Workflow reliability writing portfolio
These four long-form samples translate automation risk into decisions a business, operator, or engineer can inspect. Each piece states its sources and boundaries, separates executed proof from inference, and avoids presenting self-directed work as a client result.
Four distinct assignments
01 - Business audience
2,537 words
Hours saved are not enough. An automation investment should account for expected failure loss, review cost, and the value of the work it displaces.
Evidence: Browser-only risk-adjusted ROI calculator, NIST AI RMF, Google SRE, and n8n primary sources.
Read this field guide02 - External input
2,709 words
Retry support is not outcome proof. Stripe, AWS, and n8n each cover a different layer of identity, execution, and recovery.
Evidence: Current first-party Stripe, AWS Builders' Library, and n8n documentation translated into one explicit contract model.
Read this field guide03 - Advanced technical topic
2,241 words
A duplicate-delivery guard must distinguish safe pre-effect failure from an ambiguous crash after an external consequence.
Evidence: Deterministic local replay harness, atomic claim model, failure matrix, and tested manual-reconciliation boundary.
Read this field guide04 - Uncommon argument
2,457 words
A review branch can increase risk when demand, capacity, ownership, expiry, and overload behavior are imaginary.
Evidence: NIST oversight responsibilities, Google SRE overload guidance, and n8n's approval primitive kept in distinct roles.
Read this field guideEditorial boundary
The collection proves that One Workflow Fixed can research current primary sources, explain a technical system to different audiences, expose assumptions, and connect an argument to reproducible tests or calculations. Source links are attached to the claims they support, and generated examples are labeled rather than passed off as field observations.
These are independent portfolio pieces created with Codex assistance and manual verification. They are not commissioned articles, interviews, paid-client work, professional-journalism credits, or evidence that a third party achieved the illustrated business outcomes.
Every article names where the evidence came from, where the claim stops, how the implementation can fail, and what record would prove the intended outcome.
Commission a technical field guide
The pilot covers one 2,000-2,800-word technical field guide or comparison, primary-source research, a linked source list, one revision, and a seven-business-day delivery target after written scope and payment.
Before any charge, we confirm the audience, topic, acceptance criteria, byline, rights and exclusivity, AI-assistance policy, deadline, and payment terms in writing. Interviews, original reporting, confidential-system access, and legal, medical, or financial advice are outside this pilot unless separately agreed.
Need the same standard applied to a workflow?
The $25 review returns five prioritized reliability risks, concrete failure examples, and exact retest steps. No credentials or production access are requested.