How to vibe with AI
"AI writes the draft. We define the value."
Safety isn't "slow down." It's what makes full throttle possible.
Before we go fast, we set clear boundaries and stick to them:
Security & Privacy, Reliability, Operability, and Cost & Latency limits (SLOs).
We don't trade safety for speed.
We build a codebase that's easy for both people and AI to work with.
The code stays consistent. We save our creativity for decisions: what matters, what we trade off, and how things can fail.
We don't spend it on clever implementation tricks.
We care first and foremost about delivering value.
Whether the code comes from a person or an AI doesn't matter.
We're willing to fix it, discard it, or accept it, as long as it leads to a better result.
The question is always the same: is this better for users and the system?
Do not fear.
Trust yourself.
Stay curious.
Enjoy learning how things work.
Enjoy making them better.
We define what problem we are solving and why it matters.
We define the what and the why, and let AI handle the how.
We curate project intent, working assumptions, terminology, and the reasons behind past decisions.
We keep context current, relevant, and actionable for both humans and AI.
We design clear boundaries and durable structure so AI can move fast without turning speed into debt.
We align people and AI around a single direction and style.
We take responsibility for the work until it ships.
- State intent explicitly
We share why we're making a change before focusing on what we're changing. - Demand evidence
We trust what we can measure and verify, not what merely sounds confident.
We always ask: is this actually correct? - Declare constraints upfront
Performance, cost, security, and operability are not "later." They are decisions we make now.
See also:
- Human Guardrails for Vibe - What not to delegate to AI