A discussion with my programming students about creativity sent me down a side-project rabbit hole. I wanted to make something for fun, then see whether it could become useful outside the joke.

That became DeployAnyway, a collection of five small open-source Node.js tools. My huskies, Dallas and Benji, supplied the inspiration: plenty of curiosity, enthusiasm, and the occasional questionable decision.

The motto is “Software nobody requested, built with questionable priorities, and shipped with absolute confidence!” The software still needs to preserve failures and check actual results.

Start with something you can run

With Node.js 22.13+ or 24, try:

npx @deployanyway/bro-say "The tests passed. Suspicious, but welcome." 

That puts your message into an original terminal character's speech bubble. There are character, mood, theme, wrapping and thought-bubble options, plus a structured API.

The part I hope earns its place in build scripts is the build summary. Install the package:

npm install @deployanyway/bro-say 

Then render facts collected by your script:

import { renderBuildSummary } from '@deployanyway/bro-say'; const summary = renderBuildSummary({ label: 'Release build', exitCode: 0, tests: { passed: 42, failed: 0, skipped: 2 }, durationMs: 1520, commit: 'example-commit', }); console.log(summary); 

Those numbers are illustrative input, not a test runner's output. Your script supplies its real results and preserves the underlying command's exit status. Passing test counts alone do not establish that a build succeeded.

The rest of the collection

Each package works independently:

Tool What it does
error-translator Explains supported Node.js errors and follows Error.cause chains with practical debugging checks. Unknown errors remain unknown.
excuse-js Generates developer excuses for comic relief, or drafts incident updates from known facts. Public incident updates keep the jokes off.
doggo-log Provides log levels, JSON output, async request context and configurable secret redaction, with optional dog commentary.
ship-it-meter Evaluates test reports, coverage and build receipts against release policies, and explains missing evidence or blockers.

There is a runnable workflow combining all five: logging a request, diagnosing a failure and checking release evidence. The docs explain which behavior runs in the browser and which requires Node.

Where Codex helped

I used OpenAI Codex extensively for implementation, tests, declaration types, documentation and release workflows. I supplied the ideas, direction and feedback; I maintain the projects and am responsible for what ships. AI also helped prepare this article.

The project went through repeated revisions. One lesson was that “it builds and the tests pass” is a different milestone from “another developer wants this in their project.” That's the part I'm trying to learn about now.

The libraries do not require an AI service at runtime.

Try one and tell me what you'd change

If you try one, would it earn a place in a real script or codebase? What would make you keep it, and what feels unnecessary? A confusing example or a reason you wouldn't use it is useful feedback too.

Disclosure: these are my projects. They're MIT-licensed, and contributions and bug reports are welcome.