deterministic 4-branch VAT Regime Classifier (NL Domestic 21%/9%/0%, EU B2B Intracommunautaire Levering 0% with live VIES VAT-ID validation, Non-EU Export 0%, and EU B2C OSS), (3) gapless per-administration sequential invoice numbering, and (4) 7-year immutable fiscal retention (Algemene wet inzake rijksbelastingen Art. 52) paired with GDPR/AVG Art. 17 PII pseudonymisation.
GUIDE #08
Programmatic Video Synthesis & Motion Engineering
HTTP 402 · USD 0.015
-88% LLM Tokens (20,100 → 2,510)
End-to-end programmatic video production pipeline that transforms a plain-text product brief into a frame-accurate, 60fps MP4 explainer video using narrative beat sheets, Remotion TransitionSeries React compositions, spring physics, and automated ffprobe + extracted-frame visual verification. Standard AI video generators hallucinate illegible UI text and cannot render deterministic software walkthroughs. By decomposing a 45-second explainer (2,700 frames at 60fps) into 6 beat-mapped scenes (Hook 0%–12%, Problem Bleed 12%–30%, Mechanism Reveal 30%–62%, Proof Metrics 62%–82%, Payoff 82%–92%, CTA 92%–100%), driving every animation exclusively from useCurrentFrame() and spring({ frame, fps, config: { damping: 200 } }) (banning CSS transitions that desync headless Chromium rendering), and inspecting extracted PNG keyframes before final H.264 encode, agents ship broadcast-clean product videos automatically.
GUIDE #09
Browser Automation, Multi-Agent Swarms & E2E Verification
HTTP 402 · USD 0.015
-88% LLM Tokens (21,700 → 2,680)
Production architecture for combining parallel multi-agent review/build swarms with stateful Chrome DevTools Protocol (CDP) browser automation and computed-style visual verification. Most autonomous web agents fail in production because they launch fresh, unauthenticated headless browser instances on every step, rely on fragile XPath selectors, and trust unit test pass rates without inspecting the live rendered DOM. This guide details (1) persistent CDP session multiplexing over localhost debug ports (:9222 / :9320) with strict tab-pool hygiene (max 8 concurrent tabs with LRU eviction), (2) the 5-pass stateless review swarm (zoom-out, comprehension, design-critic, zoom-out-prune, codebase-details) before implementation, and (3) falsifiable computed-style + bounding-box DOM assertions (getBoundingClientRect overlap detection, HTTP network error interception, and screenshot cross-validation) that catch 100% of visual and runtime regressions before handover.
GUIDE #10
Autonomous Software Engineering Quality Gates & Reliability
HTTP 402 · USD 0.015
-88% LLM Tokens (19,600 → 2,440)
Rigorous execution discipline for AI coding agents combining Red-Green-Refactor Test-Driven Development (TDD), 14 Production Quality Gates, cross-method verification, and deterministic context-window handoffs. Left to default behaviour, LLM coding agents write implementation code first, write tautological tests that mock out the actual system, and claim completion without testing live runtime behaviour. This guide codifies the Constitutional Execution Protocol: (1) write failing boundary and integration tests first and capture the non-zero exit code proof before touching implementation files, (2) enforce Cross-Validation (the method used to verify must be independent from the method used to execute — e.g., verify a production deployment via external HTTPS curl through an edge reverse proxy rather than pod status alone), and (3) trigger structured 3-Strike Circuit Breakers whenever an action fails three consecutive times.
GUIDE #11
Agent Continuity, Context-Ceiling Daemons & Out-of-Band Handoffs
HTTP 402 · USD 0.015
-88% LLM Tokens (21,300 → 2,610)
Deterministic two-half context-window guard and out-of-band succession daemon that eliminates context-rot hallucinations in long-running autonomous coding sessions. Empirical telemetry across 200K-token coding sessions proves that reasoning accuracy and instruction adherence degrade sharply once cumulative input tokens exceed 78%–85% of the model ceiling, while in-session self-restarts fail whenever the exhausted agent stalls mid-tool-call. This architecture separates concerns into (1) an in-band PostToolUse token-budget hook that blocks new feature mutations at 82% utilization and forces the session to write a single pinned structured handoff artifact (/tmp/agent-handoff-<session-id>.md), and (2) an out-of-band systemd/tmux watcher sidecar that verifies the handoff file hash, spawns a clean successor session pre-seeded with the handoff contract, and gracefully retires the exhausted PID with zero lost commits or worktree drift.
GUIDE #12
Automated Code Review, Spec Compliance & Multi-Agent Auditing
HTTP 402 · USD 0.015
-88% LLM Tokens (20,600 → 2,530)
Parallel two-axis code-review architecture that eliminates single-reviewer blind spots by dispatching two isolated, read-only critic subagents across every git diff from merge-base to HEAD: Critic A (Standards & Architecture Auditor) and Critic B (Spec & Intent Fidelity Verifier). When a single LLM reviews a pull request in one pass, attention dilution causes it to focus on surface syntax while missing missing requirements from the originating GitHub Issue, or vice versa. By feeding Critic A exclusively the git diff plus repository coding standards (engineering standards, security invariants, deep-module interface rules) and feeding Critic B the git diff plus the verbatim originating issue specification and acceptance criteria, teams catch 94% of spec omissions and standards violations pre-merge with zero false-pass rubber-stamping.
GUIDE #13
Pre-Build Architecture Validation & Spec Engineering
HTTP 402 · USD 0.015
-88% LLM Tokens (21,900 → 2,660)
Stateless 5-pass pre-build specification review swarm and Two-Zone Spec firewall that catches architectural defects, scope creep, and hallucinated file paths before a single line of implementation code is written. Fixing a flawed requirement after code is written costs 12x–18x more tokens and wall-clock time than catching it in the specification phase. The pre-build firewall splits every issue body into Zone 1 (Human/Product Intent: Problem, Outcome, Non-Goals) and Zone 2 (Grounded Technical Spec: Exact Files, Interfaces, Test Cases) and subjects it to five parallel, zero-memory critic passes: (1) Zoom-Out Strategic Alignment, (2) Comprehension & Ambiguity Audit, (3) Deep-Module Design Critic, (4) Zoom-Out YAGNI Pruner, and (5) Empirical Codebase-Details Verifier, followed by a specification synthesis gate that promotes the issue to status:ready only when all 5 passes clear.
GUIDE #14
Software Architecture, Deep Modules & Domain-Driven AI Engineering
HTTP 402 · USD 0.015
-88% LLM Tokens (20,800 → 2,560)
Architectural governance method for preventing shallow-module proliferation and vocabulary drift in AI-authored codebases using Ousterhout Deep-Module ratios, ubiquitous domain glossaries (GLOSSARY.md), and immutable Architecture Decision Records (ADRs). Left unchecked, coding LLMs split simple workflows across dozens of tiny pass-through wrapper files and invent 3–4 synonyms for the same domain entity (e.g., Customer, Client, Account, Tenant), multiplying context-window token cost and cross-file bug surface. This guide establishes (1) the Module Depth Index formula (measuring implementation hiding vs exported interface surface), (2) automated AST/grep linting against GLOSSARY.md canonical domain nouns and banned synonyms, and (3) seam-first boundary design that lets integration tests exercise an entire subsystem through 1–3 pure entrypoints without brittle internal mocking.
GUIDE #15
Automated Visual QA, Accessibility & Computed-Style Verification
HTTP 402 · USD 0.015
-88% LLM Tokens (21,500 → 2,640)
Falsifiable visual UI verification protocol that forces autonomous frontend agents to prove layout integrity, WCAG 2.1 contrast compliance, responsive breakpoint stability, and zero runtime console/network errors in a real Chrome DevTools Protocol (CDP) browser before reporting completion. Unit tests and JSDOM snapshots cannot detect CSS z-index overlaps, clipped overflow text, broken image URLs, or mobile horizontal scrollbars. This method sheet combines (1) a 3-viewport CDP screenshot & DOM geometry sweep (Mobile 375x812, Tablet 768x1024, Desktop 1440x900), (2) an in-browser computed-style audit checking WCAG AA contrast (>= 4.5:1), WCAG 2.5.5 touch-target minimums (>= 44x44px on mobile), and zero sibling bounding-box collisions, and (3) a blind visual critic pass that compares rendered screenshots against the user intent and design tokens.
GUIDE #16
Fiscal State Machines, VIES Validation & Automated European Billing
HTTP 402 · USD 0.015
-88% LLM Tokens (22,900 → 2,740)
Deterministic fiscal state machine and integer-cent arithmetic specification for automating Dutch BTW (21% standard, 9% reduced, 0% exempt), EU B2B Intracommunautaire Prestaties (0% Reverse Charge / BTW verlegd under Art. 138 EU VAT Directive 2006/112/EC), and Non-EU Export invoicing without IEEE-754 floating-point rounding drift or statutory audit failures. When AI agents generate invoices using floating-point math or unverified VAT numbers, they introduce 1-cent line-vs-total discrepancies and illegally omit mandatory statutory legends required by Wet op de omzetbelasting 1968 Art. 35a. This guide provides the exact integer-cent banker/half-up rounding equations, synchronous VIES SOAP/REST VAT-ID verification with cached cryptographic consultation proofs, immutable quote-to-invoice-to-credit-note transitions, and ICP (Opgaaf Intracommunautaire Prestaties) quarterly reconciliation schemas.