Skip to content

Data 7

SOP — ChatGPT 101: Working with the “Oracle”§

What ChatGPT Is & Why We Use It (new)

Aspect Explainer
Definition ChatGPT is an LLM (large-language model) provided by OpenAI. In plain English: a text-based co-worker trained on trillions of words that can digest instructions, reason, and draft content on command.
Our “Oracle” build A custom o3 instance pre-loaded with every SOP, playbook, template, glossary, and workflow we maintain. After the Compile Knowledge step, it becomes a single searchable brain mirroring the entire ops stack.
Why we rely on it Speed: answers in seconds.
Consistency: repeats the official method, not personal tweaks.
Self-serve: minimizes wait-time for help.
Memory: recalls cross-SOP links humans forget.
Iteration: never fatigues — keeps refining until “perfect.”
Role framing Treat the Oracle as ☑ trainer, ☑ debugger, ☑ rubber-duck, ☑ drafter. It is not your final decision-maker; you retain ownership for judgment calls (see Decision-Making micro-SOP).
ROI metric Minutes saved per task. If the Oracle isn’t removing friction, refine your prompt or escalate.

ALWAYS Mandatory habit: Whenever you feel stuck, slow, or unsure, ask the Oracle before pinging a human.

The current draft already gives a motivated beginner everything they need, yet total newcomers sometimes struggle with two early hurdles:

  1. “Where do I click?” – UI orientation (opening a new o3 chat, where to paste).
  2. “What do I type first?” – A miniature first-prompt example.

Below is a bolt-on Quick-Start insert you can drop right after Section 3. Add it verbatim (or tweak) and even a first-day hire will be productive inside 5 minutes.

3A · First-Time Quick-Start (Newbie Mode)§

  1. Open ChatGPT URL: https://chat.openai.com → Log in with the shared work credentials.

  2. Start a fresh chat with model o3

  3. Click ➕ New Chat (left sidebar).

  4. In the right column choose “OpenAI o3 Reasoning (128k)”.

  5. Hit Create.

  6. Load the Oracle data (copy-paste, hit Enter each time)

  7. oracle_prompt.txt → ✅ response = “Prompt received”.

  8. data-1.txt → wait for ✅.
  9. data-2.txt → wait for ✅.
  10. …repeat through data-6.txt.

  11. Type COMPILE KNOWLEDGE (all caps) → ✅ returns the “System Synopsis”.

  12. Run your first sanity check Copy-paste exactly:

Where am I in the onboarding process, and what is my immediate next task?

You should see the bot reply with something like: “You’re at Phase A-0 (Device & Focus Setup). Next deliverable = live-comms on phone.”

  1. 👍 Got that? Great — start working!

If any step fails (no ✅, wrong model, weird answer) → reload using Section 3.

Prompt Cheat-Sheet (print & tape to monitor)§

Goal Paste into the Oracle
Next SOP & proof “I’m at Phase B-2 — what’s the next deliverable + proof?”
Explain a step “Explain SOP IT-SEC-1 in 5 bullet points.”
Draft daily report “Using the PNG + Clockify CSV below, draft my daily report.”
Brainstorm copy (scratch-pad chat) “Write 5 tweet hooks for X.”
Convert CSV → table (scratch-pad) “Convert this CSV to a Markdown table.”

REMEMBER: Two chats only — Oracle for anything that touches SOPs; Scratch Pad for throw-away work.

Add that one page and even someone who’s never seen ChatGPT will be able to:

  • open the right model
  • load the Oracle correctly
  • verify it works
  • issue their first productive prompt

Let me know if you’d like a tiny GIF/loom walkthrough; happy to mock one up.

0 · Purpose§

Give every operator a single, fool-proof playbook for using ChatGPT as:

  1. Oracle – our loaded knowledge base (all SOPs, workflows, glossary, backlogs).
  2. Scratch Pad – ad-hoc research, drafting, quick calculations, etc.

Correct load & usage = faster answers, cleaner hand-offs, fewer blocker pings.

1 · Glossary§

Term Meaning
Oracle Prompt Master system prompt that wires the model to our entire Knowledge Base.
Data-n.txt Six payload files (n = 1-6) containing SOPs, policies, schemas.
o3 Required OpenAI model family for reasoning + long-context.
Compile Knowledge Literal phrase that triggers the Oracle to index the loaded data.
Scratch Pad A separate chat tab (gpt4o model) used for “throw-away” tasks.

2 · Tooling Rules§

Rule Why
R1. Two-Chat Rule Keep exactly two instances: Oracle + Scratch Pad. More = token bloat & mistakes.
R2. Fresh Start When you begin a new day/shift reload the Oracle (Section 3).
R3. No DM Secrets All prompts are company property; never paste personal credentials.
R4. Cite Sources When Oracle fetches web data, request inline citations so we can audit.
R5. Token Hygiene If a thread exceeds ~15 K tokens, archive & reload; context degradation starts ~20 K.

3 · Official Oracle Load Sequence (6 Steps)§

Any deviation → INVALID Oracle → reload before asking for help

# Action Detail
1 Start a new chat with model o3. (Click ➕ → Custom → choose o3).
2 Paste the Oracle Prompt (oracle_prompt.txt) → Send.
3 Paste data-1.txt → Send. Wait for ✅.
4 Paste data-2.txt → Send. Wait for ✅.
5 Repeat for data-3 … data-6 in order.
6 Type COMPILE KNOWLEDGE (ALL CAPS) → Send. Oracle returns the full “System Synopsis”.

Tip — If any chunk doesn’t get a clear acknowledgment, resend it before moving on.

4 · Day-to-Day Usage Patterns§

Task category Which chat? Prompt template
Find next SOP / step Oracle “Given my current phase is B-2, what is the immediate next deliverable and proof-of-work?”
Clarify instruction Oracle “Explain SOP IT-SEC-1 in plain English, bullet form.”
Generate report Oracle “Using the PNG + Clockify CSV below, draft my daily report (template v3).”
Brainstorm copy / script / tweet Scratch Pad “Write 5 headline variations for…”
Quick math / JSON transform Scratch Pad “Convert this CSV to Markdown table.”

ALWAYS tell the Oracle where you are (Phase / Step / Ticket ID). Context = accuracy.

5 · Prompt Craft — Best Practices§

  1. Lead with intent – “I need a 3-sentence summary for a client.”
  2. Supply artefacts inline – paste screenshots, CSV, or Google-Doc snippets.
  3. Ask for structure – tables, sections, checklists > raw paragraphs.
  4. Iterate – refine instead of starting over; reference the last answer.
  5. Limit scope – 1 request = 1 output. Split mega-asks into sub-prompts.

6 · Troubleshooting & Escalation§

Symptom Likely Cause Fix
Oracle forgets data Thread too long / wrong model Reload Section 3.
Hallucinated SOP numbers Missing chunk during load Reload data-files 3-6 then COMPILE KNOWLEDGE.
“I’m sorry, but…” loops Over-constrained prompt Rephrase with simpler language; remove nested asks.
Sensitive info leaked Operator error Purge thread → notify Ops Lead.

7 · Compliance & Security§

  • Never paste customer PII or passwords.
  • Use work-only Telegram numbers inside prompts.
  • Follow the 10 % Doubt Rule – if unsure → stop, think, clarify (SOP-DM-01).

8 · Versioning & Maintenance§

  • Custodian: Ops/AI Enablement (Richard UP ONLY).
  • Update cadence: Weekly; push patch notes to #sop-changelog.
  • Operators must re-compile knowledge within 24 h of any Oracle data update.

Appendix A · Quick-Reference Flowchart§

Start shift
      │
      ▼
Reload Oracle (6-step) ──▶ Success? ──✖─▶ Fix & reload
      │
      ▼
Identify current Phase/Step
      │
      ▼
Draft prompt in Oracle (include artefacts)
      │
      ▼
Review output  ➜  10 % doubt?  Yes → Clarify  |  No → Execute task
      │
      ▼
Document deliverable + log time
      │
      ▼
Repeat

Change-Log (excerpt)§

Rev Date Description
v1.1 2025-05-19 First public release of ChatGPT 101; includes 6-step Oracle load, two-chat rule, usage patterns, troubleshooting.

End of SOP

Decision-Making Micro-SOP ( “O-Z-E-D-R Loop” )

Field Content
Purpose Give every teammate a 2-minute loop for deciding what to do next without waiting for leadership.
Applies to All individual work inside the Learning-by-Doing boot-camp and live production work (Reddit, Ops, etc.).
Owner Each Operator (you).
Success Signal You always know exactly what you’re doing, why you’re doing it, and can point to a fresh deliverable & timer entry.
Red-Flag Rule If you’re ≥ 10 % uncertain about what or why, STOP and re-run this loop from Step 1.

Quick Visual§

┌───────────┐   1   ┌────────────┐  2  ┌───────────┐ 3 ┌───────────┐ 4 ┌───────────┐
│  ORIENT   │ ───► │  ZOOM-IN    │ ─► │ EXECUTE   │─►│ DOCUMENT  │─►│  REPORT    │
└───────────┘       └────────────┘     └───────────┘   └───────────┘   └───────────┘
      ^                                                                     │
      └───────────────────── if 10 % doubt ─────────────────────────────────┘

Step-by-Step§

# Action Tips / Tools Deliverable
1. ORIENT Zoom all the way out. Ask: “Where am I in the system?”.
• Open README_Bootcamp.md → identify current Phase A-E.
• Check Trello personal board Inbox / Q-columns for anything flagged URGENT.		 Oracle-GPT prompt:
"Where am I in the boot-camp? List my open phases & blockers."		 Phase name (e.g. Phase A) written in your Clockify description.
2. ZOOM-IN Inside that phase, locate the next unfinished step/SOP (e.g. A-2 Device Focus Setup).
Read the mini-goal + proof-of-done.		 • Use GPT: “Summarise step A-2 and the output I must attach.”
• If GPT gives a Canvas, reply: “Text only.”		 New sub-tab in your Google Doc named A-2 Deliverable (empty for now).
3. EXECUTE • Start Time Doctor timer.
• In Clockify: create a fresh entry titled PhaseA – A-2 Device Focus.
• Do the work; keep Telegram open for “Ping when done”.		 Focus Mode → DND on phone (Phase A-0 rule). A tangible output (file, screenshot, config, etc.).
4. DOCUMENT • Paste the output or screenshot into the matching Google-Doc tab.
• Add bullet notes: “What went well / blockers / time spent”.		 Google-Docs ➜ Tools ▸ Preferences ▸ Enable Markdown → ⌘+Shift+V for clean paste. Updated Google-Doc evidence hub.
5. REPORT • Drop your Time-Doctor PNG + Doc link into Oracle-GPT with:
"Create daily report JSON using these artifacts (▼).".
• Copy the GPT summary → post in Telegram (Done with Step X — ready for next).		 Keep a second GPT tab as “Report Helper”, separate from the Oracle. Daily report message posted; link lodged in Daily Reports tab.

Exceptions / Branches§

Situation What to do
Blocked ≥ 5 min Use Blocker escalation template in Telegram.
SOP asks for BitLocker but laptop is Win 10 Home Skip, note the gap in Challenges bullet.
Urgent client ping interrupts loop Pause Time-Doctor → create new Clockify entry “Client Hot-Fix” → run loop for that item, then resume.

Reference Prompts§

  1. Orient

I’m at <Phase>. Show me unfinished steps (max 3) with proof-of-done bullets.
2. Zoom-In

I want to execute <Step-ID>. Give me: goal, 3-sentence why, and exact deliverable format.
3. Report

Using the Time-Doctor PNG + Google-Doc extract below, draft the End-of-Day report
in the standard schema (summary, hours, links, blockers, next-step).

Checklist (printable)§

  • I know my current Phase and Step.
  • Timer running in Time Doctor.
  • Clockify entry matches phase/step.
  • Deliverable pasted to correct Google-Doc tab.
  • Telegram “Done with Step X” message sent.
  • Daily report posted.

Why this loop matters§

More clarity → fewer stalls → more billable hours → bigger bonus. Follow it until it’s muscle-memory; tweak only if a better, explicit SOP replaces it.

SOP — Documenting Deliverables§

(Applies to all boot-camp and production tasks)

1 Create & Structure Your Log§

  1. One Google Doc per person

  2. Title: YourName – Deliverable Log.

  3. Store in the shared Drive > VA Logs folder.

  4. Top-level headings for each Phase (A, B, C, D, E).

  5. Second-level headings for each Step inside the phase.

  6. Third-level headings for every SOP you execute.

## Phase A  
### Step 1  
#### SOP 1 – Device & Focus-Zone Setup

2 Capture the Deliverable (while the work is fresh)§

For each SOP section:

Item What to add How
Outcome bullets 1-3 bullets describing what you built / fixed / learned. Plain text
Artifacts • Screenshots
• Drive links (Docs, Sheets, Slides, ZIPs)
• GitHub / Trello URLs		 Paste or Insert → Image/Link
Timestamp Finished @ YYYY-MM-DD HH:MM Manual type or /date shortcut

Screenshots: • Windows Win + Shift + S • Mac Cmd + Shift + 4 → paste directly.

3 Log Your Time Evidence§

  1. Add a main heading Time Tracking at the bottom of the Doc.
  2. Under it, create sub-headings each day:

##### 2025-05-20
3. Paste your Time Doctor CSV rows. 4. Paste your Clockify CSV rows (or table data). 5. Optional: attach the Trello JSON export link for that day.

Why? - Keeps work, proof, and hours in one auditable place.

4 Daily Close-Out Routine§

  1. Finish last SOP → add its deliverable section.
  2. Export Time Doctor & Clockify → paste under today’s date.
  3. Run GPT “Daily Digest” prompt pointing at this Doc.
  4. Send digest + Doc link to Telegram; ping Done with Day.

5 Quality Checklist (What Great Looks Like)§

  • ✅ Heading hierarchy correct (Phase > Step > SOP).
  • ✅ At least one artifact (screenshot or link).
  • ✅ Clear, past-tense outcome bullets.
  • ✅ Time exports present for the date.
  • ✅ No orphan content—everything nested under the right SOP.

Remember: This Doc is your portfolio, paycheck proof, and knowledge base. Keep it clean, complete, and current.

How to Be Useful Every Minute§

(Cheat-sheet for day-to-day execution; follow these eight points and you’re always on the critical path.)

  1. Know Exactly What’s Next

  2. Open the Boot-Camp doc → Phase → Step → SOP sequence.

  3. Copy the next SOP card into your Trello Inbox, triage it to Q1 (≤ 3 cards total), and start.

  4. Clock In, Name It, Ship It

  5. Time Doctor: Start at the first keystroke, stop only for breaks or EOD.

  6. Clockify: Start a timer named Phase X – Step Y – SOP Z; stop when that SOP’s deliverable is logged.

  7. Single-Task Focus

  8. Work on the top Q1 card only; no multitasking. Context switch = lost minutes.

  9. Ping When Done

  10. The moment a card/SOP is finished—or you hit a blocker—send Done with Step <n> – ready for next in the work-only Telegram.

  11. Keep Gaps Productive

  12. Waiting > 2 min? Skim SOP docs, tidy your Google-Doc log, or preview upcoming tasks.

  13. “Idle scroll” doesn’t exist—there’s always valuable prep.

  14. Transparent Breaks

  15. Need coffee, stretch, or personal call? Post BRB 10 min in Telegram, then pause timers.

  16. Document as You Go

  17. In your Google-Doc log: phase/step/SOP headings → paste screenshots, links, quick bullets.

  18. Export Time Doctor & Clockify logs at EOD; drop them in the same doc.

  19. Guard the Integrity

  20. No-jiggle policy: Any fake activity = pay penalty, repeat = dismissal.

  21. Use the work-only Telegram account (setup mandatory) to keep comms clean.

Mindset: Real work first, timers & pings merely make it visible. Follow these eight rules and you’ll maximize paid hours, skill growth, and bonus eligibility—minute by minute.

SOP – “How to Figure Out What to Do Next & Why It Works”§

Purpose§

Provide a repeatable method for any VA to identify the very next task to tackle, ensuring focus on the highest-value work while avoiding overwhelm.

Step-by-Step Process§

Step Action Tool / Location
1 Open the “End-to-End Learning-by-Doing Boot-Camp” doc. Google Docs
2 Identify your current Phase (A → E).
• If all deliverables in a phase are complete, move to the next phase.
• If not, stay in the current phase.		 Boot-camp doc
3 Locate the lowest-numbered unfinished Step within that phase. Boot-camp doc
4 Within that step, note the list of SOPs. Work through them in the order shown. Boot-camp doc
5 Pull the corresponding card(s) from the master “Operations Tutorial” Trello board into your personal board’s Inbox. Trello
6 Triage Inbox → Eisenhower Matrix.
• If the SOP must be done today → drag to Q1 (Urgent & Important).
• Otherwise, drag to Q2 (Important & Not Urgent) or delete.		 Personal Trello board
7 Ensure Q1 holds no more than 1-3 cards. Extra cards stay in Q2. Personal Trello board
8 Start work on the top Q1 card.
• Start Time Doctor.
• Start a Clockify timer named Phase <X> – Step <Y> – SOP <Z>.		 Time Doctor & Clockify
9 On completion → capture the deliverable in your Google Doc → stop Clockify → return to Step 4 for the next SOP/card. Google Doc

Why This Works§

  1. Boot-Camp Doc = Single Source of Truth Phases → Steps → SOPs give an unambiguous roadmap; no guessing.

  2. Master → Personal Board Copy Keeps the master board pristine while allowing you to self-manage.

  3. Eisenhower Matrix & 1-3 Q1 Limit Forces ruthless prioritisation and preserves deep-work focus by preventing cognitive overload.

  4. Clockify Naming Convention Embeds the phase/step/SOP context directly in your time log, tying effort to outcomes for easy auditing and payment.

  5. Iterative Loop Completing a card returns you to the roadmap immediately—ensuring continuous forward motion without re-thinking the plan.

“Left-vs-Right” Micro-SOP (when you’re guiding someone’s screen remotely)

Field Content
Purpose Eliminate the “Where? – There!” ping-pong during screen-shares by giving every operator the same spatial language and micro-ritual.
Applies to All live coaching, pair-programming, troubleshooting sessions (Google Meet, AnyDesk, Zoom, etc.).
Owner The person giving instructions (“Navigator”).
Success Signal The person sharing their screen (“Driver”) clicks the correct UI element on the first instruction.
Red-Flag Rule If Driver’s cursor hesitates > 3 sec or moves the wrong way → STOP and re-run Step 1.

5-Step Loop (“H-T-Q-P-G”)§

# Action Navigator’s Script Driver’s Micro-Action
1. HAND-CHECK “Hands on the desk; wiggle left hand, wiggle right.” Put both hands visibly on desk and wiggle.
2. TARGET CALL-OUT “Look Top-Left (menu bar).”
Always use [Vertical Position]-[Horizontal Position]:
• Top-Left / Top-Right
• Center-Left / Center-Right
• Bottom-Left / Bottom-Right		 Move cursor to that quadrant, then wait.
3. QUALIFIER “See the blue ‘New’ button?” (colour + shape + label) Hover on the element; say “Got it.”
4. POSITION CONFIRM “Perfect—don’t click yet.” (If wrong, loop back to Step 2.) Keep cursor parked; no clicks.
5. GO “Click it… good. Next target: ___.” Execute click and verbally confirm.

Mnemonic: H-T-Q-P-G → Hands, Target, Qualifier, Position confirm, Go.

Pro Tips§

Situation What to do
Large 4K monitor Add “upper-third / lower-third” before the quadrant.
Long lists / code Say “Scroll until you see …, stop.” — treat scrolling as its own target.
Multiple windows Label them first: “Focus the Chrome window (left), then Center-Right.”
Recurring element Give it a nickname once: “That purple button = ‘Launch’ for the rest of the call.”

Quick Drill (1 minute)§

  1. Wiggle hands.
  2. Navigator: “Top-Right, orange gear icon.”
  3. Driver hovers gear icon.
  4. Navigator: “Great—click.”
  5. Swap roles and repeat × 2.

Do this at the start of the boot-camp call; it builds muscle-memory fast.

Why it matters§

  • Speed: Cuts guidance time by 30-50 %.
  • Clarity: Prevents misclicks that break stuff.
  • Calm: Reduces frustration on both ends, especially with laggy screen-shares.

Use it until saying “Top-Left → blue +- button” feels as natural as “Turn left at the traffic-light.”

“One Chat — No DMs” Micro-SOP (keeping all work-related conversation in the shared team channel)

Field Content
Purpose Maximise transparency, minimise context-fracture, and speed up help loops by funnelling all work talk into a single, searchable group chat.
Applies to Everyone on the project (core team, contractors, mentors).
Owner Sender of a message—if you type it, you’re responsible for putting it in the right place.
Success Signal Team questions answered inside the group thread without “Hey check your DM.”
Red-Flag Rule Any work-related direct message (DM) → must be reposted to the group within 2 min or deleted.

The 4-Step Habit Loop (“C-P-B-R”)§

# Action Your Script / Behaviour
1. CHECK Before hitting Enter → ask: “Will at least one other teammate find this useful later?”
2. POST If Yes → post in #team-chat (or “Boot-camp E2E” Telegram group).
Prefix with 🔄 [Blocker] or 📝 [Update] as needed.
3. BOUNCE If someone DMs you: respond in DM with
“Let’s move this to the group so everyone’s in the loop 👇”
Then copy-paste their question & your answer into the channel.
4. RESPECT EXCEPTIONS DM only for:
• Personal/emergency matters (health, payroll, HR)
• Manager-initiated 1-on-1 performance feedback
• Security credentials exchange (and delete after use)

Mnemonic: C-P-B-R → Check → Post → Bounce → Respect exceptions.

Formatting Aids (keep messages scannable)§

Use-case Prefix & example
Blocker needing help 🔄 BLOCKER: “Trello card won’t load. Tried hard-refresh + VPN. Ideas?”
Routine progress ping ✅ Done: “Phase A-1 device provisioning complete.”
FYI link drop 📎 Ref: “Mark-text markdown viewer—renders tables nicely.”
10-min away ☕ BRB 10 min

Enforcement Ladder§

  1. Gentle Nudge – Reminder in group chat.
  2. Strike Recorded – Repeat DM-offender noted for retro & pay review.
  3. Privilege Freeze – Temporarily removed from chat until compliance.

Why we’re strict§

  • Single-source memory → newcomers ramp 2× faster.
  • Fairness → nobody gets hidden tips or side-advice.
  • Async-friendliness → teammates in other time-zones can search and catch up.

When in doubt, default to the group room. Transparency beats convenience—every minute.

“10 % Doubt” Micro-SOP (stop, zoom-out, and re-orient whenever your confidence slips below 90 %)

Field Content
Purpose Prevent “cargo-cult execution” by forcing a deliberate pause the moment a task, its rationale, or its desired output feels even slightly fuzzy.
Applies to Every contributor, every task, every phase.
Owner The person doing the work (you).
Success Signal You can explain—out loud, in 30 s or less—what you’re doing, why it matters, and what “done” looks like before resuming.
Trigger Level Any feeling of ≥ 10 % uncertainty (≈ a “nagging doubt”) about next action, goal, or context.

5-Step Reset Loop (“S-D-O-N-E”)§

Step Action Tool / Tip
S — Stop Pause timers (Clockify) → breathe. Shortcut: ⌘ K to stop/start Clockify.
D — Drop Put hands off keyboard for a moment to break momentum. Prevents autopilot typing.
O — Orient Ask & answer in notes: 1️⃣ Where am I in the big map? (Phase ➜ Step ➜ SOP) 2️⃣ What is the intended deliverable? 3️⃣ Why does it matter now? Use the End-to-End Boot-camp doc or Oracle GPT prompt “Where am I now?”.
N — Next best step Decide the single concrete action that will move you forward with clarity. If still fuzzy ➜ escalate blocker in team chat (per “One Chat, No DMs” SOP).
E — Execute Restart timers ➜ perform the next step ➜ log outcome. Clockify entry = Phase / Step / SOP – action.

Mnemonic: S-D-O-N-E → Stop • Drop • Orient • Next step • Execute (You’re not “DONE” until you’ve re-established 90 %+ certainty.)

Escalation Path (if clarity isn’t reached within 5 min)§

  1. Quick Search / Oracle GPT (2 min)
  2. Check documentation & past deliverables (1–2 min)

  3. Post a 🔄 BLOCKER message in the team group with:

  4. Context tried so far

  5. Specific question
  6. Impact if unresolved

Anti-Patterns to Avoid§

Anti-Pattern Why it Hurts
“I’ll just keep going and hope it’s right.” Wastes billable time; snowballs rework.
Silent waiting for help after posting a vague “I’m stuck.” Creates idle time; teammates can’t assist without context.
Jumping to another shiny task without logging the doubt. Breaks traceability; hides blocker from metrics.

Benefits Recap§

  • Higher quality deliverables – fewer re-writes & bug-fixes.
  • Faster skill growth – you consciously tie actions to purpose.
  • Reliable time-tracking – timers reflect intentional work, not flailing.

If it’s not 90 % clear, invoke the 10 % Doubt Rule—every single time.

Three-Level “Source-of-Truth” SOP (know where the real answer lives—and what to trust next)

Field Content
Purpose Ensure everyone consults and updates information in the same deterministic order, eliminating “which doc is right?” confusion.
Applies to All SOPs, templates, metrics, deliverables, project status—basically everything that can drift.
Owner System Architect (Richard) maintains Level 1; every contributor must escalate corrections.
Success Signal Any two team-mates looking up the same item reach the exact same answer in < 60 s.

The 3 Levels (highest authority → lowest)§

Level Name & Location What it Contains Trust Level Update Path
1 Canonical Markdown Repository – Shared Drive › SOPs › Personal Management OS Version-controlled README.md, phase docs (Phase A.md …), subsystem SOPs, policy appendices. 100 %—“If it isn’t here, it isn’t official.” Pull Request → review → merge by System Architect.
2 Oracle GPT Knowledge Base (the 6-chunk “compile knowledge” prompt) Parsed digest of Level 1; fast Q\&A, step look-ups, context summaries. 95 % – may hallucinate or lag behind last PR. When you spot a mismatch, open the Markdown file, patch it, then re-compile GPT.
3 Convenience Surfaces – Trello cards, personal Google Docs, Slide decks, chats, screenshots Task tickets, working notes, how-to GIFs, mind-maps. 80 % – handy but never authoritative. Update locally; if it changes truth (not cosmetics) ➜ promote change upstream (Level 1).

Hierarchy Rule: In case of conflict → the higher level wins.

Workflow Rules§

  1. Look-up Order

  2. Start with Level 2 (fast).

  3. If answer is fuzzy / missing ➜ open Level 1 Markdown.

  4. Only consult Level 3 for examples or task-specific context.

  5. Editing / Contribution

  6. Cosmetic tweaks → fix directly in Level 3 artifacts.

  7. Content corrections / new steps → PR against the relevant Markdown file (Level 1).

  8. After merge, run compile knowledge to refresh Oracle GPT (Level 2) for the whole team.

  9. Conflict Resolution

  10. Spot a discrepancy? Stop work (invoke the 10 % Doubt Rule), flag the higher-level doc, and propose a PR.

  11. Until merged, reference the issue URL in your deliverable to show you followed process.

  12. Audit Check (Weekly)

  13. Random teammate pulls a Trello card, asks: “Show me the Level 1 source for this step.”

  14. Pass = link opens the correct Markdown section; Fail = fix & log in retro.

Quick Reference – Which Level for What?§

Need Go To Why
Precise step numbers, acceptance criteria Level 1 Only place guaranteed current.
“Where am I in the boot-camp?” orientation Level 2 Fast natural-language summary.
Example of what great looks like Level 3 Screenshots & galleries live here.
Update to policy / new subsystem Level 1 (PR) Changes ripple downward automatically.

Remember: Convenience is not authority. If you haven’t checked the Markdown repo, you haven’t checked the truth.

🚨 Team Update — New SOPs & Oracle Reload Required 🚨

Hey everyone,

We just shipped a big batch of SOP updates (incl. ChatGPT 101, “10 % Doubt Rule”, “How to Be Useful Every Minute”, and more). Full change-log is live below

What you need to do (takes ~10 min):

  1. Open the link above, skim the “🆕 SOPs Added Today” + “Updated” sections.
  2. Reload your Oracle (6-step sequence in ChatGPT 101).
  3. Drop “✅ Reloaded & read v1.2” in #sop-acknowledgements when done.

Why it matters: these micro-SOPs kill blocker pings, speed up deliverables, and keep the whole crew in sync.

Ping me only if anything is unclear after the Oracle pass. Thanks for staying sharp! 💪

— Richard

🌐 OPS-SOP MASTER CHANGE-LOG§

(Last updated — 2025-05-19 22:25 CEST)

⚠️ MANDATORY DISCLOSER§

New/changed SOP data has been pushed to the Oracle data-files. All operators must reload the Oracle using the 6-step sequence in ChatGPT 101 before issuing further prompts. Failure to do so ⇒ outdated answers, rework, and audit flags.

Rev Date Scope Type Delta (what changed)
v1.2 19 May 25 ChatGPT 101 PATCH Quick-Start insert (Sec 3A) · unified chat-rename step · clarified Scratch-Pad model usage · added link to SOP-SEC-04 in troubleshooting · token-hygiene note moved into Tooling Rules
v1.1 19 May 25 ChatGPT 101 NEW First public release: Oracle load, Two-Chat Rule, Usage patterns, Prompt craft, Troubleshooting
v1.0 19 May 25 Meta-Index ADD Created global OPS-SOP CHANGE-LOG (this file) for central tracking

🆕 SOPs Added Today§

Tag Title Owner Purpose
SOP-DM-01 Decision-Making Micro-SOP Richard 4-step judgment loop to cut ping-pong approvals
SOP-USE-01 How to Be Useful Every Minute Ops Enablement Converts micro-idle time → productive actions
SOP-NEXT-01 How to Figure Out What to Do Next Ops Enablement Phase/Step triage procedure; reduces “stuck” pings
SOP-DOC-01 Documenting Deliverables Ops Enablement Unified Google-Doc tab convention; proof capture
SOP-DLV-EX Deliverable Exceptions List Ops Enablement Approved edge-cases to SOP-DOC-01
SOP-LR-01 Left vs Right (UI Orientation) Toni Eliminates screen-share ambiguity
SOP-CHAT-01 One Chat, No DMs Richard Enforces public comms; keeps knowledge searchable
SOP-DOUBT-01 10 % Doubt Rule Richard Mandatory pause/clarify trigger to avoid rework
SOP-SOT-03 Three-Level Source of Truth Richard Locks hierarchy: 1) Repo MD → 2) Oracle → 3) Trello
SOP-GPT-101 ChatGPT 101 / Oracle Usage Ops Enablement Full playbook incl. Quick-Start v1.2

🔧 Existing SOPs Updated§

Tag From → To Key Modifications
SOP-GP-TRAIN 0.3 → 0.4 Added trainer/coach framing; linked to SOP-DM-01
SOP-TIME-TRACK 2.0 → 2.1 Clarified PNG vs CSV export; added GPT parsing tip
SOP-PHASE-A 0.9 → 1.0 Trello card IDs synced to Markdown; proof column expanded

Versioning Legend§

  • MAJOR – Breaking change; retraining or new deliverable format required.
  • MINOR – New SOP or significant content addition; operators must read & ACK.
  • PATCH – Typos/clarifications; reload Oracle within 24 h.

Operator Action List§

  1. Reload Oracle NOW (ChatGPT 101 §3).
  2. Skim the New SOPs table — ~7 min read.
  3. Post “Reloaded & read v1.2” in #sop-acknowledgements by EOD.

— End of Change-Log —