Blog
Practical guides to AI-driven development, full-stack architecture, and solving production problems.
Showing 37–48 of 100 articles
Separate domain rules from framework plumbing without overengineering
Keep business decisions separate from HTTP and database plumbing with a small function boundary, explicit inputs, and behavior-focused verification.
Write an architecture decision that explains the tradeoff
Write a concise architecture decision with concrete constraints, credible alternatives, accepted costs, and evidence that would trigger a review.
Draw a system boundary from one real user journey
Trace one user outcome across UI, authorization, storage, and providers to draw a useful system boundary with explicit ownership and failure states.
Choose a modular monolith before reaching for microservices
Compare deployment, ownership, and failure boundaries before choosing microservices, then build module boundaries you can verify inside one application.
Use synthetic data to test AI generated database changes
Build deterministic synthetic fixtures around invariants, legacy shapes, races, and privacy boundaries so AI-generated database changes can fail meaningfully.
Avoid exposing private fields through populated database records
Trace selected and populated fields through public serialization, use explicit response views, and test records that contain private extras and missing relations.
Debug date grouping that changes across server timezones
Store instants separately from calendar dates, choose an IANA business timezone, and group with explicit boundaries that survive server moves and daylight shifts.
Decide when several database writes need a transaction
Identify the invariant shared by multiple writes, compare single-document modeling, and test rollback without including external side effects in a retryable transaction.
Normalize account email addresses without merging users accidentally
Define email identity policy, preview normalization collisions, and protect login uniqueness without silently combining distinct accounts or provider-specific aliases.
Design a migration that can be checked before it writes
Preview migration candidates, enforce write-time preconditions, process bounded batches, and verify invariants so a dry run predicts a reviewable change.
Fix a cached database connection that never recovers
Reset failed connection initialization without discarding healthy pooled clients, distinguish query errors from connection setup, and test recovery in one process.
Design deletion behavior without leaving invisible orphan records
Choose deletion semantics for each relationship, prevent new dependents during cleanup, and verify retained history and removed private data separately.
Explore topics
Subscribe to Newsletter
Get the latest articles and updates delivered to your inbox.