Hakaru Lab | Published: 2026-09-03 | How we verify
Methodology M1.0 (Established 2026-09-03)
This methodology is subject to revision. When revised, we will update the version number and clearly state the revision history and comparability (comparable / limited / not comparable).
Who this is for: Readers of our lab's articles who want to know how much they can trust the conclusions for their own decisions before reading the main text.
Key points (3 lines)
・This page outlines the verification rules common to all articles by our lab (hakarulab.com), which clearly state their verification status as "hands-on verification," "research review," or "preliminary observation" at the beginning and strictly separate facts, estimates, and impressions.
・It helps readers (including those planning to start using a product or gathering information) who want to use our lab's articles to decide on adopting products or services quickly understand how reliable the conclusions are before reading the main text.
・The rules stated here apply to all articles, while the specific verification environment, timing, and conflicts of interest for individual articles are explicitly stated at the beginning of each article; this page itself is version M1.0 as of September 3, 2026.
Our lab (hakarulab.com) is a media outlet that introduces products, services, and tools based on records of actual use. This page summarizes what we verify before writing an article, how we separate the results in our writing, and how readers should interpret those results. It is a page we ask you to read through once before using our articles' conclusions to make your decisions.
Our lab is particular about how we verify things not to make ourselves look trustworthy, but to enable readers to make correct judgments on their own just by reading our articles. To achieve this, our fundamental rule is to never claim anything stronger than what we have confirmed. If the evidence is weak, we avoid definitive statements, and we present limitations, conflicts of interest, and unverified areas before our conclusions.
We provide verifiable links for technical terms appearing for the first time and the tools and services used for measurements. This is to maintain a state where readers can verify things with their own eyes rather than blindly trusting our lab's words.
There is one more thing we want to convey. Our lab does not declare completion. We operate by continuously verifying and correcting. If an article mentions "something not yet implemented," it is not a confession of a hidden secret, but a record of the result of our choice to keep correcting. This is why this methodology has version numbers and articles maintain correction histories.
What our lab verifies is how useful products, services, and tools incorporated into work and daily life are in actual use cases, and whether they present pricing and contract terms in a way that users can understand. We cover subjects our lab has actually verified without limiting ourselves to specific fields or genres.
The intended readers are those who want to read our lab's articles to decide whether to adopt or use a product or service. This includes not only those who are already knowledgeable but also those who want to start using them or are gathering information. However, our lab's articles are not personalized advice tailored to each reader's individual situation. Please make the final decision to contract or purchase based on your own circumstances.
Our lab's articles clearly state one of the following verification statuses at the beginning. There is no hierarchy of superiority among these terms; they simply differ in the subject handled and the time of confirmation.
We proceed through the following 5 stages to reach these statuses.
The records of each stage are shown for each article as a badge (a heading indicating the status) and a condition line (verification environment, duration, confirmation date, etc.) at the beginning of the article.
In our lab's articles, we change the way we write depending on the nature of the information, even within the same text.
The results of the verification apply only within the range of confirmed conditions. At the beginning of the article, we clearly state the verified environment (practical work / isolated environment), duration, confirmation date, target plan or version, region, and language, and distinguish any other conditions as "unconfirmed."
For example, an article that only confirmed the free plan cannot be directly used to evaluate paid plans. Content confirmed in a specific region or language environment does not necessarily apply directly to plans for other regions or other language environments. We restrict the scope in our writing to ensure readers do not interpret the content beyond the confirmed scope.
Whether our lab introduces a product or service is judged by confirming the following three points. We do not sum up scores to determine approval or rejection. This is because a total score would make it seem more precise than what was actually verified.
If the above three conditions are not met, we will not introduce it, regardless of compensation terms. The compensation terms are not used in these three judgments.
For articles with referral links, we include a statement to that effect and a sentence stating, "compensation terms are not included in the evaluation," within the article. The conditions for currently active referral programs are as follows.
| Provider | Compensation terms | Confirmed date |
|---|---|---|
| systeme.io | 60% referral fee (recurring: continuously generated as long as the referred party's usage continues) | 2026-08-30 |
Products or services that fall under any of the following are excluded from introduction before the above three confirmations: those where exaggerated advertising expressions are normalized, those that make cancellation procedures difficult to understand, and those with unclear contract terms.
To maintain the accuracy of our articles, we take action in three stages.
Prevention: Before writing an article, we organize the claims, evidence, confirmation dates, and applicable conditions. We distinguish between similar but different states, such as "planned" and "actual," or "processing started" and "actually delivered." Errors that actually occurred in the past are recorded and cross-checked every time we write or verify a new article.
Detection: In our lab's verification, AI agents (automated programs that perform tasks according to instructions on behalf of humans) handle the practical work under the operator's design and judgment. In a session separate from when the article was written, the results are cross-checked by multiple independent AIs and the operator. Verifying factual relationships and checking from the perspective of a first-time reader are conducted separately. The mere number of entities involved in the confirmation is not used as the basis for reliability.
Response: Depending on the magnitude of the error, we either add notes, temporarily unpublish, or retract the article, and we also check other related articles simultaneously. Corrected records are left at the end of the article and are not deleted. We consider corrections not as "evidence that the verification mechanism worked well," but as "a record of fulfilling accountability later for something we failed to fully verify once."
Furthermore, generalized versions of the tools actually used in this mechanism (such as automated article inspection and proofreading rule sets) are published on GitHub. You can confirm them in their actual form, rather than just through verbal explanations.
Please contact [email protected] regarding errors in articles or feedback on this methodology.
We assign version numbers to this methodology (e.g., M1.0). We increase the major version for changes that could alter the interpretation of results, and the minor version for additions that do not change the meaning. When increasing the version, we clearly state the comparability in the revision history as either "comparable," "limited," or "not comparable."
For major changes where we want to discuss past measurement results alongside new results, we re-measure representative targets using both the old and new measurement methods to connect the results (bridging measurements). We do not rewrite pages of past versions; we leave them as they are as records. We believe that honestly showing when things are not comparable is more trustworthy than making them superficially appear aligned.
| Version | Date | Changes | Comparability |
|---|---|---|---|
| M1.0 | 2026-09-03 | Established. Consolidated and finalized rules for the 3 verification statuses, separation of facts/estimates/impressions, 3 criteria for introduction judgments, prevention and response to errors, and version control. | - (First edition) |
This methodology itself will also be reviewed as we operate. When changes are made, we will record them by stacking versions in this table.
Even if the subjects or languages our lab handles increase, we will not change the following principles.
Even when publishing articles in another language, we do not directly translate the conclusions from the original language. We re-verify the differences in pricing, provided regions, legal systems, language environments, and verification environments for that specific language.