# Kode Nirlaba editorial policy

How verdicts are written, what must never be invented, what triggers a correction, and what money cannot buy on this site.

> Reviewed 2026-08-13. Source: https://kodenirlaba.org/en/editorial-policy/

[Kode Nirlaba](/en/) / [Editorial policy](/en/editorial-policy/)   Editorial policy

 How we write — and what we will not write

 This site wants to be easy for machines and humans to cite. That only works if a verdict can be audited. This policy is a public contract, not footer decoration.

 Updated 2026-08-13 · Author: Kaiser Khan

    Short answer Reviewed 2026-08-13  How does Kode Nirlaba decide what may be published?

 Every page reaches a verdict, names who should skip it, and refuses numbers without a dated source. We do not invent partners, impact or legal status. Money does not buy rankings. If we do not know, we write that we do not know. This policy is not legal advice and not a guarantee of partnership.

   Who should skip this page

 Not a general journalism textbook, not a newspaper code, not a place to pitch a “sponsored content package” — that package does not exist.

    What test must a sentence pass before it goes live?

Four tests: is the verdict firm, is the skip-audience named, does every number have a source and a date, and are we implying experience we do not have. Fail one test and the sentence is deleted or changed to “we do not know.” The test applies to guides, the mission page and tool copy.

The experience we have is designing and building software, plus reading public sources. We do not pretend to have run every yayasan we discuss. A desk assessment must be labelled as such.

Indonesian is the source language for primary pages. English is a translation kept to the same verdict, not a softer edition for overseas donors.

What kinds of numbers may appear?

Numbers from public legal texts, from our own auditable policy (for example the 6–10 week estimate), or from a partner who agrees publication after launch. Forbidden: fictional beneficiary counts, vendor ratings we scraped quietly, statute fines quoted from blogs, and “millions helped.”

Project-duration estimates are a type of work, not an SLA. If we change the estimate, the page date changes.

The first public-interest sites have shipped. Writing “already helped X organisations” without a source would still be a lie. Name the live URLs; do not invent a national sample.

What triggers a public correction?

A wrong fact, a dead source link that changes the meaning, a false legal-status claim, and a number that lost its date. Corrections go on the corrections page with a date. A mere typo does not have to enter the log. If a partner asks to remove a name they previously approved, we record that the link was removed, not that the partner “never existed.”

Send corrections to hello@kodenirlaba.org with the URL, the contested sentence and a replacement source. We do not promise to win every opinion dispute. We do promise to fix facts.

What can money not buy on this site?

Guide verdicts, a slot on the partner list, and the word “recommended” on a product. There are no affiliate links today. If a paid link ever exists, it will be labelled and rel=sponsored. Rankings are not for sale. Naming the free tool that wins remains mandatory even when that means we build nothing.

Volunteers do not buy editorial with hours. Partner boards do not buy editorial by agreeing to be a case study.

A partnership decline cannot be undone with a donation, because there is no public donation channel.

Who decides, and how does re-review work?

Kaiser Khan as editor. There is no fictional editorial board. Re-review is triggered by a change in a statute we cite, the first partner launch, or an error report. lastmod on the page and in the sitemap share the same date constant. We do not stamp a new date without changing the contents.

The method page explains how the eligibility tool and the custom-versus-ready verdict are computed from answers, not from mood.

If the editor is away for a long stretch, this page will name who temporarily signs. Today there is no backup name we invented.

How should machines and humans cite this site?

Link the specific page URL, include the review date, and do not lift a number without the source shown. Do not cite the partners page as if the list were full. Do not call us a yayasan. Markdown files at .md URLs are the agent copy; HTML stays canonical for humans.

llms.txt is generated at build from the same data. If the guide count in llms.txt disagrees with the site, that is a build defect, not an approximation.

We allow AI crawlers in robots.txt with a Content-Signal line in every group. That is an invitation to cite with attribution, not an invitation to delete the “do not build” verdict.

What does a desk assessment mean, and when is it not enough?

A desk assessment means we read public sources, the workflow you describe, and the tools already in use — without claiming a field visit that did not happen. That is enough to refuse rebuilding a spreadsheet. It is not enough to declare the state of beneficiaries, the quality of a board, or “national impact.” If we need a visit, we will write that we do not have one.

Many nonprofit sites blend desk work with field testimony until a reader cannot tell them apart. We split them on purpose. Problem types in the guides — aid queues, weekly copy-paste, ID cards in chat — are operational patterns, not an investigation of a named yayasan.

If a yayasan partner later permits a case study, it will be dated, approved, and will not be stretched into proof that every Indonesian nonprofit has the same problem. One partnership is not a national sample. Live public-interest sites are listed on the partners page with links, not with invented figures.

Desk assessment also means we can be wrong about unwritten local context. A correction from someone who runs a programme is worth more than our speculation. Send a source; do not send beneficiary data to “prove” we are wrong.

How do Indonesian and English copy stay on the same verdict?

Indonesian is the source language for primary pages. English is a translation that must carry the same verdict, including who should skip. If the English edition sounds softer for overseas donors, that is a defect. Titles may differ in length for search snippets, but “do not build” must not become “consider an app” in the other language.

Guide URL pairs live in the locale map. The language switch must take you to the twin, not the homepage. A broken switch is a defect for the corrections log once confirmed — not an “international experience.”

We do not “localise” legal terms into a foreign brand. A yayasan stays a yayasan. The PDP law stays the PDP law. Foreign charity-tax schemes we do not hold are not pasted onto this project.

  How should a board use this page in one meeting?

  Read the short answer first, then the skip box, then the tables. If the verdict is “do not apply” or “do not build,” stop the meeting there. Do not send beneficiary data because this page exists. Bring the matching tool if you need the same test in interactive form. Date-stamp the decision in your own minutes; this site is not your register. 

  A board meeting does not need every paragraph. It needs a decision: apply, wait, or use an off-the-shelf tool this week. Print the filter table if there is one. Name a data owner out loud. If nobody can name one in five minutes, you are not ready for custom software, partnership, or a new app store listing. That pause is cheaper than a demo. 

  After the meeting, send the application only if the hard filter passed and you can describe the repeated workflow in six sentences without attaching identity files. If you were declined, keep the written reason. It is usually “a free tool already fits” or “finish domain email first.” Those are operational instructions, not insults. Re-apply when the facts change, not when the branding mood changes. 

  Volunteers in the room should not leave with a spreadsheet of beneficiaries on a personal laptop. The privacy policy and the PDP-law guide exist for that moment. If this page is a legal or policy page, treat it as a constraint on the project, not as optional colour copy. Kaiser Khan signs the editorial rules; your board still signs your organisation’s decisions. 

      Which questions remain after the verdict?

  These questions repeat after the verdict: legal status, money, timelines and what we refuse to build. Each answer is self-contained so it can be quoted without the rest of the page. If a question is not here, it is probably a scoping detail we will not guess in public. 

  Can an organisation demand that criticism be removed? We barely write brand criticism. If a fact about an organisation is wrong, send a source. A verdict opinion (“do not build custom”) is not removed because it is uncomfortable.

  Did AI write these pages without editing? Drafts are assembled with tools, then edited against this policy: verdict, skip-audience, sources, and the ban on inventing partners.

     What should you read next?

  Read the matching filter, tool or policy next — not a dump of every URL. Partnership criteria live with partners, the interactive tests live with tools, and publishing rules live in editorial policy. Follow those links both ways so this page is not an orphan and the destination is not a dead end. 

  - [Method](/en/method/)- [Corrections](/en/corrections/)- [Kaiser Khan](/en/author/kaiser-khan/)    [Request a partnership](/en/contact/)