Context is everything
Brief AI like a smart new hire.
AI Hub
Golden Rules
The principles I come back to every time I build with AI. Not theory — the hard-won shorthand from actually shipping tools.
If you've done it manually, you can describe it
The bottleneck is the description, not the tool.
Garbage in, garbage out
Clean your data before you upload.
AI is a first draft, not a final answer
You're still accountable.
Wrong direction? Start a new chat
Don’t keep adding "actually..."
Doing it twice? It should be a tool
Think prevention, not just execution.
How I decide whether something should become a tool
Not every annoying task deserves software.
Before I build, I ask a few questions.
- 01
Am I doing this more than once?
If it keeps coming back, pay attention. The second time is usually the signal.
- 02
Can I explain the process clearly?
If I cannot explain it to a new employee, I probably cannot explain it to AI either.
- 03
Is this rules-based or judgment-based?
Rules are good candidates for tools. Judgment needs a human in the loop.
- 04
Does this live in one person's head?
If the process disappears when that person is out, the problem is bigger than efficiency.
- 05
What does done look like?
Define the finish line before building. Otherwise every useful tool becomes a playground for scope creep.
Doing it twice is the signal. Defining done is the guardrail.