Which DevOps practices do DeFi protocols actually use?

A practice-by-practice reading of the public repositories of influential, original Ethereum DeFi protocols. Every cell links to the file that produced it, pinned to a commit.

Count

How common is each practice?

Share of protocols where the practice applies and was found. Practices that don't apply to a protocol (for example upgrade checks on an immutable one) are left out of that practice's denominator.

Count a tool as adopted when it is

Yes/no practices are unaffected by the threshold. n = protocols where the practice applies. The thin line over each bar is the 95% interval.

Table view

Said vs. run in CI

For each tool-like practice: how far did adoption get? “Mentioned” means only docs or comments. “Configured” means a config file or script exists. “Run in CI” means a workflow triggered by pushes or pull requests invokes it.

Whether a CI check is required before merging is a branch-protection setting held on GitHub's servers, not in the repository, so it cannot be observed here.

Table view

Do practices come in a fixed order?

If teams matured along one path, a protocol with a rare practice would also have every more common one. Across all practices that is not the case. Within testing and CI it largely is: test suite, then fuzz tests and CI on pull requests, then invariant tests, then formal verification.

The testing and CI ladder, protocol by protocol

How far up does each protocol get?

Which other practices follow the ladder?

H of each remaining practice against the five rungs. Above the 0.3 mark, a practice tends to appear in step with the ladder; near zero, teams adopt it regardless of how far up they are.

H is unstable for practices found in fewer than five protocols; those bars are faded.

Table view

Where in the lifecycle is the effort?

For each protocol, the share of applicable practices found at each lifecycle stage, averaged across protocols. Stages group the ten dimensions: build → launch → governance → operate → assure.

Post-launch practices are the least visible in public repositories; a low bar there partly reflects what teams keep private.

Table view

Upgradeable vs. immutable

Do protocols that can't be patched after launch invest more before it? Share adopting each practice, by whether the contracts holding user funds can be upgraded.

Table view

Toolchain by launch cohort

What the repositories use today, grouped by when the protocol launched. This is a snapshot of current tooling, not a history of when each practice was adopted.

Foundry-era practices, by cohort

Table view

Explorer

One row per protocol, one column per dimension. Shade is the share of applicable practices found. Select a cell to see each practice, its level, and the evidence, linked to the exact file and line at the scanned commit.

Methods