Engineering notes — Luis Carrera

Notes

How I run a team and make decisions: playbooks, decision records, postmortems, runbooks and technical articles.

01Playbook

Engineering playbook

The standards I set for a team: a definition of done anyone can check, how code is reviewed, how branches reach production, how new people ramp up, and which meetings exist.

leadershipprocesscode reviewquality

2026-10-14

02Templates

Team templates: 1:1, sprint planning, RFC and risk matrix

Four copy-ready templates for running a team: a 1:1 agenda, a sprint planning sheet, an RFC for technical proposals and a risk matrix with a scoring rule.

templatesleadershipplanningrisk

2026-10-14

03Test strategy

Test strategy: VetNova

What gets tested in a multi-tenant clinic SaaS run by one person directing an AI agent: the layers of the pyramid, what stays manual on purpose, a risk matrix and the known gaps.

VetNovaQAtestingrisk

2026-10-14

04Runbooks

Runbook: customer implementation, from kickoff to hypercare

The phases, checklists and rollback for bringing a new customer's historical data into production: kickoff, migration rehearsals, go-live and hypercare.

implementationdata migrationgo-liverollback

2026-10-14

05Architecture decision records

ADR-001: Hybrid storage for dynamic insurance attributes

How REYBER stores risk attributes that change by insurance line: JSONB plus trigger-kept columns and a projection table for lists, chosen by a benchmark on 10 million policies.

REYBERPostgreSQLdata modelbenchmark

2026-09-22

ADR-002: Tenant isolation with PostgreSQL row-level security

Why VetNova keeps each clinic's data apart in the database itself, with row-level security set per request and a non-owner app role, instead of filters in application code.

VetNovaPostgreSQLRLSmulti-tenant

2026-08

ADR-003: Safety rules enforced in code, not in the prompt

Why JARVIS refuses dangerous tools without a confirmation checked by code, has no destructive tools at all, and answers routine requests on a fast path before the language model.

JARVISAI safetyLLMPython

2026-10

06Postmortems

Postmortem: a dry run that saw less than a third of the data

A blameless review of a bulk correction whose rehearsal counted about 3,400 packages when 11,934 existed: root cause, contributing factors and the checks that came out of it.

postmortemSQLdata qualityroot cause

2026-06-04

07Articles

Dynamic attributes at 10 million policies: what the benchmark taught me

Storing fields that change by product line is easy; searching them fast under row-level security is not. Four strategies on 10 million synthetic policies, and the index that was never used.

PostgreSQLJSONBindexingRLSbenchmark

2026-10-14

Put the safety rule in code, not in the prompt

A prompt is a request, not a guarantee. How a local assistant refuses dangerous actions in code, why some tools should not exist at all, and how to test it with the real model.

AI agentsLLMsafetyPython

2026-10-14

Pagination without ORDER BY: the bug that hides on small tables

Why paging through rows without a stable order can skip or repeat records, why tests on small tables rarely catch it, and the three fixes I now require in any script that reads in pages.

SQLpaginationdata quality

2026-10-14