Every item below is a working artefact from real engagements, lightly edited to remove client detail. Half are ungated because a checklist behind a form is a checklist nobody reads.
How to retire a system that runs the business, without a cutover weekend. Includes the strangler facade pattern, business-rule extraction and the reconciliation harness we use on every migration.
Building a scorecard your risk function will accept. Gold-standard dataset construction, weighted scoring for domain severity, and wiring evaluation into CI as a release gate.
Anonymised spend data across 84 client estates: cost per transaction, commitment coverage rates and the ten line items that account for most overspend.
The 78 items we verify before any service takes real traffic — observability, failure modes, rollback, capacity, security and the runbook that has actually been tested.
The ADR format we use on every engagement, with three worked examples from real projects — including one that documents a decision we later reversed, and why.
Auditing against WCAG 2.2 AA without drowning in false positives. Manual test scripts, assistive technology matrices and how to prioritise findings by real user impact.
Severity definitions, incident command roles, communication templates and the blameless postmortem format we use — plus how to make the follow-up actions actually land.
Amara Diallo on the evaluation gap: why the systems that impress in demos fail in deployment, and the measurement discipline that closes it. 52 minutes, with the Q&A.
These are the generic forms. A discovery sprint produces the same artefacts written against your actual architecture.
hello@cymbiote.com