Jeff Bullas

Upgrade the two-fix idea with one small pro move: force structure. When you tell the AI exactly how to format its answer, you get clearer teaching, faster fixes, and fewer detours. Think of it as giving your coach a checklist.

Why this works

Beginners don’t need more code; they need clarity. A structured response turns a scary error into a tidy lesson: what broke, how to patch it now, how to fix it properly, and how to test it. You learn while you ship.

What you’ll need

• Your tiny failing snippet (1–20 lines) and the exact error text
• Your language and version (e.g., Python 3.11 or JavaScript in Chrome)
• An editor you like and an AI chat tool

Do / Do not

• Do state your environment (language + version). It prevents wrong advice.
• Do ask for two fixes: quick patch and robust fix, plus trade-offs.
• Do ask for 2–3 tiny tests (one normal, one edge case).
• Do run the code yourself after each change.
• Do not paste giant files. Share the minimal failing block.
• Do not accept big rewrites. Ask for a diff or changed lines only.
• Do not skip the “why.” Learning sticks when you hear the reason.

Step-by-step (10–15 minute loop)

1. Reduce to the smallest reproducible example.
2. Note your environment: language, version, runtime (e.g., terminal, browser).
3. Paste your code + exact error into the prompt template below.
4. Apply the quick patch first. Run it. Confirm the error disappears.
5. Run the AI’s suggested tests. Add one edge case of your own.
6. Study and apply the robust fix (usually input checks or clearer logic).
7. Ask the AI for a one-line rule to remember and one question to check understanding.

Copy-paste prompt — “Two-Fix Debugging Template”

Prompt: I’m using [language + version]. Here’s the smallest code that fails and the exact error/output. Please answer using these sections only: 1) Diagnosis (plain English), 2) Quick Patch (smallest change), 3) Robust Fix (future-proof, with a one-line trade-off vs the patch), 4) Tests (3 cases: typical, edge, error), 5) Prevention Tip (one sentence), 6) Next Step if it still fails. Code: [paste code]. Error/output: [paste exactly].

Follow-up prompt — “Diff Only”

Prompt: Show only the changed lines needed for the Quick Patch, with line numbers based on my snippet. Then explain the change in one sentence.

Worked example (Python)

Code: def average(nums): return sum(nums) / len(nums)

Call: print(average([]))

Error: ZeroDivisionError: division by zero

What the AI should deliver with the template

• Diagnosis: You’re dividing by zero when the list is empty.
• Quick Patch: If nums is empty, return 0.0 (or None). Small change, fast success.
• Robust Fix: Validate input and raise a clear error; document behavior. Safer and teaches good habits.
• Tests: average([2, 4]) → 3.0; average([]) → 0.0 or raises ValueError; average([5]) → 5.0.
• Prevention Tip: Guard against empty inputs before dividing.

Possible Quick Patch

def average(nums):
if not nums:
return 0.0
return sum(nums) / len(nums)

Possible Robust Fix

def average(nums):
if not isinstance(nums, list):
raise TypeError(“nums must be a list of numbers”)
if len(nums) == 0:
raise ValueError(“cannot compute average of an empty list”)
return sum(nums) / len(nums)

Insider trick: force a “teaching shape”

• Add this to your prompt: “End with a one-question quiz to check my understanding.”
• Ask for analogies: “Explain it like a recipe in two lines.” It sticks.
• When fixes are long, require sections and word limits to keep it readable.

Common mistakes & quick corrections

• Big pastes: If the AI gets vague, your snippet is too large. Trim to 5–20 lines.
• Version mismatch: Wrong advice often comes from missing versions. Always state them.
• No tests: Ask for three tiny tests every time. Run them.
• Over-fixing: If the AI rewrites everything, use the “Diff Only” prompt.
• Blind trust: Treat AI answers as hypotheses. Verify by running the code.

Action plan (one-week, low-stress)

1. Day 1: Pick a 10–20 line script. Use the template to get a Diagnosis and Quick Patch.
2. Day 2: Ask for the Robust Fix and apply it. Re-run tests.
3. Day 3: Break your code on purpose. Use the template to recover in under 15 minutes.
4. Day 4: Add one guard (input check) to a function. Ask AI for a one-line rule to remember.
5. Day 5: Practice “Diff Only” changes on a new bug.
6. Day 6: Teach-back: ask AI to quiz you with five quick questions on today’s bug.
7. Day 7: Do a full loop: isolate → template → patch → tests → robust fix → reflect.

Expectation check

• Good answers look structured, short, and runnable.
• Great answers make you faster and smarter: a fix, a test, and a rule you can recall.
• If you’re still stuck after two loops, reduce the snippet and restate your environment.

Closing nudge

Two fixes plus structure turns AI into a calm coach. Keep your snippets small, your questions clear, and your feedback loop tight. You’ll learn faster while shipping small wins—exactly what builds confidence.

Quick win: You can lift your site’s accessibility score fast by pairing AI-written alt text with a simple ruleset and a light human review. Think hours and days, not weeks and months.

Why this works: Most images are simple. AI handles those well. The tricky bits are context, text inside images, and brand terms — that’s where a reviewer and a tight prompt do the heavy lifting.

What you’ll need

• Image list or sitemap with image URLs and page URLs
• CMS access (or staging) to update alt attributes
• A short style guide: character limits, tone, brand terms, and a decorative rule
• One reviewer who knows your product/content

High‑leverage steps

1. Bucket images: product, hero, infographic/chart, screenshot, logo, decorative, functional (icons/buttons).
2. Capture context: for each image, grab the page title, nearby caption/heading, and whether the image is a link or button. Context turns mediocre alt into meaningful alt.
3. Run the AI with structure: use the copy‑paste prompt below to force consistent output, including a decorative decision, extended description when needed, and a confidence score.
4. Auto‑apply low‑risk groups: decorative, logos, generic hero images. Queue product, charts, and screenshots for review.
5. Review with a 7‑point checklist: accuracy, context relevance, brevity, no SEO fluff, include visible text, correct people roles, correct handling of linked/functional images.
6. Feedback loop: note recurring errors (e.g., missed text-in-image), update the prompt/style guide, and rerun. Track edit rate and aim to get it under 10% by iteration two.

Insider upgrades that pay off

• Context injection: pass page title and surrounding text so the AI knows why the image exists.
• Link-awareness: if an image is a link or button, alt should describe the destination or action, not the pixels.
• Decorative gate: many hero backgrounds/confetti are decorative; require the AI to explicitly choose empty alt (“”).
• Brand dictionary: give the AI your product names and forbidden phrases (e.g., “best, amazing”).
• Extended descriptions: for charts and complex screenshots, store a 1–3 sentence long description and reference it via aria-describedby or nearby text.

Copy‑paste AI prompt (context‑aware, structured)

Act as an accessibility specialist. You’ll receive: page_title, surrounding_text, file_name, link_destination (optional), image_role (product, hero, chart, screenshot, logo, decorative, functional). Task: write alt text and, if needed, a short extended description. Rules: 1) Alt ≤100 characters; no marketing or keyword stuffing; no “image of.” 2) Include visible text in the image verbatim in quotes. 3) Describe people by role/activity, not names. 4) If decorative, alt = empty string. 5) If image is a link/button, alt describes the destination or action. 6) For charts/screenshots, add a 1–2 sentence extended description. Output exactly in this schema:
ALT:
LONGDESC: <1–2 sentences or “none”>
ROLE:
DECORATIVE:
LINK_FOCUS:
CONFIDENCE: <0–100>
ISSUES:

Style guide template (fill in once, reuse forever)

• Alt length: 60–100 chars (≤80 for icons/buttons).
• Tone: factual, neutral, no adjectives unless informative.
• People: role/action only (e.g., “nurse drawing blood”).
• Logos: company or product name only.
• Charts: state what, where, and trend (e.g., “Line chart showing 12% YoY growth, 2022–2024”).
• Decorative: empty alt (“”). If unsure, return for review.
• Forbidden: “image/picture of,” “best/top #1,” repeated keywords.

Worked examples

• Product photo: a stainless 1‑liter travel mug on a white background; page title “ThermoPro Go Mug,” not a link. Alt: “ThermoPro Go Mug, stainless 1‑liter travel mug with flip lid.” Longdesc: none.
• Decorative confetti banner: hero background behind the headline; not a link. Alt: “” (empty). Longdesc: none.
• Sales chart screenshot: bar chart with labels “Q1–Q4 2024,” caption mentions “12% YoY growth.” Alt: “Bar chart of 2024 quarterly sales, Q4 highest; labels ‘Q1–Q4 2024’.” Longdesc: “Quarterly sales rise across 2024, peaking in Q4; overall growth about 12% YoY.”
• Linked image (CTA): image is a button linking to pricing. Alt: “View pricing plans.” Longdesc: none.

Common mistakes and quick fixes

• SEO stuffing: ban marketing words in the prompt and sample 10–20% for repeats.
• Missing text-in-image: require “include visible text verbatim.” If OCR is weak, mark for review.
• Over‑describing decorative art: force a decorative decision; empty alt if it adds no information.
• Wrong focus on linked images: ensure the prompt checks link_destination and writes action/destination.
• Vague chart summaries: insist on a trend + timeframe + notable peak/low in longdesc.

What to expect

• First pass: 80–95% correct on simple images.
• Human review time: 15–60 seconds per image on average; longer for charts/screenshots.
• After two iterations: edit rate under 10% is realistic with a tuned prompt and style guide.

7‑day action plan

1. Day 1: Export image and page URLs. Fill the style guide. Tag image roles by bucket.
2. Day 2: Run 100‑image pilot using the structured prompt with context fields populated.
3. Day 3: Reviewer checks all flagged items and 20% sample of auto‑applied. Log recurring errors.
4. Day 4: Update prompt (ban list, people roles, link logic). Add brand dictionary.
5. Day 5: Roll out to remaining images; auto‑apply low‑risk groups; queue product/charts/screenshots.
6. Day 6: QA sampling; spot‑fix weak buckets (e.g., charts) with a longdesc‑focused re‑run.
7. Day 7: KPI report: meaningful alts %, average review time, edit rate. Lock the process for monthly refresh.

One more ready‑to‑use prompt (charts/screenshots)

Write a concise alt (≤100 chars) and a 2–3 sentence extended description for this chart or screenshot. Include the chart type, timeframe, key trend, and any labeled peaks or lows. Include visible text in quotes. Avoid marketing language. Output exactly: ALT: … NEWLINE LONGDESC: …

Bottom line: Treat AI like a fast apprentice with great eyesight. Give it context, enforce crisp rules, and keep a human in the loop. You’ll get accessible, meaningful image descriptions at scale — and results you can measure.

Spot on, Aaron: your 3‑minute verification and KPIs are the backbone. Quoting exact source lines and limitation sentences is the fastest way to kill hallucinations. I’ll add a simple “Source Map + Fact Fence” system that gives you two extra safeties: number locks and a contradiction check.

Why this works

• Source Map turns every claim into a traceable line back to the paper.
• Fact Fence limits what the AI is allowed to say (and how strongly it says it).
• Contradiction check automatically flags anything the AI wrote that isn’t backed by your quotes.

What you’ll need

• Paper PDF or full text open.
• A note or spreadsheet with four columns: Claim, Verbatim quote, Location (page/section/table), Confidence (High/Medium/Low).
• Your checklist (citation, study type, n, main outcome, exact numbers/CI/p-values, one limitation).

Step-by-step (10 minutes, repeatable)

1. Build the Source Map (3–4 min):
   • Copy the exact line for sample size (n) into the “Verbatim quote” column. Add the page/section/table.
   • Copy the exact main outcome numbers (effect size, CI, p) and their table/figure number. No rounding yet.
   • Copy one limitation sentence verbatim with its location.
2. Lock the numbers (1 min):
   • Paste numbers as text, not retyped. Note the units (%, percentage points, mg/dL, etc.).
   • Write “No rounding” next to any key figure you plan to quote.
3. Set the Fact Fence (1 min):
   • Pick your certainty bucket: Observational = “associates with” or “linked to.” Randomized = “reduced” or “improved,” but still cautious if small/short.
   • Ban words: “proves,” “causes,” “definitive,” unless the paper is a large, well-powered RCT with clear primary endpoints and even then use sparingly.
4. Draft the 3 lines (2–3 min):
   • Line 1: question + design (and population).
   • Line 2: main result with exact numbers and the table/figure location.
   • Line 3: one limitation and an honest certainty statement (preliminary/associational/short follow‑up).
5. Run the contradiction check (1 min):
   • Ask the AI to list anything in your draft that is not directly supported by a quote/location in your Source Map. Fix or label as “not reported.”

Premium template: the “Source Map” you can reuse

• Claim: Sample size
• Verbatim quote: “We enrolled n=___ participants …”
• Location: Methods, p.__
• Confidence: High

• Claim: Main outcome
• Verbatim quote: “Risk ratio 0.__ (95% CI 0.__–0.__; p=0.__)”
• Location: Table __, p.__
• Confidence: High

• Claim: Key limitation
• Verbatim quote: “This study is limited by …”
• Location: Discussion, p.__
• Confidence: High

Copy‑paste AI prompt (structured, low‑risk)

Use this when you already have your quotes pulled:

“You are an evidence‑bound assistant. Use only the quoted excerpts and locations I provide. If a detail is missing, write ‘not reported.’ Output exactly this structure: 1) Full citation line, 2) Study design and population, 3) Main result with exact numbers and the table/figure/page location, 4) One verbatim limitation sentence with location, 5) A three‑sentence plain‑English summary using cautious language (associates with/suggests for observational; reduced/improved for randomized, without overstating). Do not add new numbers, methods, or causal claims that aren’t in the quotes.

Quotes and locations: [paste n line + location] | [paste main outcome line + location] | [paste limitation sentence + location]

Optional “red‑team” prompt (30‑second contradiction check)

“Compare this draft summary to the quoted excerpts and locations. List any claim, number, or causal wording in the draft that is not directly supported by a quote. For each item, suggest the minimal fix (remove, soften wording, or mark ‘not reported’).”

Worked example (illustrative)

• Design: Randomized trial, adults with condition X.
• Numbers: “n=120” (Methods, p.3); “Mean change −2.1 vs −0.8; between‑group difference −1.3 (95% CI −2.4 to −0.2), p=0.02” (Table 2, p.6).
• Limitation: “Follow‑up was 8 weeks and the sample was small” (Discussion, p.9).

• 3 lines:
  • Trial tested intervention X in 120 adults (randomized).
  • Main outcome improved vs control: difference −1.3 (95% CI −2.4 to −0.2; p=0.02; Table 2, p.6).
  • Short 8‑week follow‑up and small sample make results preliminary.

Common mistakes and quick fixes

• Mixing % and percentage points. Fix: write “pp” for percentage points; keep percent signs for relative change.
• Using subgroup numbers as the headline. Fix: headline must reflect the primary analysis (usually intention‑to‑treat).
• Rounding away meaning. Fix: quote exacts; round only in parentheses.
• Paraphrasing limitations into nothing. Fix: include one verbatim limitation line.
• Copying press release language. Fix: ignore secondary sources; use the PDF only.

Action plan (one week)

1. Day 1: Create your Source Map template and paste it as a reusable note.
2. Day 2–3: Do two papers end‑to‑end (10 min each). Track time and whether every number has a location.
3. Day 4: Add the contradiction check to your routine.
4. Day 5–6: Reduce time to 7–12 minutes without losing traceability.
5. Day 7: Review KPIs: 100% numbers quoted or “not reported,” corrections down 50%+.

Closing thought: Slow is smooth; smooth is fast. Lock the numbers, quote the limitation, and let your Source Map do the heavy lifting. Confidence rises, corrections fall.

Nice point — you’re right: AI tutors explain things in plain language and can show the math step‑by‑step, but they aren’t perfect so a quick check is smart. Here’s a practical way to get reliable, useful guidance right away.

Quick win (under 5 minutes): Ask the AI to convert grams to moles for a simple sample and show each arithmetic step. You’ll see how it labels units and uses the molar mass — immediate confidence booster.

What you’ll need

• The exact problem text (numbers, units, and question).
• Any work you’ve already done.
• A periodic table or molar mass values (optional — AI can provide them).
• Note: the level of detail you want (hint vs. full solution).

Step-by-step: how to get a clear AI walkthrough

1. Paste the full problem and say what you want: hint, full solution, or step-check.
2. Ask the AI to label units at every step and to explain why each formula applies.
3. Request intermediate arithmetic (don’t skip steps) and the final answer with significant figures.
4. Try a similar problem yourself, then paste your steps and ask the AI to check them.

Worked example

Problem: You dissolve 5.85 g NaCl in 250.0 mL water. What is the molarity (M)?

1. Find molar mass: Na (22.99) + Cl (35.45) = 58.44 g/mol.
2. Calculate moles: 5.85 g ÷ 58.44 g/mol = 0.1001 mol.
3. Convert volume to liters: 250.0 mL = 0.2500 L.
4. Molarity = moles ÷ liters = 0.1001 ÷ 0.2500 = 0.4004 M → report 0.400 M (three significant figures).

Common mistakes & fixes

• Forgetting to convert mL → L: always check units. Fix: write units on each line.
• Using wrong atomic masses: copy values or ask AI to show them.
• Rounding too early: keep extra digits, round at the end.
• Missing stoichiometric coefficients in reactions: include balanced equation first.

Copy-paste AI prompt (use this exact text)

“Solve this chemistry problem step-by-step: [paste your problem]. Show every formula used, label units on each line, include intermediate arithmetic, and give the final answer with correct significant figures. Explain why each step is done and list two common mistakes students make on this type of problem. Finally, provide one similar practice problem with its full solution.”

Action plan (next 30 minutes)

1. 5 min: Try the quick win prompt with a simple grams→moles or molarity question.
2. 15 min: Ask the AI for a full, labeled walkthrough of a course problem you’ve done and compare results.
3. 10 min: Do the AI’s practice problem and ask it to check your steps.

Try it now with the prompt above. If you tell me whether you’re in general chemistry or organic, I’ll suggest the best phrasing and a tailored practice problem.

Quick answer: Yes — AI can surface podcast and YouTube topics that are more likely to attract sponsors. It speeds research, spots patterns in your audience, and helps craft sponsor-friendly angles you might miss.

Why this matters

Sponsors buy attention that converts: the right topic, framed for a specific audience and a clear sponsor benefit, sells. AI helps you find those topics faster by analyzing your analytics, competitor content, search demand, and advertiser intent.

What you’ll need

• Access to your show/channel analytics (audience age, location, watch/listen time)
• Episode transcripts or video descriptions/comments
• A short list of industries you’d like as sponsors
• AI access (chat models like ChatGPT or an AI tool that can analyze text)
• A spreadsheet to capture ideas and scores

Step-by-step

1. Collect data: export top-performing episodes, keywords, audience demographics, and 50 recent comments.
2. Define your sponsor avatar: list 3 ideal sponsor types and what they want (leads, brand awareness, sales).
3. Ask AI to analyze: feed transcripts + comments + sponsor avatar and ask for topic ideas that align to sponsor value.
4. Score ideas: rate each idea for audience appeal, sponsor fit, production effort, and revenue potential (1–5).
5. Test and measure: produce 2–3 episodes, include clear sponsor-ready segments, then track engagement and sponsor conversations.

Copy-paste AI prompt (use as-is)

“You are a podcast and YouTube growth strategist. Here are inputs: 1) Audience: [insert age, location, interests], 2) Top episodes: [list 3 titles], 3) Sponsor targets: [list industries], 4) Tone: [e.g., conversational, expert]. Generate 20 episode topics that: a) match audience interest, b) highlight clear sponsor benefits, and c) include a suggested 30-second sponsor integration line. For each topic include a short title, one-sentence description, and why a sponsor would pay for it.”

Short example

Topic: “Home Office Upgrades that Boost Productivity” — Description: Tips and product picks to transform a home workspace. Sponsor angle: ideal for ergonomic furniture or software brands; integration: a 30-second testimonial-style mention linking product to increased output.

Common mistakes & fixes

• Mistake: Generating topics without sponsor value. Fix: Always add a sponsor-benefit line when brainstorming.
• Mistake: Ignoring audience data. Fix: Let analytics guide topics—AI should interpret your data, not replace it.
• Mistake: Testing too many at once. Fix: Pilot 2–3 ideas, measure, then scale winners.

30-day action plan

1. Week 1: Export analytics and assemble sponsor list.
2. Week 2: Run the AI prompt, shortlist 10 topics, score them.
3. Week 3: Produce 2 episodes with sponsor-ready segments.
4. Week 4: Review metrics, reach out to 3 potential sponsors with episode briefs.

Reminder: AI speeds discovery and creativity, but sponsorships close with clear audience results and simple sponsor ROI. Start small, measure, and iterate.

Nice — your workflow is exactly right: small batches, one question type, and iterate. That keeps things simple and prevents overwhelm.

Here’s a practical add-on you can use right away to go from spreadsheet to QTI with minimal fuss.

What you’ll need

• A spreadsheet (Excel or Google Sheets) with these columns: QuestionID, QuestionText, OptionA, OptionB, OptionC, OptionD, CorrectOption, Feedback (optional).
• An AI chat tool you already use (tell it you want QTI 2.1 unless your LMS requires a different version).
• A plain text editor and your LMS to import and test the file.

1. Step 1 — Prepare one clear sample row
   1. Example row (copy into a cell or plain text):

      Q1 | What is the capital of France? | Paris | London | Berlin | Rome | A | Good job — Paris is correct.

   2. Keep wording short; avoid special characters like “&” (use “and” instead) or ensure AI encodes them correctly.
2. Step 2 — Use this copy-paste AI prompt

   AI prompt (copy-paste):

   “I have a spreadsheet with columns: QuestionID, QuestionText, OptionA, OptionB, OptionC, OptionD, CorrectOption, Feedback. Use QTI 2.1 format and create multiple-choice assessmentItem XML for these rows. Output ONLY the XML (no explanations). Here is one sample row to format:

   Q1 | What is the capital of France? | Paris | London | Berlin | Rome | A | Good job — Paris is correct.

   Create a complete, importable QTI XML snippet for that single question, with UTF-8 encoding and a clear question ID. Keep answers scored so option A is the correct response.”

3. Step 3 — Create the file and test
   1. Copy the AI’s XML output into Notepad (Windows) or TextEdit (Mac set to plain text). Save as quiz.xml with UTF-8 encoding.
   2. Import one item into your LMS. If it fails, copy the LMS error and one XML snippet back to the AI and ask: “Why did this error occur and please provide a corrected XML snippet.”
4. Step 4 — Batch convert and import
   1. Once one item imports cleanly, convert the next 5–20 rows into XML and repeat.
   2. Keep a copy of the working XML template so you can paste new items into it easily.

Common mistakes & quick fixes

• XML syntax errors — often an unescaped ampersand (&). Fix: replace & with & or ask the AI to escape characters.
• Wrong answer key — confirm CorrectOption uses A/B/C/D and AI maps those to the right responseChoice identifier.
• Encoding issues — save as UTF-8 to avoid strange characters.

Action plan (15–30 minutes)

1. Create 5 sample questions in your spreadsheet.
2. Use the copy-paste AI prompt above with one sample row.
3. Save the XML, import one item, fix any errors, then batch convert the rest.

Reminder: start small, validate one item, then scale. You’ll be surprised how quickly you can build reliable QTI quizzes without writing XML by hand.

Make it bulletproof: lock in constraints, learn from every trip, and use small modules that plug in based on weather and activities. That’s how you go from “good list” to a repeatable system you trust.

Context

• Your rules + counts are solid. Constraints and a 2‑minute audit reduced weight and misses. Now we’ll add modules (kit cards), a delta method (compare trips), and a simple readiness score so the list tunes itself and stays lean.

What you’ll need

• Your current mappings, thresholds, and weight limits.
• A mini weight sheet for common items (close enough is fine).
• Three to six “kit cards” (small, reusable bundles of items).
• Last trip’s “unused/missed” notes (or recreate from memory once).
• A luggage scale and your shoe limit (2–3 pairs).

How to make it work (step-by-step)

1. Build kit cards (modules)
   • Core capsule: 6–8 multi-use items that appear on every trip.
   • Work kit: laptop, charger, adapters, presentation USB.
   • Weather kits: Rain (shell, compact umbrella), Cold (base/mid, hat/gloves), Heat (sun hat, sunscreen, electrolytes).
   • Health/Sleep: meds, first-aid, earplugs, eye mask.
   • Active kit: gym/hike items (shoes, quick-dry tee, bottle).

   Assign each kit an approximate weight. You’ll toggle kits on/off by forecast and activities.

2. Set a readiness score (fast, objective)
   • Coverage (0–4): are activity essentials present?
   • Constraints (0–4): under 90% of weight, shoe limit respected, toiletries deduped.
   • Risk buffer (0–2): one versatile extra layer only if forecast certainty is low and risk appetite is low.
   • Charge status (0–2): phone, headphones, power bank fully charged and packed.

   Score out of 12. 10–12 = green, 7–9 = amber (tune swaps), 0–6 = red (rework).

3. Use a delta method between trips
   • Start from last trip’s final list.
   • Apply new weather and activities; only add/remove what’s changed.
   • Use weights to swap heavy items for lighter equivalents if you exceed limits.
4. Run a cut list if overweight
   • Sort items by “weight × likelihood unused.”
   • Cut from the top until you’re ≤90% of your limit.
   • Favor substitutes: jeans → chinos; boots → trail runners; paper book → e‑reader app.
5. Do the charge + refill check
   • Charge all electronics; pack the charger next to the device.
   • Top up consumables (toiletries, meds, electrolytes) in your kit cards.
6. Post‑trip micro‑audit (2 minutes)
   • Log 3 unused items with reasons (redundant, weather changed, wrong dress code, comfort issue).
   • Log 1 “missed” item and add it to the right kit card.

Premium prompts (copy/paste)

1) Delta + constraints prompt

Act as my packing optimizer. Start from this last-trip checklist: [paste list]. New trip: Location [city, country], Dates [start–end], Activities by day [list], Forecast [high/low °C, precip %, wind km/h, conditions, UV, humidity], Forecast certainty [low/med/high]. Preferences: carry-on only [Y/N], dress code [casual/smart/business], cold tolerance [low/med/high], laundry [Y/N], toiletries provided [Y/N], shoes limit [2/3], risk appetite [low/med/high]. Constraints: bag weight limit [kg], size [cm]. Kits available: Core, Work, Rain, Cold, Heat, Health/Sleep, Active [describe contents briefly]. Item weight hints: [list 10–20 approximations].

Tasks: 1) Produce a delta plan (what to add/remove/swap) vs last trip. 2) Enforce pack-counts (tops = days, bottoms = ceil(days/2), underwear = days+1, socks = days [+1 if hiking], layers = 1 mid + 1 shell). 3) Keep shoes ≤ limit (travel in bulkiest). 4) Respect toiletries provided. 5) If certainty low and risk low, add 1 versatile layer; if risk high, remove 1 nonessential. 6) Estimate total weight; if >90% of limit, propose a cut list ranked by “weight × likelihood unused” with lighter substitutes. Output: a categorized checklist with quantities, a 6–8 item capsule, selected kits (on/off), 3 compact substitutes, 3 contingency items, a 5‑item last‑minute grab list, and the readiness score (0–12) with one-line trade-offs.

2) Kit card builder prompt

Create packing kit cards for my travel pattern. Inputs: typical activities [list], climate range [e.g., −5 to 35°C], dress code mix [casual/smart/business], shoe limit [2/3], airline weight limit [kg]. Output: 5–7 kits (Core, Work, Rain, Cold, Heat, Health/Sleep, Active). For each kit: 3–6 items, compact substitutes, approximate total weight, when to include (rules tied to forecast/activities). Keep items minimal and multi-use.

Mini example

• Trip: 4 days, Berlin, meetings + one museum day, chance of showers 50%, highs 21°C, lows 12°C, wind 18 km/h, UV moderate. Preferences: carry-on, smart-casual, cold tolerance medium, shoes limit 2, laundry no. Bag limit 10 kg.
• Outcome highlight: Capsule = neutral tee, oxford shirt, smart chinos, quick-dry tee, light sweater, rain shell, walking sneakers (travel in), dressy sneakers. Kits on: Work, Rain, Health/Sleep. Estimated weight 8.6 kg. Readiness score 11/12 (green). Trade-off: skip jeans, use chinos; add one backup shirt because certainty is medium and meetings are formal-ish.

Common mistakes and quick fixes

• Static kits that bloat over time → Review kit cards monthly; remove anything unused twice in a row.
• Counting per day, not per trip → Use pack-counts; laundry access reduces counts further.
• Too many chargers/cables → One multi-port charger + short cables; verify voltage and plug type.
• Forgetting airline personal-item rules → Enforce shoe limit and put the bulkiest pair on your feet.
• Electronics weight blind spot → Weigh laptop + power brick; swap to lighter sleeve or tablet if feasible.

1‑week action plan

1. Today (10 min): Draft 5–7 kit cards with rough weights; set your readiness score thresholds.
2. Tomorrow (10 min): Run the Kit Card Builder prompt; merge its output with your draft.
3. Day 3 (10 min): Weigh your heavy hitters (shoes, laptop, jacket). Update weight hints.
4. Day 4 (10 min): Save the Delta + Constraints prompt with your defaults.
5. Day 5 (10 min): Test on your next trip; aim for ≤90% of weight limit.
6. Day 6 (2 min): Post‑trip audit: 3 unused, 1 missed. Update one kit.
7. Day 7 (5 min): Review KPIs; set one swap to cut weight next time.

Bottom line: Modular kits + a delta method + a quick readiness score turn AI‑generated lists into a light, reliable packing system that adapts each trip and keeps stress low.

Nice point: verifying numbers and copying author-stated limitations up front is the simplest habit that prevents most hallucinations. Small routines, big payoff.

Here’s a compact, beginner-friendly routine you can use every time you summarize a study. Short, repeatable, and designed for non-technical readers who want reliable takeaways.

What you’ll need

1. Paper PDF or full text open (or the journal page).
2. Notepad or document for live notes.
3. A short checklist: citation, study type, n (sample size), main outcome, exact numbers/CI/p-values if present, and authors’ limitations/conflicts.

Step-by-step routine (do this every time)

1. Skim for the headline: Read title and abstract. Write the question and claimed result in one line.
2. Verify methods: Open Methods — note design (randomized, observational, meta‑analysis), population, and n. If you can’t find n, stop and look for tables/figures.
3. Lock numbers: From Results or tables, copy exact values (e.g., “n=120,” “risk ratio 0.75, 95% CI 0.60–0.95, p=0.01”). Don’t estimate or round unless you note it.
4. Copy limitations: Find the Limitations/Discussion and paste the authors’ own caution sentence(s). These are your guardrails.
5. Write the 3-line summary: 1) question + design, 2) main result with exact qualifier (effect size or p-value if important), 3) one limitation and the level of certainty (e.g., preliminary, associative, needs replication).
6. Flag anything uncertain: If you can’t find a method or number, don’t guess — write “not reported” or “not located in the text.”

Quick worked example:

Suppose a cohort study claims reduced risk of outcome X with diet Y. Use the routine: note it’s observational (cohort), n=4,500, report the adjusted hazard ratio and CI from Table 2 (do not invent confounders), and copy the authors’ limitation that residual confounding may exist. Your 3-line summary: the cohort of 4,500 found an adjusted HR 0.82 (95% CI 0.70–0.96) associating diet Y with lower X; because it’s observational, this shows association not causation; authors note possible residual confounding and call for trials.

Common mistakes & fixes

• Skipping tables — fix: always open the primary table with the main outcome.
• Paraphrasing away limitations — fix: copy the limitation sentence and paraphrase below it.
• Memorizing numbers — fix: write them down or copy directly from the PDF.

Copy-paste AI prompt (use when asking an AI to draft the summary):

Read this paper [paste title and link or paste abstract and key excerpts]. Summarize in three sentences: 1) study question and design, 2) main result with exact numbers/CI/p-values copied from the Results or Tables, and 3) one clear limitation taken from the authors’ Discussion. If you cannot find a number or method in the text, state “not reported” rather than guessing. Include the full citation line at the top.

Action plan — next time you summarize:

1. Apply this routine once — timed: 7–12 minutes for a short paper.
2. Keep the checklist as a reusable snippet in your notes.
3. Force yourself to copy one limitation sentence verbatim each time for accuracy.

Little habits beat big willpower. Do the routine three times and you’ll notice fewer corrections and more confidence in your summaries.