BE-7025 Spike: Experiment with vibe coding

Description

We want to run a lightweight experiment encouraging engineers to try “vibe

coding” (AI-assisted development) in their day-to-day work and report back on

how useful it actually is.

The goal is not to mandate usage or optimize for speed at all costs, but to

better understand where AI meaningfully helps, where it falls short, and how to

set realistic expectations across the company.

Engineers should start with low-risk work (tech debt, refactors, cleanup, tests,

documentation, or non-feature tasks), and only apply it to feature work if no

suitable candidates exist.

Engineers should capture anecdotal observations such as productivity impact,

code quality, iteration speed, debugging usefulness, failure modes, and any

surprising wins or frustrations. Findings will be shared with the team to help

establish best practices and align on where AI is genuinely additive versus

noise.

Acceptance Criteria

Engineer intentionally uses AI/vibe coding on at least one task

select from parking lot

Task preferably falls under tech debt, refactoring, or non-feature work

Notes captured on what worked, what didn’t, and overall value

Learnings shared with the team (Add to spreadsheet)

Remember the "vibe code gospel" -