claudeholic.me — One question. Several hours.
Skip to the intervention
Department of Prompt Health · Bureau of Context Hygiene
Case file 001 · Status: still typing
Unofficial self-assessment
I only opened Claude
to ask one question.
Hours later...
Your tea is cold. Your context is warm. You have used the phrase<br>“one last refinement” three times.
Check your dependency ↓<br>Estimated time: shorter than your current prompt.
Known cause<br>Curiosity
Primary vector<br>“While I have you...”
Prognosis<br>Probably fine. Hydrate.
Not a medical instrument. Barely an instrument.<br>Read the responsible part.
In the past seven days...
01<br>I ask Claude to explain code I wrote yesterday.
02<br>I paste errors before reading them.
03<br>I ask Claude to name variables I have not created yet.
04<br>I say “please” because the machines may remember.
05<br>I ask Claude to write prompts for Claude.
06<br>I optimize prompts more carefully than my sleep schedule.
07<br>I have opinions about context windows.
08<br>I maintain a folder called prompts-final-final.
09<br>I refer to a new chat as “a clean room.”
10<br>I have installed an MCP server to avoid clicking twice.
11<br>I have a preferred model for different kinds of emotional weather.
12<br>I read release notes recreationally and call it research.
Live departmental estimate
Prompt Withdrawal Index
0 / 12<br>No symptoms filed
Recorded symptoms<br>0 of 12
Check the statements above to begin the entirely unofficial review.
Recommended next step
Protect the habit
Keep verifying confident answers, leave private material out,<br>and stop when the useful work is done.
Check consequential claims against a primary source.
Keep secrets and personal data outside the prompt.
Leave the conversation when the question is answered.
Clear the record
Automatic scoring is unavailable, which may be your healthiest<br>browser setting. Count your marks and compare them with the stages<br>below.
Stage 0
Healthy curiosity
You ask a useful question. You verify the answer. You leave.
Clinical note: Astonishing form.
Stage 1
The polite return
You come back tomorrow. It remembers nothing. You forgive it.
Clinical note: Normal tool use with early anthropomorphism.
Stage 3
Prompt engineer
Your question now has a role, constraints, examples, and a small constitution.
Clinical note: The preamble exceeds the task.
Stage 6
Local MCP collector
You connected six tools so you would not need to open a seventh tab.
Clinical note: There are now eleven tabs.
Stage 9
Context window maximalist
You no longer provide context. You provide an archive.
Clinical note: Somewhere, a token counter is crying.
Final stage
Release notes recreationalist
You know what changed before your colleagues know there was a release.
Clinical note: You are reading this website recreationally.
External field report · Exhibit A
“It’s like I have a jet pack. I can’t stop thinking about it.”
Boris Cherny, quoted by Steven Levy in
“AI Agents Plunged the Tech World Into Chaos”<br>, WIRED, May 26, 2026.
01 · Biological maintenance
Drink water.
The model can wait. Your kidneys have a stricter rate limit.
02 · Primary sources
Read the documentation.
AI can orient you. The source tells you what the source actually says.
03 · Manual mode
Write the first draft yourself.
Struggle is not always inefficiency. Sometimes it is the lesson arriving.
04 · Peer network
Ask a human.
They may disagree, remember context, or suggest lunch. All useful features.
05 · Context hygiene
Protect private information.
Remove secrets, personal data, and anything you would not place on a projector.
06 · Verification
Check the confident parts twice.
Fluency is presentation. Evidence is evidence.
07 · Carbon-based rendering
Go outside.
Observe a tree with no loading state. Maintain one hobby without a roadmap.
Initiate emergency analog protocol
Remain calm. Your context window cannot follow you here.
Save your work.
Close one browser tab.
Put both feet on the floor.
Look at something that was not generated.
No account cancellation is required. This is just a break.
Generative AI can be a remarkable companion to curiosity. It can<br>help us cross unfamiliar ground, test an idea, find the shape of a<br>question, or begin when the blank page feels heavier than it should.
But speed is not the same as understanding. A polished answer is not<br>necessarily a true one. When we stop checking, practicing, and<br>making decisions of our own, convenience begins to charge interest.
Craft lives in attention: choosing what matters, noticing what is<br>wrong, and staying with a problem long enough to be changed by it.<br>No tool can outsource that responsibility. Good tools give it more<br>room.
Use AI to expand your agency, not replace it.
Keep learning the fundamentals. Keep private things private. Ask<br>humans what they see. Make something inefficiently, just because<br>your hands and mind deserve the practice. Let an answer surprise<br>you, then verify it.
Curiosity brought us here....