Featured
API Gateway Strategy in the Agent Era for Retail
Shopping agents look like scrapers to a gateway built for humans. Separate north-south, east-west, and agent traffic, and decide which machine clients you actually want before peak.
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.
Shopping agents look like scrapers to a gateway built for humans. Separate north-south, east-west, and agent traffic, and decide which machine clients you actually want before peak.
Platform engineering is not DevOps renamed. DevOps is a way of working; a platform is a product with users. The difference shows up in how you decide what to build.
Blanket restart automation quiets the pager and hides defects across thirty teams. Narrow, verified, rate-limited remediation loops fix known failures while keeping real problems visible.
Agent traffic breaks the assumptions your gateway was built on. Separate north-south, east-west, and AI traffic deliberately, because they need different limits and different failure behavior.
Namespaces, vclusters, or separate clusters is a blast-radius decision, not a cost decision. Pick the isolation your worst tenant justifies, and price the operational overhead honestly.
Infrastructure agents are moving from suggesting changes to making them. In a multi-team SaaS org, what makes them safe is scope, permissions, and a blast-radius budget, not a better model.
In fintech, the isolation model has to satisfy both blast radius and auditors. Namespaces, vclusters, or separate clusters is a decision you must be able to justify in writing.
In energy, a Terraform module is where your controls actually live. Narrow the interface, put compliance in the defaults, and version it so evidence stays consistent.
Operators are the right abstraction when a resource has real lifecycle logic that many teams need. Most of the time a Helm chart is enough, and writing one anyway costs you a maintainer.
Control plane or pipeline is the real question. Crossplane reconciles continuously and suits self-service; Terraform plans deliberately and suits change review. Most SaaS orgs need both.
Energy workloads carry long-lived data and heavy compute. Make budgets and retention platform primitives enforced at provisioning, so cost is a design decision rather than an annual surprise.
Agent traffic breaks rate limits built for humans and stresses idempotency on payment APIs. Separate north-south, east-west, and agent paths, and make every retry safe.
A Terraform module is an API your whole org consumes. Design the interface for the caller, version it properly, and stop exposing every provider argument as a variable.
Agents that act on infrastructure are arriving in energy engineering. What makes them safe near regulated and operational systems is scope, permissions, and a blast-radius budget, not model quality.
Buying a developer platform does not remove the work, it moves it. The integration, golden paths, and ownership are yours either way. Budget for the part nobody puts on the slide.
Telemetry tells you what happened; surveys tell you what it felt like. Measure both, never rank teams with them, and watch the friction metrics that actually predict retention.
Self-healing infrastructure in energy earns trust one narrow remediation at a time. Start with the failures you already fix the same way every time, and never let automation hide a real problem.
Retail spend is seasonal, so annual cost targets hide the real problem: capacity provisioned for peak that never comes back down. Make scale-down a guardrail, not an intention.
Monthly cost reviews find waste six weeks after it started. Make budgets a platform primitive enforced at provisioning, so thirty teams cannot create spend nobody approved.
AI assistants on the energy developer platform are now expected. The value is grounding them in your real systems, standards, and OT boundaries, not a generic chatbot that guesses about regulated infrastructure.
In fintech the build vs buy question includes controls and evidence. Buying the portal is usually right; the golden paths, approvals, and audit trail are always yours to build.
Retail engineering runs on a seasonal clock. Buy the portal, build the golden paths, and time the whole programme so nothing lands near peak trading.
In fintech, automated remediation has to fix the symptom without hiding the defect or breaking the audit trail. Narrow, verified, rate-limited loops are what earn trust here.
In fintech, self-service can't mean a free-for-all. Guardrails must enforce compliance and security at provision time, so teams move fast without creating regulatory risk.
In a fast-scaling SaaS org, a platform team approving every provision is the bottleneck. Self-service with guardrails lets product teams move at their own speed, safely.
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.