Dargel Solutions

Field report · Adrian Verdan · English

Nobody Opens a Passing Check

Thirty-nine logged days of a two-agent experiment, plus checklists to test whether your own controls held or only looked as if they did.

  • PDF, 193 pages
  • 15 chapters
  • 3 calculators
  • 7 editable example files
  • Data as of day 39 of 90 · 25 Aug 2026

€29.90 Final price · no VAT charged under § 19 UStG

Buy direct

  • EU private buyers
  • Order page, then PayPal

Business or outside the EU? Buy on Gumroad ↗

Created with AI assistance. A human editor reviews and takes responsibility for every title. How we work

See the preview

Real Pages 1, 13, 30 of 193 · unaltered

Real pages, and chapter 1 in full.

Four pages, unaltered from the PDF; chapter 1 in full on its own page.

Read before you buy

Read chapter 1 free: When two agents, and when one is enough →

Complete, with the decision tree. No email required.

Try the seven-question silent-pass audit ↓

The field study in the book: the FENNEC experiment, not Dargel Solutions
Day 39 of 90 · 25 Aug 2026

  1. Page 1: title page “Nobody Opens a Passing Check” with its subtitle and openingPage 1 Title page
  2. Page 6: the dating and vocabulary rules of the book, including the explicit state UNKNOWNPage 6 · Chapter 1 Reading rules
  3. Page 13: a check list, whether anything you publish is irreversible or costs moneyPage 13 · Chapter 1 Four checks
  4. Page 30: the record of the days the loop stood still, failed runs in a rowPage 30 · Chapter 3 When the loop stood still
Page 1 of 193
Page 1: title page “Nobody Opens a Passing Check” with its subtitle and opening
Page 6 of 193 · Chapter 1
Page 6: the dating and vocabulary rules of the book, including the explicit state UNKNOWN
Page 13 of 193 · Chapter 1
Page 13: a check list, whether anything you publish is irreversible or costs money
Page 30 of 193 · Chapter 3
Page 30: the record of the days the loop stood still, failed runs in a row
Page 57 of 193 · Chapter 6
Page 57, Chapter 6, with numbered markers explained in the legend
Page 58 of 193 · Chapter 6
Page 58, Chapter 6, with numbered markers explained in the legend

€29.90 final price Buy direct

The spine of the book, opened.

Chapter 6 is the longest and opens with an admission.

Page 57, Chapter 6, with numbered markers explained in the legend Page 58, Chapter 6, with numbered markers explained in the legend
  1. Chapter 6 opens with an admission: “We built the gate.”
  2. Fourteen green signals and one capability treated as a control, kept apart on purpose.
  3. The limit is stated in the text, not in the small print.
Real Pages 57–58 · Chapter 6

Fourteen green lights, and the question stayed unasked.

Across fifteen cases in one small, carefully watched experiment, fourteen controls supplied a green or pass-like signal while the question that mattered sat outside what they enforced. In the fifteenth, a merely available capability was treated as coverage even though it was not invoked correctly during the documented nine-day period.

Four classes to help you find comparable failures in your own setup; not measured in other setups.

Filled: a green or pass-like signal. Dashed: case C2, an available capability not invoked correctly in nine days. One square per case, not a seal.

Sketch 14 + 1 cases from the book

Try a tool from chapter 6 · English only

What does your passing check leave unchecked?

Pick one workflow you run and name what its green lights actually cover.

Try the seven-question silent-pass audit

No sign-up. This tool never sends or stores your answers; the questions come from the book.

1. Does any entry in your decision log name the same principal as author and approver?

Resolve aliases to the same identity before comparing author and approver. Two spellings of one name do not make an independent review.

2. Could the approval note have been written before the review happened, and would anything show it?

Different names show who is recorded, not whether review preceded approval. Check the sequence against recorded times.

3. For each control: through which branch is a red result reachable?

Name a deliberately faulty test input and the branch that should reject it. A control needs a demonstrated way to fail.

4. Does each control report how many objects it examined, and is that count compared against how many exist?

Compare the examined count with an independently known total. Printing a count alone does not detect missing coverage.

5. Is there a number in any of your passing reports that would alarm you if you noticed it?

A suspicious number in a passing report needs an explicit condition that can change the result.

6. Which of your controls check presence, and are any of them counted as checking correctness?

A check that a file or field exists does not establish that its content is correct. Label those two claims separately.

7. Who reads the output of your controls when they pass — and if nobody does, what changes as a result?

Name the person who reads passing output, or the condition that turns the important finding into a failed check.

Your next check

Answer all seven questions to see your next step.

This counts gaps in your own answers. It is not a system inspection, a benchmark or proof of compliance. Zero reported gaps still needs a real test.

Read the interpretation without JavaScript

Count the second answer in each question as one gap.

  • 0 gaps: Give the control you trust most a deliberately faulty test input, and check that it rejects it.
  • 1–2 gaps: Write them down with a date. An unwritten gap is indistinguishable from a closed one six weeks later.
  • 3–7 gaps: Start with question 3: a control that cannot produce a red result makes every other answer optimistic.

From an answer to evidence

The book shows what happened when these checks met real work.

Read the fifteen cases behind the questions, then adapt the charter, decision log and weekly report examples to your own operation.

15 chapters · 193-page PDF + offline HTML · 3 calculators · 7 editable example files. English original edition; also available in German as a separate edition.

€29.90 Final price · no VAT charged under § 19 UStG

Buy direct

  • EU private buyers
  • Order page, then PayPal

Business or outside the EU? Buy on Gumroad ↗

Read chapter 1 free first →

The FENNEC experiment, not Dargel Solutions

What was measured in the experiment.

The experiment did not run unattended: 16 commits came from the owner.

As of: 25 Aug 2026
39 / 38calendar dates in the action log, with activity on 38
236commits in a separate 40-calendar-day git population, 16 of them carrying the founder commit prefix
81posts published by the harvest
55 / 20decisions logged, 20 with a machine-readable record of who approved them
−€13.00cumulative profit: the experiment earned nothing.

Sketch Six reporting slots, width by days

Five reports produced; W4 has no report, only a row filled in afterwards from the ledger.

Fifteen chapters, every number dated.

Chapter 6 carries the book.

p. 1Ticks: preview pagesp. 193
  1. How to read this book.
  2. When two agents, and when one is enough.
  3. The charter.
  4. The operating loop.
  5. State that survives.
  6. Two agents, not one.
  7. The gate is not the answer.
  8. The second model.
  9. Gates, caps and stop conditions.
  10. Tools at the boundary.
  11. Monitoring and steering.
  12. The record.
  13. Four times we were wrong about ourselves.
  14. What a red light is worth.
  15. The next experiment.
All 15 chapters with a summary
  1. How to read this book. The two voices, and what this was measured on and what it was not.
  2. When two agents, and when one is enough. With a decision tree that has a leaf saying one agent.
  3. The charter. The autonomy grid, the three kinds of switch, and a template with no pre-written legal text.
  4. The operating loop. Eight steps in the order that makes them work, and the outage that cost a weekly report nobody missed for seven days.
  5. State that survives. What crosses the end of a session, and why a prose log cannot be audited.
  6. Two agents, not one. The separation of author and approver, and the four times it broke.
  7. The gate is not the answer. The fifteen cases, in four classes, plus a seven-question self-audit.
  8. The second model. A read-only reviewer with a daily cap: twelve blockers, seven of them inside its own guard.
  9. Gates, caps and stop conditions. Three constructions that fail differently, and the defect that only exists across a series.
  10. Tools at the boundary. Three permission gates a human owns, and a contract clause nobody read before building.
  11. Monitoring and steering. A queue that grew to three dense pages a day against a reading budget of twenty to thirty minutes a week.
  12. The record. Six scheduled reporting slots within thirty-nine days: a two-day W1, five seven-day slots, five produced reports — and for W4 no report, only a row filled in afterwards from the ledger.
  13. Four times we were wrong about ourselves. Four documented self-refutations, and the rule the team applied each time.
  14. What a red light is worth. Five produced reports printing red under a definition that never moved; W4 has no report.
  15. The next experiment. What is open on day thirty-nine, and the criterion by which this book’s own launch counts as a failure.

A book about the middle of the job.

It is about day twenty-two: the loop has run for three weeks, everything is green, the weekly report says red. Is it telling the truth?

A good fit if you

  • already run an agent team on a schedule and want the operating half
  • need to tell a control that held from one that only looked like it did
  • would rather read one case measured honestly than ten described vaguely

Not a fit if you

  • are choosing your first framework or wiring your first agent
  • want a benchmark, a framework comparison or a revenue playbook
  • are looking for a set of prompts to paste

What you download.

  • 193-page tagged PDF with bookmarks, page numbers and a running disclaimer
  • Offline HTML bundle of all fifteen chapters, with no external dependencies
  • Three browser calculators that send and store nothing: autonomy matrix, self-approval check, cycle costs
  • Exactly seven editable example files
    All file names

    CHARTER.md, autonomy_matrix.xlsx, cycle_prompt_pack.md, DECISIONS_format.md, weekly_report_template.md, field_study_numbers.csv and field_study_numbers.xlsx

  • The two number twins are granted under the MIT licence: copy, edit, redistribute and sell work built on them

The film about the field report.

Film · 0:36 · Instrumental music, no voice-over

What the film shows
  1. 0:00 Every check passed. So did the bug. · run-checks --all · gate ........... PASS · tests .......... PASS · review ......... PASS · report ......... sent · report opened by: nobody
  2. 0:04 Day 22. Every control green. · Gate: pass · Tests: pass · Review: pass · Audit: pass · Alarm: pass · Backup: pass · Schema: pass · Links: pass · Budget cap: pass · Kill switch: pass · Sign-off: pass · Health: pass · Its failure looked exactly like its success.
  3. 0:12 Field report · Adrian Verdan · Nobody Opens a Passing Check · Two agents, 39 logged days.
  4. 0:16 Seven questions for every green check. · Can it go red at all? · How many did it actually check? · Did one agent write and approve it? · Who reads it when it passes?
  5. 0:22 All fifteen cases. None held back. · Fig. 6 · four classes
  6. 0:26 An honest promise. · The book, opening page
  7. 0:30 Be the one who opens it. · Field report · 15 chapters · 193-page PDF

Two ways to buy, the same edition.

Direct from Dargel Solutions

€29.90 Final price · no VAT charged under § 19 UStG

Buy direct

  • EU private buyers
  • Order page, then PayPal
  • Instant download available

Via Gumroad

  • Businesses and outside the EU
  • Seller: Gumroad
  • Your country’s VAT is added; Gumroad shows the final price before you pay

View on Gumroad ↗

Special territories and details: How we work

Quick answers.

What exactly is included?

Fifteen chapters as a 193-page tagged PDF and as an offline HTML bundle, three calculators that run in your browser, and exactly seven editable example files, two of them XLSX workbooks, plus the complete number series as XLSX and CSV. Everything works offline.

Is this a framework comparison or a set of prompts?

No. It is one experiment, measured on one day, written down while it was still running. Where a lesson depends on the specific stack it ran on, the template that carries it says so on its own line.

Will the numbers still be current when I read it?

The field study was read on day thirty-nine of ninety. Figures that move are marked as moving, and every number carries its own date and commit.