If you’ve ever poured your best ideas into AI on your phone, then sat down at your PC and couldn’t find any of it—you’re not crazy. Two simple truths explain most of the pain:
Truth #1: Cross-device memory isn’t guaranteed.
You’re often using the same service in different sessions. Some apps sync history; many don’t, or they do it spottily. Even if you can see old chats, the AI won’t automatically “know” what’s in a different conversation unless you paste it in. That’s by design (privacy, safety, cost).
Truth #2: There’s no background magic.
Consumer AI doesn’t run offstage or “get back to you later.” No timers. No inbox. No push updates. If an AI says “I’ll follow up,” that’s just a figure of speech. If you don’t prompt again, nothing happens.
So here’s a no-nonsense field guide you can actually use.
The Field Guide
Rule 0 — No background magic.
Assume the AI does nothing after you close the window. If you need a result, get it now in this reply.
Rule 1 — End every session with a recap you can carry.
“Summarize everything we covered today as: Date, Ideas (bullets), Drafts (titles + links/text), Next Steps (numbered). Plain text only.”
Rule 2 — Bridge phone → PC with one habit.
At the end of the bed session, copy the recap into Notes/Keep/one Google Doc (“Bedside Stream”), or email it to yourself (subject: “AI Recap — YYYY-MM-DD”).
At your PC, paste that recap into the chat and say, “Pick up at Next Steps #2–3.”
Rule 3 — Name the project.
Use a code name in every message (“Project: Ferguson Files” or “Project: AI Field Guide”) so you can find it later.
Rule 4 — Ask for deliverables, not promises.
Skip “look into this and get back to me.” Say: “Right now, draft a one-page outline with A/B/C. If you need anything else, list it at the bottom.”
Rule 5 — Force structure.
Tell the AI how to format outputs (bullets, sections, table, CSV). Structured text travels well across devices.
Rule 6 — Expect mistakes.
Ask for a self-check: “Scan this for contradictions or missing steps. Fix what you find.”
Rule 7 — Keep a master doc.
Long projects live in a single running document you and the AI update. Paste the relevant slice when you resume.
Respect & Restrictions: What AI Won’t Do (and Why)
AI has guardrails. That’s by design.
Won’t help with:
- Hate speech, slurs, or demeaning content about groups (race, religion, ethnicity, gender, orientation, etc.).
- Harassment, revenge, doxxing, or “get back at my neighbor” requests.
- Promoting harm or illegal activity.
Why: Safety policies, basic decency, and keeping tools useful for everyone—like range safety rules that protect the whole line.
How to get useful help instead:
- Ask for a firm, respectful message that sets boundaries (no insults).
- Request a neutral summary of the dispute plus options to resolve it.
- Draft a complaint or incident log (dates, times, facts only) you can send to a landlord/HOA, school, or local authority.
- For heated issues, ask for de-escalation language and a cool-off template.
Quick scripts (copy/paste):
- “Write a short, firm note that states my concern and the requested fix, without insults or threats.”
- “Turn this angry draft into a professional complaint I can send to the HOA. Keep dates/times, remove opinions.”
- “Give me a calm one-paragraph reply that declines further argument and offers one next step.”
Why it works this way
- Privacy & safety: Assistants don’t automatically pull info from other chats or devices.
- Costs & limits: Long-term memory is expensive and bounded.
- Design: Most tools are request-response. They answer when you prompt; they don’t run on their own.
What AI is great for today
Plain-English letters, summaries of long bills/emails, organizing photos and notes, checklists, and turning bullet points into clean drafts—fast.
Your turn: What would you put in this field guide for first-time AI users? Drop your rule or horror story in the comments—I’ll add the best ones to a printable checklist.