Objective
This skill guides a demanding technical review in which several analyses are compared, criticized, and consolidated in order to produce a more robust result than any individual contribution.
The objective is not to obtain an artificial consensus, but to progressively improve the technical, pedagogical, and documentary quality of a piece of work.
Use Cases
This method is particularly suited to:
- cryptography;
- cybersecurity;
- software architecture;
- technical documentation;
- RFCs and specifications;
- audits;
- library design;
- GitHub documentation.
Principles
Disagreements are sought
A reasoned divergence is an opportunity to improve the document.
Every criticism must be justified
An objection must explain:
- why it exists;
- which assumptions it calls into question;
- which references or reasoning support it.
Concessions are explicit
When a criticism is accepted, the correction is documented.
When it is rejected, the justification is also documented.
Terminology is standardized
Terms are harmonized with the standards and the literature of the field whenever possible.
Process
Step 1
Analyze the document.
Identify:
- errors;
- ambiguities;
- approximations;
- shortcuts;
- strong points.
Step 2
Evaluate the different levels of reading.
For example:
- popularization;
- engineering;
- academic literature.
A statement may be acceptable for one level and insufficient for another.
Step 3
Identify blind spots.
Look in particular for:
- absent concepts;
- implicit assumptions;
- counterexamples;
- historical incidents;
- limits of the models.
Step 4
Propose corrections.
Corrections must:
- improve precision;
- preserve readability;
- avoid unnecessary jargon.
Step 5
Consolidate.
Produce a revised version incorporating the accepted corrections.
Step 6
Reread like an external reviewer.
No longer defend the text.
Look for what can still be improved.
Expected style
- constructive;
- factual;
- well-argued;
- precise;
- pedagogical;
- transparent about uncertainties.
Avoid:
- the argument from authority;
- personal judgments;
- forced consensus;
- peremptory wording.
Deliverables
Depending on the context:
- technical commentary;
- article;
- README;
- documentation page;
- ADR;
- RFC;
- pedagogical guide;
- GitHub repository.
Success criteria
A review is successful when:
- errors are corrected;
- ambiguities disappear;
- limitations are made explicit;
- pedagogy is improved;
- rigor progresses without sacrificing understanding;
- the final document is more robust than each of the intermediate versions.
Philosophy
Disagreement is not a failure of the review.
It is its driving force.
Quality does not come from the absence of contradictions, but from their argued resolution.
The final goal is not to be right, but to produce a durable, useful and technically reliable document.