Long-form essays from the engineers shipping AI inside payers, hospitals, energy operators and proptech platforms. Written for technology leaders who care more about what runs in production than what trended last week.
Most architecture reviews are a rubber-stamp meeting or a fight. A serious review pressure-tests scale, failure, data, and cost before you build. What it actually covers and how to run one.
In healthcare, AI code that looks right but mishandles a dose, a unit, or PHI is a patient-safety and privacy risk. Know how AI code fails so you review for clinical and compliance harm.
In fintech, AI code that looks right but mishandles money, precision, or a control is a reportable incident, not a bug. Know how AI code fails so you review for money and compliance risk.
AI-generated code fails in patterns, plausible but wrong, insecure defaults, subtle duplication. Knowing its quality profile lets your SaaS team review for what AI actually gets wrong.
When AI generates much of the code, output-based productivity metrics like lines and commits stop meaning anything. What to measure instead: outcomes, flow, and the quality of what ships.
In fintech, shipping fast means little if a change breaks payments or fails an audit. DORA metrics reveal delivery health, but stability and change control carry extra weight. What they predict…
DORA's four metrics predict delivery performance well and are gamed easily. Used honestly they reveal how your SaaS team ships; used as targets they mislead. What they measure and what they miss.
Long-lived branches feel safe and quietly kill velocity through merge hell and integration pain. Trunk-based development keeps everyone on one line, integrating daily. When it fits and how to…
In retail, a bad release during a sale costs real revenue. Feature flags let you ship dark, roll out to a store or region, and kill a broken checkout instantly, if you manage their lifecycle.
Feature flags are release infrastructure, not scattered if-statements. Done right they separate deploy from release, enable safe rollouts, and stay clean. Done wrong they become permanent…
Repo strategy now shapes build times, code ownership, and how much context AI tools can see. Monorepo or polyrepo is a real tradeoff, not a religion. How to choose for a scaling SaaS team.
An energy operations dashboard shows "live load" by having each open browser poll a telemetry endpoint every few seconds. It looks responsive in the demo. In production, dozens of operators and…
A fintech app shows a "live balance" by having the client poll the ledger every few seconds. It looks instant in the demo. In production the polling multiplies across every open app, hammers the…
A hotel platform adds a "rooms left" indicator to its booking flow by having the browser poll an availability endpoint every few seconds. It demos beautifully. In production the polling multiplies…
A retailer adds a "live stock" badge to product pages by having the browser poll an inventory endpoint every few seconds. It looks great in the demo. In production the polling multiplies across…
A product page, a mobile app, and a partner feed each want different data from your commerce API. GraphQL, REST, and RPC fit different retail consumers. How to choose by who calls, not by hype.
GraphQL, REST, and RPC each fit different consumers. Picking by fashion instead of by who calls your API is how SaaS teams end up with the wrong one. How to choose by consumer, not hype.
In fintech, an API change can break a partner's payment flow or a regulator's reporting feed. Versioning that ages well lets your API evolve without breaking money movement, integrations, or…
Every API change risks breaking a customer's integration. Good versioning lets your SaaS API evolve without breaking consumers, including your own agents. The strategies that age well, and the…
A hospitality platform serves many hotel groups and properties from one system. Multi-tenant architecture decides how you isolate them, survive one brand's peak-season surge, and keep guest data…
One codebase, many customers. Multi-tenant architecture decides how you isolate tenants, contain noisy neighbors, and see per-tenant cost. The patterns that scale, and the ones that trap you.
Most SaaS scale-ups reach for microservices too early and pay for it. A modular monolith gives you clean boundaries without the distributed-systems tax. When it is the right call, and how to build…
In fintech, a shared model tangles payments, ledger, risk, and compliance until every change is a negotiation. Domain-driven design draws bounded contexts that survive org changes and keep each…
When a SaaS platform grows, the model tangles and every team steps on every other. Domain-driven design draws bounded contexts that survive org changes and let teams ship independently.
One long-form essay every other Wednesday. Written by the engineers shipping production AI for our clients, not by a content team. No promotional emails. Unsubscribe in one click.
