aaron

Quick win: use a short, tweakable prompt template — not a one-size-fits-all paste — and you’ll cut wasted images and speed from idea to deploy.

Problem: people overproduce AI images that don’t fit UVs, scale, or engine constraints. That costs time, money, and motivation. Fixing this is a tiny process change: a repeatable prompt template, early device checks, and a strict asset budget.

Why it matters: in AR/VR performance and consistency matter more than photorealism. One optimized prop that runs well on a phone is worth ten beautiful models that drop frames or show seams.

Lesson from the field: I iterate with a 3-part prompt — concept views, material instructions, and texture map specs. That forces outputs suitable for UVs and baking and reduces rework.

What you’ll need

• Blender (free) for modeling and UVs.
• Unity Personal or Unreal Engine for preview and AR export.
• AI image tool (Stable Diffusion or cloud generator) able to output high-res and seamless textures.
• Basic phone for on-device testing.

Step-by-step (do this once per asset)

1. Define constraints: target (AR phone/VR headset), max tris (1–5k for props), and scale in meters.
2. Use the prompt template below to generate 4–6 concept views and a seamless texture map.
3. Block out low-poly model in Blender, match scale, and test against a human proxy.
4. UV unwrap with consistent texel density; bake normal/AO maps; apply AI-generated textures.
5. Optimize: remove hidden faces, merge where possible, and make 2 LODs.
6. Export glTF/FBX, import to engine, test on device; iterate until stable 30+ FPS on target phone.

Copy-paste prompt template (tweak placeholders)

“Generate [output-type: concept images / seamless albedo texture / normal map] for a [object type, e.g., mid-century wooden lounge chair] in [style: realistic/stylized], provide 4 views (front, side, top, close-up of [feature]), resolution 4096×4096, include seamless UV-ready wood grain texture map (tileable), natural lighting, neutral HDRI reflections, color profile sRGB.”

Prompt variants (copy-paste)

• Concept: “Generate high-res concept images for a stylized ceramic vase, 4 views (front, side, top, close-up lip), soft studio lighting, consistent proportions.”
• Texture: “Seamless albedo texture map for worn walnut wood, 4096×4096, tileable, visible grain direction, subtle wear at edges.”
• Maps pack: “Produce albedo, normal map, and roughness map for aged leather cushion, 4096px, aligned for UV baking, sRGB albedo, non-color for normal and roughness.”

Metrics to track

• Time to first usable asset (goal: <7 days).
• FPS on target device (goal: 30+ steady).
• Triangle count vs. budget (stay within 10% of target).
• Texture memory per asset (MB).

Common mistakes & fixes

• Wrong scale — Fix: set Blender units to meters and use a 1.8m human proxy.
• Non-tileable textures — Fix: request “seamless/tileable” and test in a checker UV.
• Too many images — Fix: limit to 3–6 images and iterate on the template, not quantity.

1-week action plan

1. Day 1: Define asset + constraints; run 2 prompt variants, pick best image set.
2. Days 2–3: Block out low-poly model in Blender, set scale, basic UVs.
3. Day 4: Generate seamless textures with tuned prompt; apply and bake normals/AO.
4. Day 5: Optimize, create LODs, export to glTF.
5. Day 6: Import to Unity/Unreal, place in scene, test on phone.
6. Day 7: Fix issues found, measure FPS and memory, finalize asset.

Your move.

Start simple: run an AI-assisted accessibility audit that surfaces real fixes you can ship this week.

Problem: most sites pass generic scans but fail real-world accessibility — keyboard focus issues, missing labels, poor color contrast. AI can analyze automated results plus your page HTML/screenshots and return prioritized, developer-ready fixes. But you need a repeatable process and acceptance criteria.

Why it matters: accessibility reduces legal risk, improves conversion, and opens your product to more customers. Fixes are measurable: better Lighthouse/axe scores, fewer support tickets, and higher engagement from assistive-tech users.

Lesson: use AI to translate diagnostics into tasks. Automated tools find issues quickly; AI turns those into step-by-step fixes, code examples, and estimated dev time so a non-technical PM can prioritize work.

1. What you’ll need
   • Access to an automated scanner (Lighthouse/axe output or HTML report).
   • 1–3 representative page URLs or HTML snippets and screenshots.
   • AI access (ChatGPT-style) to convert reports into fixes.
   • Basic QA: keyboard test and a screen reader (or checklist for a contractor).
2. How to do it — step-by-step
   1. Run an automated scan across key pages; export results (JSON/HTML).
   2. Gather 3 screenshots and the HTML of the top-converting page(s).
   3. Feed the scan + HTML + screenshots to the AI prompt below to get prioritized fixes and code snippets.
   4. Create tickets with acceptance criteria and estimated dev time from the AI output.
   5. Implement quick wins (labels, alt text, tabindex order, contrast) first; re-scan and repeat.
   6. Do manual keyboard and screen-reader checks for the top 10 flows.

Copy-paste AI prompt (use with ChatGPT or your AI):

Here is an automated accessibility scan result (paste JSON) and the HTML for this page (paste HTML). Provide a prioritized list of fixes, each with: issue title, severity (High/Medium/Low), exact location (CSS selector or HTML snippet), a short explanation in plain English, step-by-step implementation instructions, one copyable code fix (HTML/CSS/JS) where possible, estimated developer time (low/med/high in hours), and an acceptance test to confirm the fix. Limit to 12 items and mark the top 3 quick wins.

Metrics to track

• Accessibility issues found (total and by severity).
• Pages remediated / percentage complete.
• Developer hours spent vs estimated.
• Lighthouse/axe score improvement.
• Support tickets related to accessibility.

Common mistakes & fixes

• Relying only on automated tools — fix: add manual keyboard & screen-reader tests.
• Vague tickets — fix: include selector + acceptance test + code snippet from AI.
• Trying to change everything at once — fix: prioritize quick wins and top user flows.

1-week action plan

1. Day 1: Inventory top 10 pages; run automated scans and capture screenshots/HTML.
2. Day 2: Run the AI prompt on each page; collect prioritized fixes and code snippets.
3. Day 3–4: Implement 5 quick wins (labels, alt text, keyboard focus, contrast, form errors).
4. Day 5: Manual QA on fixed pages; update tickets for remaining items with estimates.
5. Day 6–7: Re-run scans, report metrics, and plan next sprint based on remaining high-severity items.

Your move.

Aaron

Good call on category + UTM tags. Now turn your listings and posts into a repeatable revenue engine that you can measure every two weeks.

The gap: Most profiles stop at NAP, category and a few posts. Missed: full services/products, offer cadence, message testing, and call tracking. That’s where AI earns its keep.

Why it matters: Google Business Profile (GBP) is your highest-intent local touchpoint. Tight, testable posts and complete listings convert map views into calls, directions and bookings in weeks—not quarters.

What to set up (once)

• GBP access, master NAP, service list, 10 recent photos.
• One spreadsheet with tabs: Listings, Posts, Reviews, Metrics.
• AI assistant for drafting content and comparisons.
• Optional: a call-tracking number (set as primary on GBP; set your main number as additional to protect NAP consistency across directories).

Playbook: precision + scale

1. Complete your GBP inventory
   • Add 8–12 Services written in customer language. Add 6–10 Products (or menu items) with short benefits and prices/ranges.
   • Use the “From the business” field to add proof points: years in area, nearby landmarks served, guarantees, response times.
   • Enable attributes (accessibility, payment types, ownership) and add Booking/Appointment URL if relevant.
   • Double-check pin placement and service areas. Small map shifts can affect visibility.
2. Run a simple A/B post system (2×2)
   • Two themes: Offer vs Local Tip. Two CTAs: Call vs Get Directions.
   • Format every post with your “3-3-1” rule (you outlined). Keep 80–120 words. Attach a relevant photo.
   • Publish 2 posts/week. Add UTM tags to each post link so you can see sessions and calls from posts in analytics.
3. Build a “language bank” from reviews
   • Paste 10–20 customer reviews into AI. Extract top phrases, recurring benefits, and anxieties.
   • Mirror those exact phrases in Services, Posts, and Q&A. This increases message-market fit and can trigger local “justifications.”
4. Offer cadence with urgency
   • Use Offer and Event post types monthly. Include a start/end date and a clear redemption step.
   • Tie at least one monthly offer to a local event/season for stronger relevance.
5. Visual trust, not stock
   • Monthly set: exterior sign + known landmark, team at work, before/after, product in use, customer-safe testimonial shot.
   • Name files descriptively; in the post caption, name the neighbourhood and service.
6. Calls you can actually attribute
   • If you use call tracking: set the tracking number as primary in GBP and your main number as additional. Keep your main number on the website and all other directories to preserve NAP consistency.
   • Log calls by source (GBP profile vs GBP posts via UTMs) in your sheet.
7. Tighten with a competitor gap check
   • Scan the top 3 local competitors. Note their primary category, secondary categories, services, and post angles.
   • Close the gaps: missing services, weaker CTA, fewer local cues, no offers.

Copy-paste AI prompts (use as-is)

• Inventory Builder: “You are a local SEO assistant. For a [business type] in [city/neighbourhoods], create: a) one 240–300 character ‘From the business’ paragraph with 2 local landmarks, b) a list of 10 Services in customer language with 1-line benefits, c) 6 Products (or packages) with short blurbs and price ranges, d) a checklist of relevant GBP attributes and why each matters.”
• Post A/B Test Kit: “Generate 8 Google Business Profile posts for a [business type] in [city]. Use 80–120 words, include 3 benefits, 3 local cues, and 1 CTA. Create a 2×2 matrix: Theme (Offer vs Local Tip) x CTA (Call vs Get Directions). Provide UTM-ready link text and a one-sentence hypothesis for each variant.”
• Review Language Miner: “Analyze these reviews: [paste 10–20]. Output: a) top 10 customer phrases, b) 5 objections and how to address them, c) 5 Q&A entries, d) a 120-word post using the most common phrases, with a clear CTA.”

What to track every 14 days

• Discovery vs Direct searches (GBP Insights).
• Calls, Direction requests, Website clicks (absolute numbers and percentage change).
• Sessions from UTM campaign “gbp_posts_[month]”.
• Review volume, average rating, and response time.
• Category/keyword positions for 2–3 target phrases (simple grid or radius check).

Targets: Aim for steady 10–20% lifts in post-driven sessions month-over-month, +1–2 new reviews weekly, sub-48h reply time, and at least one winning post variant with 2x higher clicks vs the baseline.

Frequent mistakes & quick fixes

• Thin services list — Fix: add 8–12 services in customer words; match them in posts.
• Generic imagery — Fix: use real photos with visible signage or landmarks.
• Untrackable posts — Fix: every post link gets a UTM tag; log results.
• One-and-done copy — Fix: run the 2×2 test, keep the winner, iterate the loser.
• Overstuffed keywords — Fix: 1–2 local phrases max; mirror review language.

7-day execution plan

1. Day 1: Run Inventory Builder; update Services, Products, and “From the business.”
2. Day 2: Confirm categories/attributes; set Appointment URL; verify pin/service area.
3. Day 3: Upload 5 fresh photos; draft 8 posts via Post A/B Test Kit.
4. Day 4: Publish 2 posts (Offer+Call, Local Tip+Directions) with UTM tags.
5. Day 5: Mine reviews; add 5 Q&A entries; update one post to include top phrases.
6. Day 6: If using call tracking, set numbers as noted; test and log sources.
7. Day 7: Snapshot competitors; close one gap (missing service or stronger CTA). Log baseline metrics.

Expectation: You should see early movement in calls/directions within 2–4 weeks if you publish consistently, keep offers current, and iterate from the data.

Your move.

— Aaron

Strong call on p80 + spread — that’s the reliability lens most people miss. Let’s turn it into a weekly errand system that reduces trips, locks in predictable windows, and feeds your calendar automatically.

Outcome in plain terms: fewer drives, fewer surprises, more on-time arrivals. You’ll cluster stops, pick micro-windows (20–30 minutes), and buffer smartly based on confidence.

Do / Don’t checklist

• Do use p80 + low spread for window selection; widen windows only if sample size is small.
• Do log dwell times (e.g., 8 min pharmacy pickup, 20 min grocery) and store hours; they make or break the plan.
• Do bundle nearby stops (within 2–3 miles) into one errand run; sequence them to avoid backtracking.
• Do use a morning-of live check as the final gate before you leave.
• Don’t optimize on averages only or trust 2–3 trips; require n≥5 per 20–30 min bucket, else fall back to 60–90 min windows.
• Don’t ignore edges of rush periods; depart at the start of a stable plateau, not midway into a rising spike.
• Don’t mix different routes without labels; keep origin–destination names consistent.

Insider upgrade: micro-windows + bundling

• Micro-windows: slice each hour into 20–30 minute buckets (e.g., 09:00–09:20, 09:20–09:40, 09:40–10:00). Pick buckets with the lowest p80 and low spread. If sample size is thin, merge adjacent buckets or roll up to the full hour.
• Bundling: cluster stops you can do in 60–90 minutes. Choose a departure micro-window that keeps all legs inside low-traffic zones given your dwell times.
• Reliability buffers by confidence: High = add 10% or 5 mins (whichever larger); Medium = 20% or 7 mins; Low = 30% or 10 mins.

What you’ll need

• Maps/Waze on your phone
• Spreadsheet (Excel/Google Sheets)
• AI assistant
• 7–14 days of quick logs (14+ ideal)
• List of stops with opening hours and typical dwell minutes

Step-by-step (build once, reuse weekly)

1. Log routes: keep columns: Date | Day | Time (HH:MM) | Origin | Destination | Travel_minutes | Weather | Event_flag. Add at least 5 samples per micro-window you care about; else use hourly windows.
2. Add stops data: Stop_label | Opening_hours | Dwell_minutes | Priority (High/Med/Low) | Deadline_window (if any).
3. Sample ahead: Use “Depart at” to add predicted times for key buckets. Mark those rows Event_flag = 2 so AI can down-weight them.
4. Analyze: run the prompt below to compute p80 by micro-window and propose bundles with leave-by times and buffers.
5. Commit: block chosen windows on your calendar; include the sequence and buffers. Do a 5-minute live check the morning of.

Copy-paste AI prompt (weekly plan, micro-windows + bundles)

“I have two datasets. Trips: date, day_of_week, depart_time (HH:MM), origin_label, destination_label, travel_minutes, weather, event_flag (1=anomaly exclude, 2=predicted down-weight). Stops: stop_label, opening_hours, dwell_minutes, priority (High/Med/Low), deadline_window (optional). Tasks: 1) For each origin–destination and day_of_week, compute mean, median, p80, spread for 20–30 minute buckets; if any bucket has n<5, merge with adjacent or roll up to 60–90 minute windows and say what you merged. 2) Rank buckets by lowest p80 then low spread; assign confidence (n<5 Low, 5–14 Medium, 15+ High; reduce by one level if most samples are event_flag=2). Exclude event_flag=1 from baseline. 3) Propose 2–3 errand bundles per weekday (60–90 minutes each) that cluster nearby stops. For each bundle, choose a leave-by micro-window that keeps all legs inside reliable buckets using p80 values, include buffers (High=10% or 5m min, Medium=20% or 7m, Low=30% or 10m). 4) Return: a) plain-English rules (Avoid/Prefer), b) calendar-ready lines: Day, Bundle#, Leave_by, Sequence (Stop→Stop), leg p80s, buffers, total ETA, confidence, fallback window, c) a CSV table with weekday, window_start, window_end, mean, median, p80, n, spread, confidence, rule.”

Worked example

• Inputs: 16 trips Home→Grocery (Mon), 12 trips Home→Pharmacy (Mon), 10 trips Home→PostOffice (Mon); dwell: Grocery 20m, Pharmacy 8m, PostOffice 6m.
• AI summary: Mon micro-windows — 09:40–10:00 p80=12 (n=6, High), 10:00–10:20 p80=13 (n=5, Med), 16:40–17:00 p80=25 (n=7, High, avoid). Avoid 07:30–09:00 rising edge.
• Bundle: Leave 09:42 (inside 09:40–10:00). Sequence: Home→PostOffice (p80 8m + buffer 5m) → Pharmacy (p80 7m + buffer 5m) → Grocery (p80 10m + buffer 5m). Total drive p80 ≈ 25m; buffers ≈ 15m; dwell ≈ 34m; total ≈ 74m. Confidence: High. Fallback: 13:00–13:30 (Medium).

Metrics to track (weekly)

• On-time arrival rate to deadlines (target ≥ 95%)
• Minutes saved per week vs. prior baseline (target 30–90 minutes)
• Variance reduction: difference between p80 and median (smaller is better)
• Hit rate: percent of trips within the chosen window’s p80 (target ≥ 80%)
• Replans required due to surprises (target ≤ 1 per week)

Common mistakes & quick fixes

• Mistake: Ignoring dwell times and store hours. Fix: include them in the stops dataset and choose windows that keep all legs feasible.
• Mistake: Departing in a rising spike. Fix: shift 10–15 minutes earlier into the stable plateau.
• Mistake: Treating predicted samples equal to real drives. Fix: down-weight predictions; confirm with a live check.
• Mistake: Overfitting to one weekday. Fix: merge Tue/Thu or Mon/Wed when patterns match and n is low.

1-week action plan (crystal clear)

1. Today: Add dwell minutes and opening hours to your stops; label routes clearly (e.g., Home–Grocery).
2. Next 5 days: Log trips and add 3–5 “Depart at” predictions per target micro-window.
3. Day 6: Run the prompt; pick 2–3 bundles for next week with leave-by times and fallbacks.
4. Day 7: Block calendar with leave-by, sequence, and buffers; during the week, do a 5-minute live check before leaving and record actuals.

This turns your reliable windows into a full errand playbook: fewer drives, fewer spikes, more on-time arrivals — measured weekly, improved monthly. Your move.

Smart call on the calibration pack. You’ve nailed the “what.” Here’s how to operationalize it so anyone can produce on-brand copy, fast—and you can measure the impact.

Goal: turn your one-pager into a repeatable Voice Ops system: draft faster, edit less, stay consistent across channels, and track results.

What you’ll need

• Your one-pager (audience line, behaviors, Dos/Don’ts, Never Words, templates)
• 5 “lighthouse lines” you love (gold standards) and 1 off-brand line
• One current draft to test (email, post, ad)
• 15–30 minutes to set up the prompts; 10 minutes per piece to run them

System steps (simple, scalable)

1. Create a Voice DNA card (one page). Add two extras beyond your calibration pack: 5 lighthouse lines (best-in-class examples) and 1 off-brand line with a note on what makes it wrong. These become your comparison anchors.
2. Stand up a “Voice Coach” prompt that enforces your rules, asks clarifying questions, and self-scores outputs before you see them.
3. Build a scenario set of six moments that stress-test tone: new offer, apology, price change, delay, how-to tip, testimonial. You’ll reuse these for training and audits.
4. Run the Compression Test (cut to 60% without losing tone) and the Expansion Test (add details without getting formal). If the voice breaks, tighten your rules.
5. Quantify fit with a simple scorecard—pass/fail threshold at 24/30. This keeps review quick and objective.

Copy-paste prompt: Voice Coach (use as your default)

“You are our Brand Voice Coach. Use our guide to write and self-audit before delivery. Guide: Audience = [one line]. Voice statement = [2 sentences]. Behaviors = [3 actions]. Dos = [4]. Don’ts = [4]. Never Words (with swaps) = [list]. Signature moves = [open/close]. Templates = [3 short examples].

Process:
1) Ask up to 3 clarifying questions if needed.
2) Draft the piece.
3) Self-audit with scores 0–10 for Warmth, Clarity, Confidence; plus Plain English, Specificity, Momentum. Show a 30-point total.
4) If total < 24/30, revise once and rescore.
Output sections:
A) Final draft (ready to paste)
B) Scorecard with 6 scores + total
C) 2 alternates: i) shorter, ii) punchier CTA.
Task: [e.g., “Email opener about our spring tune-up offer; 25 words max.”]”

Prompt variant: Scenario Trainer

“Use our brand guide to write six short pieces for these scenarios: new offer, apology, price change, service delay, how-to tip, testimonial. 40–70 words each. Keep behaviors visible. Include the self-audit scorecard and one note on how each piece follows the signature moves. Guide: [paste your one-pager].”

Prompt variant: Consistency Scorecard (fast audit)

“Audit this draft against our voice. Guide: [paste one-pager]. Score 0–10 for Warmth, Clarity, Confidence, Plain English, Specificity, Momentum. Total /30. List 3 off-brand risks with fixes. Return A) minimal edits (same structure), B) stronger rewrite (tighter, benefit-first). Draft: [paste].”

Prompt variant: Compression + Expansion

“Rewrite this draft in our voice at 60% of the length (same meaning, tone intact), then expand to 120% with two concrete details and one clearer next step. Return both versions. Guide: [paste]. Draft: [paste].”

What to expect

• First week: faster drafts and fewer rewrites because the Coach self-corrects before you review.
• After 3–5 pieces: a stable tone and reusable snippets you can drop into emails, posts, and ads.
• Within a month: measurable consistency and better team handoffs.

Track these

• First-draft acceptance rate (approved with minor edits). Baseline it; aim for steady improvement.
• Edits per 100 words (track in comments). Lower is better.
• Time to publish (draft start to approved). Watch it drop as rules stabilize.
• Channel outcomes: email reply rate/CTR, social saves/comments, ad CTR. Compare to your last 3 similar pieces.

Common mistakes and fixes

• Rules too abstract: replace adjectives with actions (e.g., “clear” → “12–16 word sentences; define jargon in 5 words”).
• Voice drift across scenarios: lock two signature moves; use them in every piece.
• Over-editing good drafts: use the 24/30 threshold. If it passes, ship with minimal tweaks.
• Letting AI waffle: force the self-audit and a shorter alternate every time.

One-week plan (simple and concrete)

1. Day 1 (30 min): Finalize your one-pager. Add 5 lighthouse lines and 1 off-brand example.
2. Day 2 (20 min): Set up the Voice Coach prompt with your details. Save it as a reusable template.
3. Day 3 (20 min): Run the Scenario Trainer. Keep the two best outputs as ready-to-use templates.
4. Day 4 (10 min): Audit one live draft with the Scorecard. Ship if ≥ 24/30.
5. Day 5 (10 min): Do Compression + Expansion on a key page or email. Keep the winner.
6. Day 6 (10 min): Document 5 Never Words with swaps you actually use. Share with your team.
7. Day 7 (15 min): Review metrics: acceptance rate, edits/100 words, time to publish. Pick one rule to tighten.

Insider tip: keep a “Best Lines” file. When a line performs or feels right, save it. Your library becomes the fastest way to maintain voice and speed up new pieces.

Your move.

Hook: Stop guessing. Build a minimum‑viable personalized plan in one hour, run short cycles, and judge success by a few simple numbers.

The problem: Most parents jump between apps and rely on gut feel. That creates inconsistent practice, fuzzy goals, and motivation drop-offs.

Why it matters: The right level (not too easy, not too hard) compounds quickly—10–20% accuracy gains in 2–6 weeks are realistic when you keep sessions short, track the same few metrics, and adjust with simple rules.

Lesson learned: Treat this like a small project. Define outcomes, run a 2‑week sprint, review the data, and iterate. AI gives options and structure; you supply judgment and motivation.

What you’ll need:

• One clear goal per subject (plain language).
• One AI tutor or adaptive app to start, plus one hands‑on alternative (cards, manipulatives, read‑alouds).
• A tracker (notebook or simple spreadsheet).
• 30–60 minutes to set up; 20–30 minutes per session; 10–15 minutes weekly review.
• Privacy guardrails: use an alias; avoid names, addresses, or health details.

How to do it (step‑by‑step):

1. Define targets: Pick 2–3 specific skills (e.g., “add fractions with like denominators,” “find the main idea in a paragraph”).
2. Baseline snapshot (5 minutes): One math item, one reading item, one “what feels hard?” question. Record accuracy, time per item, and confidence (1–5).
3. Choose tools: One app for practice + one off‑screen option. Keep the tool count to one to avoid context switching.
4. Set up your tracker: Columns: Date, Subject, Target, Activity, Accuracy %, Avg time/item, Confidence (1–5), Frustration (1–5), Error type, Next step.
5. Calibrate difficulty (the Goldilocks Band): Aim for 70–85% accuracy, 1–2 minutes per item, confidence 3–4/5. Too easy? Increase challenge slightly. Too hard? Add scaffolds or switch format.
6. Schedule cadence: 3–5 sessions/week, 20–30 minutes each. End with a 2‑minute reflection: “What felt easy? What was sticky?”
7. Run a 2‑week micro‑plan: Alternate app practice with a hands‑on or reading task. Keep one target in focus until it stabilizes in the Goldilocks Band.
8. Weekly review (10–15 minutes): Compare to baseline. Decide using simple rules (below). Note one change for the coming week.
9. Teacher alignment (optional): Share your targets and ask for one confirming skill or sample question to stay on track.
10. Privacy & guardrails: Use an alias in tools; never enter personal or health data; export/delete logs if you stop using a tool.

Decision rules (copy this into your tracker):

• If accuracy > 85% for two sessions then increase difficulty one notch or move to mixed practice.
• If accuracy 55–69% then add one scaffold (hint, worked example) and reduce item load by 25%.
• If confidence ≤ 2/5 with rising time/item then switch the next session to a hands‑on format and retry after a win.
• If time/item drops but errors repeat then run an error‑type mini‑lesson (see prompt below).

Metrics that matter (set expectations):

• Accuracy % (target band 70–85% during learning; 90%+ at mastery).
• Average time per item (steady or trending down by 10–25% over 2–6 weeks).
• Confidence 1–5 (aim for 3–4 sustained; sudden drops signal overload).
• Error types (concept, procedure, attention). Reduce the dominant error type first.

What to expect:

• Placement and decent first recommendations within a few sessions.
• Small wins (accuracy or time improvements) inside 2–3 weeks with consistent cadence.
• Clear personalization after 1–3 adjust‑and‑review cycles.

High‑value insider trick: Track a simple “Goldilocks Index” note after each session: E (easy), J (just right), H (hard). Your fastest lever isn’t switching apps; it’s keeping most sessions in “J.” When two “H”s appear in a row, apply the “reduce load + add scaffold” rule immediately.

Copy‑paste AI prompts (use as‑is):

Micro‑plan generator

“Create a 2‑week learning plan for a child (no personal details) targeting: [LIST 2–3 SKILLS]. Baseline: [ACCURACY %], [AVG TIME/ITEM], [CONFIDENCE 1–5], main error type: [CONCEPT/PROCEDURE/ATTENTION]. Provide: (1) 5 sessions/week, 20–30 minutes each, (2) a mix of app practice and off‑screen activities, (3) scaffolds for ‘hard’ days, (4) a simple tracker template (what to record each session), and (5) end‑of‑week reflection questions. Do not ask for personal data.”

Error‑fix generator

“Design a 15‑minute mini‑lesson to address this error: [DESCRIBE ERROR]. Include one worked example, three practice items with step‑by‑step hints, and one hands‑on or read‑aloud alternative. Keep language friendly and concise. No personal data.”

Diagnostic check

“Write three diagnostic questions per target skill to quickly determine level (easy/just right/hard). For each, provide: correct answer, common wrong answer, and what the wrong answer reveals. No personal data.”

Mistakes and fixes:

• Mistake: Tool‑hopping when results are slow. Fix: Commit to one tool for a full 2‑week cycle before changing.
• Mistake: Chasing 100% too early. Fix: Hold the Goldilocks Band until patterns are stable; push to 90%+ only at the end.
• Mistake: Ignoring motivation signals. Fix: Track confidence and add quick wins after any “H” day.
• Mistake: Overlong sessions. Fix: Cap at 30 minutes; stop on a win when possible.

One‑week starter plan (crystal clear):

1. Day 1 (30–45 min): Baseline snapshot; set two targets; choose one tool; build the tracker; paste the micro‑plan prompt and adopt the plan.
2. Day 2: Session 1 (20–25 min). Log accuracy, time/item, confidence, error type; mark E/J/H.
3. Day 3: Session 2 with an off‑screen activity (manipulatives or read‑aloud). Apply decision rules if Day 2 was E or H.
4. Day 4: Session 3 in the app. Aim for the Goldilocks Band. End with a 2‑minute reflection question.
5. Day 5: Light review: run the diagnostic check prompt for any sticky error; do the 15‑minute mini‑lesson.
6. Day 6: Session 4 (mixed practice). Log metrics; adjust difficulty only if you have two sessions outside the Band.
7. Day 7 (10–15 min): Weekly review. Compare to Day‑1 baseline; decide one change for next week; celebrate one specific win.

Session script (use verbatim if you like): “Today’s goal is [TARGET]. We’ll work for 20 minutes. After five items, we’ll check how it feels. If it’s too easy, we’ll bump difficulty one notch; if it’s tough, we’ll do one worked example and a hands‑on version. We finish with one question you know you can get right.”

Your move.

Agreed — your constraint-packed prompts cut noise fast. Let’s layer on a convergence system that turns those 20 seeds into 1 backed-by-metrics concept in under an hour, with zero debate spirals.

The problem: Idea volume is high; decision quality is inconsistent. Without structured convergence, you burn minutes scoring and arguing.

Why it matters: Workshops should leave with one owner, one test, one metric — and a calendar block to execute. That’s the difference between creativity and progress.

What you’ll need (additions to your list)

• One shared scorecard (columns: Impact, Feasibility, Speed, Confidence; 1–5 each)
• Pre-baked constraint toggles: budget ($0, $500, $2k), time (24h, 7d, 30d), channel (email, social, in-product)
• Three prompts: Cluster & Dedupe, Concept Card, Pre-mortem (copy/paste below)
• Decision rule: if tied, pick the idea with the highest Impact x Confidence

Experience/lesson: Two-pass generation wins. First pass creates short titles only. Second pass expands only the top titles using a fixed concept card format. This slashes noise, accelerates scoring, and avoids over-investing in weak ideas.

Run-of-show (adds to your flow — 60–75 minutes)

1. Clarify (3 min) — One-sentence problem + 2 constraints aloud. Set the single success metric you’ll optimize for in tests (e.g., qualified sign-ups).
2. Idea titles only (5 min) — Use your 20 one-line idea burst. Titles + one-line benefit, nothing more.
3. Cluster & dedupe (5 min) — Paste the Cluster & Dedupe prompt with your 20 titles. Expect 4–6 clean clusters and removal of duplicates.
4. Silent dot-vote (3 min) — Each person gets 3 votes. Top 3 titles move forward.
5. Concept cards (12–15 min) — For each top title, run the Concept Card prompt to produce a tight, comparable format: user, problem, promise, channel, asset, single metric, 3-step 7-day test, and cost.
6. Score & weight (10 min) — Everyone scores 1–5 on Impact, Feasibility, Speed, Confidence. Calculate: Total Score = (Impact×2 + Feasibility + Speed + Confidence) ÷ 5. Top score advances.
7. Pre-mortem (7 min) — Use the Pre-mortem prompt to stress-test the winner and add safeguards.
8. Owner + calendar (3 min) — Assign a named owner and book a 60–90 minute Day-1 action block before leaving the room.

Copy-paste AI prompts (use as-is)

• Cluster & Dedupe“You are assisting a creative workshop. Given these idea titles: [PASTE 20 ONE-LINE IDEAS], 1) group them into 4–6 clusters with clear labels, 2) remove duplicates or near-duplicates, 3) return a shortlist of the 8 strongest, each as a short title plus one-line benefit, 4) note the single biggest constraint risk for each (budget, time, channel, or capability). Keep it concise and numbered.”
• Concept Card (standardized)“Create a one-page concept card for idea: [TITLE]. Audience: [PERSONA]. Constraints: budget [$X], time [7 days], channel [ONE]. Include exactly: 1) Problem (1 sentence), 2) Promise (primary user benefit, 1 sentence), 3) Asset needed (max 2 items), 4) Channel + call-to-action (1 sentence), 5) Single success metric for 7 days (define threshold), 6) 3-step minimum viable test with estimated time and cost per step, 7) Risks + quick mitigations (3 bullets). Keep it tight and skimmable.”
• Pre-mortem & Countermoves“Assume this concept fails in 7 days. List the top 5 reasons it likely failed (specific, evidence-based), then propose one counter-move for each that can be done within the same constraints. End with a revised test plan that fits in 7 days and under [$X].”

Metrics to track (during and after the session)

• Ideas per minute (target ≥3/min in burst)
• Deduped ratio (unique ideas ÷ total; target ≥60%)
• Testable ratio (concept cards with clear 7-day test; target 100%)
• Decision time from shortlist to winner (target ≤15 min)
• Concept-to-experiment start time (calendar block created; target same day)
• 7-day test win-rate (met threshold metric; track trend over sprints)

Common mistakes & fixes

• Overwriting before scoring — Fix: Titles-only first, then expand the top 3.
• Scoring drift — Fix: Use the weighted formula and keep the single success metric consistent across concepts.
• Ambiguous ownership — Fix: Name the owner in the room and schedule the Day-1 block before closing.
• Unbounded novelty — Fix: Use constraint toggles (time, budget, channel) inside the prompts.

Insider trick: the “novelty dial.” Run the burst twice — once with conservative constraints ($0, 24h, existing channels), once with bold constraints ($2k, 30 days, new channel). You’ll get a safe option and a breakthrough option; let the scorecard pick the winner.

1-week action plan (crystal clear)

1. Day 1 (60–90 min): Generate titles → Cluster & Dedupe → Concept Card → Score → Pick → Book calendar.
2. Day 2: Build minimal assets (max 2). Run the Pre-mortem prompt and adjust the plan.
3. Day 3–6: Execute 3-step test. Log the single metric daily in a shared sheet. Midpoint tweak only if metric is tracking below 50% of target.
4. Day 7 (15–20 min): Review metric vs threshold. Decide: scale, iterate, or kill. If scaling, draft the next 14-day plan with the Concept Card prompt upgraded for scale.

What to expect: 20 raw ideas distilled to 3 comparable concept cards, 1 test-ready winner, and a booked first action — all within the session. By week’s end, a yes/no decision grounded in a single metric, not opinion.

Your move.