Ruslan Akchurin https://ruslanakchurin.dev/ 2026-07-06T00:00:00Z Ruslan Akchurin Appendix B: The contract, end to end https://ruslanakchurin.dev/blog/making-iac-boring/appendix-b 2026-07-06T00:00:00Z 2026-07-07T00:00:00Z One self-contained file implements the cross-tier contract: a producer publishes a typed value, a consumer names producer, key, context, and type, and a resolver returns the bound value or refuses by name. A library is not an architecture https://ruslanakchurin.dev/blog/library-not-architecture 2026-07-06T00:00:00Z 2026-07-07T00:00:00Z The Pulumi component library I built made infrastructure easier to create before ownership, tiers, contracts, and release boundaries were explicit. Delivery is a transaction https://ruslanakchurin.dev/blog/where-secrets-go/02-delivery-is-a-transaction 2026-07-05T00:00:00Z 2026-07-01T00:00:00Z Secret delivery breaks when writes, wiring, and receipts are separate chores; the transaction is the unit that keeps partial success and proof limits visible. The bundle is the competitor https://ruslanakchurin.dev/blog/where-secrets-go/03-the-bundle-is-the-competitor 2026-07-05T00:00:00Z 2026-07-05T00:00:00Z The working bundle around a secret store can deliver a value; rotation, revocation, and honest proof limits are what test whether it delivers one verifiable transaction or evidence held together by convention. The green placebo https://ruslanakchurin.dev/blog/green-placebo 2026-07-04T00:00:00Z 2026-07-03T00:00:00Z An autonomous coding agent ran a ratified security plan on a secrets manager and twice recommended the less-secure option, every fact correct. The green test suite certified a defence that did nothing. Here is what actually held the line, and why the human could not be removed from it. A secret has to arrive https://ruslanakchurin.dev/blog/where-secrets-go/01-a-secret-has-to-arrive 2026-07-03T00:00:00Z 2026-06-29T00:00:00Z Secret storage gives a value somewhere to rest; secret delivery records the route, authority, wiring, and proof that get it there. Start with the shape https://ruslanakchurin.dev/blog/making-iac-boring/01-shape 2026-06-25T00:00:00Z 2026-05-17T00:00:00Z By the time anyone reviews an IaC codebase, its shape is already decided. Resolve by contract https://ruslanakchurin.dev/blog/making-iac-boring/02-contract 2026-06-25T00:00:00Z 2026-05-19T00:00:00Z Cross-stack dependencies make another stack's outputs a public interface. A typed contract on a producer namespace and key is the alternative. Define tier membership https://ruslanakchurin.dev/blog/making-iac-boring/03-membership 2026-06-25T00:00:00Z 2026-05-24T00:00:00Z Environment boundaries blur when workload-specific members (service accounts, IAM bindings, DNS records) accumulate in the shared environment tier. Fail before apply https://ruslanakchurin.dev/blog/making-iac-boring/04-fail-before-apply 2026-06-25T00:00:00Z 2026-06-07T00:00:00Z An IaC contract holds only if the release path enforces it: a resolver checks type and context, then refuses invalid composition before apply. Re-cut the system https://ruslanakchurin.dev/blog/making-iac-boring/05-re-cut-the-system 2026-06-25T00:00:00Z 2026-06-24T00:00:00Z Refactoring a live IaC system toward a new architecture: freeze the old boundary, publish the new contract beside it, delete only after consumers move. Appendix A: The shape, runnable https://ruslanakchurin.dev/blog/making-iac-boring/appendix-a 2026-06-25T00:00:00Z 2026-06-25T00:00:00Z A companion repository makes the series architecture executable: four tiers, typed cross-tier contracts, and resolver probes that refuse bad composition before any provider call.