Jeff Bullas

Hook: Want translations that sound like you — not like a robot? AI can do that if you give it the right instructions and checks.

Why this matters: Most machine translations focus on accuracy. Tone, rhythm and personality often get lost. With a few practical steps you can keep the original style, whether it’s warm, formal, playful or authoritative.

What you’ll need

• Original text and the target language.
• A short description of the desired tone (e.g., “friendly, concise, slightly humorous”).
• 2–3 example sentences that capture the voice you want preserved.
• Optional: glossary of brand terms and preferred translations.

Step-by-step: How to translate while preserving tone

1. Prepare: Collect the original text, tone notes, and sample sentences.
2. Prompt the AI with clear role instructions (see copy-paste prompts below).
3. Ask for 2–3 variant translations (e.g., formal, neutral, playful) to compare.
4. Use back-translation: translate the AI output back to the original language to spot meaning drift.
5. Revise with micro-edits: tweak idioms, contractions, and cultural references.
6. Validate with a native speaker or small user test if possible.

Copy-paste AI prompts

Use these directly. Replace placeholders in ALL CAPS.

1) Preserve tone (friendly, concise)

“You are a professional translator. Translate the following text into TARGET_LANGUAGE while preserving a friendly, concise tone. Keep contractions and casual phrasing consistent. Original: ‘ORIGINAL_TEXT’. Provide three variants: (A) friendly & casual, (B) neutral, (C) slightly more formal. Note any idioms you changed and why.”

2) Preserve formal/authoritative tone

“You are a translator for a professional audience. Translate into TARGET_LANGUAGE preserving a formal, authoritative tone. Keep sentence structure dignified and avoid slang. Original: ‘ORIGINAL_TEXT’. Highlight two alternative word choices for key terms.”

3) Localize for cultural fit

“Translate into TARGET_LANGUAGE and localize cultural references to TARGET_COUNTRY. Maintain HUMOR_LEVEL (e.g., low, medium, high) and the original author’s voice. Original: ‘ORIGINAL_TEXT’. Explain any cultural substitutions made.”

Short example

Original: “Thanks for stopping by — grab a coffee and take a look around.”

Translation (friendly tone): “Gracias por pasarte — toma un café y mira con calma.”

Common mistakes & fixes

• Literal word-for-word translation: fix by asking for idiomatic phrasing.
• Loss of contractions or warmth: fix by specifying formality and giving examples.
• Wrong cultural references: fix with localization instructions and country context.

Action plan — start today

1. Pick one short piece (100–200 words).
2. Run the friendly prompt and review 3 variants.
3. Do a back-translation and one quick native check.

Closing reminder: Small, iterative tests win. Use the prompts, compare variants, and refine. You’ll get translations that sound human and true to your voice within a few tries.

Quick win — try this in under 5 minutes: pick one paper, copy the title + abstract + methods, paste them into the prompt below and ask for a 2–3 sentence summary plus a confidence rating. You’ll immediately see how useful a first-pass AI check can be.

Good point — a repeatable, measurable first-pass is the sweet spot. LLMs speed up reading and flag risks, but they don’t replace experts, raw data checks, or domain knowledge. Your goal: use the AI to prioritize which papers need closer attention.

What you’ll need

• The paper’s title, authors, year, DOI or PDF (abstract + methods at minimum).
• Access to an LLM (a chatbox like ChatGPT or an API).
• A simple spreadsheet or notebook to record outputs (Confidence, Risk flags, Actionability).

Step-by-step (do this)

1. Open the paper and copy the title, abstract and key methods into your clipboard.
2. Paste them into the AI with the prompt below (keep within token limits).
3. Ask for a plain-English summary first, then targeted checks (sample size, controls, stats, conflicts).
4. Record the AI’s Confidence, Risk flags and Actionability in your spreadsheet.
5. Manually verify one flagged item (preregistration, COI, or raw data link).

Copy-paste AI prompt (use as-is)

“Evaluate this research paper. Here are the details: [paste title, authors, year, DOI, and the abstract + key methods]. Tasks: 1) Give a 2-sentence plain-English summary of the main claim. 2) Rate overall confidence: High / Medium / Low and give 1-line justification. 3) List up to 6 risk flags (sample size, blinding, controls, statistics, conflicts, preregistration). 4) Rate actionability for practice on a 0–10 scale. 5) Suggest 3 concrete follow-ups (e.g., look for raw data, replication, code, protocol). Keep answers concise and non-technical.”

Example of expected output

• Summary: “The study reports X improvement in Y from a randomized trial of 120 patients.”
• Confidence: Medium — adequate design but small sample and short follow-up.
• Risk flags: small sample, unclear blinding, no raw data, single-center, industry funding.
• Actionability: 3/10 — interesting but not ready to change practice without replication.

Common mistakes & simple fixes

• Mistake: trusting the abstract alone. Fix: always paste methods and sample info.
• Mistake: using different prompts each time. Fix: use the same prompt for consistency.
• Mistake: treating AI output as final. Fix: verify one flagged item manually or consult an expert if Confidence=Low.

7-day action plan

1. Day 1: Run the prompt on 5 papers and record outputs.
2. Day 2–3: Manually verify one flagged item per paper.
3. Day 4: Triage — pick 1 for expert review, 2 to monitor, 2 low priority.
4. Day 5–7: Repeat with next batch and track KPIs (avg Confidence, % with 2+ flags).

Small, repeatable habits beat one-off deep dives. Use the LLM to sort and focus — then validate the few papers that matter most.

5-minute win: pick one tag list, speak a two-line header before every note, and run the transcript through the prompt below. You’ll get a clean title, summary, tags and 1–3 actions you can use today.

Why this works — headers give the AI clear anchors (date, context, keywords). A controlled tag list stops drift so search and reporting stay clean. The result: fewer lost ideas, faster follow-through.

What you need

• A recorder you’ll actually use (phone voice memos, Otter, or similar).
• Transcription (Otter/Rev/Whisper via any app or service).
• One store for final notes (Notion database or a dated Google Drive folder).
• Optional: an automation tool (Zapier or Make) for hands-off flow.

Your header script (say this at the start of every note)

• “2025-11-22. Project: GreenCo proposal. Keywords: pricing, pilot. Short note: …”

Controlled tag list (start with 10)

• Pilot, Pricing, Sales, Marketing, Product, Client, FollowUp, Research, Idea, Meeting

Step-by-step (start manual, then automate)

1. Record: open your recorder, speak the header, then your thoughts. Keep notes under 3 minutes for best accuracy.
2. Transcribe: upload the audio to your transcription tool. Copy the transcript text.
3. Organize with AI: paste the transcript and tag list into the prompt below. Expect a title, summary, 5 tags, up to 3 actions and a priority.
4. Store: create a note named “YYYYMMDD — Title” in your Notion database or Drive folder. Paste the AI output, attach the audio and transcript.
5. Act: move actions into your task app or Notion task board. Done beats perfect.

Copy-paste AI prompt (robust, works with any transcript)

“You are my voice-note organizer. Today’s date is {{YYYY-MM-DD}}. You’ll receive a transcript and must return only these sections as clean bullet points:
1) Title: a 6–8 word title.
2) Summary: 2–3 sentences in plain English.
3) Type: choose one — Idea, Meeting, Decision, Research, Personal.
4) Tags: choose up to 5 from this controlled list only: [Pilot, Pricing, Sales, Marketing, Product, Client, FollowUp, Research, Idea, Meeting]. If a client/company is mentioned, replace ‘Client’ with ‘Client’ plus the name (e.g., Client: GreenCo).
5) Actions: 0–3 items using this exact format — Action — Owner — Due date (YYYY-MM-DD). If no owner is stated, use ‘Me’. Set realistic due dates based on context.
6) Priority: High/Medium/Low with one sentence of reasoning.
7) If there’s a specific date or time mentioned, add: Calendar suggestion — Title — Date — Time (optional) — Duration.
If the transcript is vague, include one clarifying question at the end.
Keep language simple.”

Insider trick (big payoff): add field labels to your header. Saying “Project: … Keywords: … Decision: …” makes AI extraction 20–30% cleaner. Those colons are mini-beacons for models.

Worked example

Spoken: “2025-11-22. Project: GreenCo proposal. Keywords: pricing, pilot. Need updated pilot pricing and case study. Ask Mark for conversion benchmarks and set start date in two weeks.”

• Title: GreenCo pilot pricing and next steps
• Summary: We need updated pricing and a relevant case study for the GreenCo pilot. Mark should supply conversion benchmarks. Target start in two weeks pending data and pricing confirmation.
• Type: Decision
• Tags: Pilot, Pricing, Client: GreenCo, FollowUp, Product
• Actions:
  • Request conversion benchmarks — Mark — 2025-11-25
  • Draft pilot pricing update — Me — 2025-11-27
  • Confirm pilot start date — Me — 2025-12-06
• Priority: High — client timeline depends on pricing and benchmarks.
• Calendar suggestion: GreenCo pilot kickoff — 2025-12-06 — 30m

When you’re ready, automate (keep it simple first)

1. Create a “Voice Inbox” folder in Drive/Dropbox.
2. Zapier/Make flow:
   1. Trigger: new audio file in Voice Inbox.
   2. Action: transcribe via your chosen service.
   3. Action: send transcript to the prompt above.
   4. Action: create a Notion page (fields: Title, Date, Summary, Tags, Priority, Actions, Transcript, Audio URL).
   5. Optional: email yourself a daily digest of new actions.

What “good” looks like after a week

• Every note has a clear title and 3-sentence summary.
• Tags come only from your list (plus Client: Name when relevant).
• At least 60% of notes include one action with an owner and due date.
• You can find any note in under 30 seconds by tag or title.

Common mistakes and quick fixes

• Long rambles (10+ minutes): cap notes at 3 minutes or split into parts; accuracy and actions improve.
• Tag bloat: stick to 10 core tags for the first month; review the first 50 notes and merge similar tags.
• Mixed personal/work: add a Type field (Personal vs Work) so filters stay clean.
• Inconsistent owners: default to “Me” unless you name someone; change later if needed.
• Privacy worries: keep sensitive audio local or use services with clear privacy controls; encrypt storage if required.

Action plan

1. Today (20 minutes): write your 10-tag list, practice the header script, record three 60–90s notes, run the prompt, and file them.
2. This week (45–60 minutes): create the “Voice Inbox” folder, build one automation from audio → transcript → AI → note, and turn on a daily action digest.
3. Next week (30 minutes): review 30 notes, merge/rename tags, refine the prompt, and lock the workflow.

Closing thought: the magic isn’t the app — it’s your repeatable loop. Header, tags, prompt, one store. Set it once, then let your ideas flow and your system do the filing.

Quick 5-minute win: paste the prompt below into your image tool, generate 6 tiles, then preview a 3×3 repeat. If two tiles look clean at arm’s length, you’ve got a usable starter pattern.

Nice call on treating AI like a funnel and using the 3×3 seam gate — that saves hours. Here’s a compact, practical follow-up you can use right away to turn those keepers into production-ready files.

What you’ll need

• Target use & tile size (concept: 900–1200 px; print-ready: tile exported at 300 dpi sized to your cm/inch target).
• 3 palette anchors (2 fixed + 1 tweakable) in hex codes.
• AI image tool that exports high-res PNG with transparent background.
• Basic editor: Photoshop, GIMP, Affinity or a free online editor. Optional: Illustrator/Inkscape for vector work.
• Printer or lab for a 10 cm strike-off on your actual fabric.

Step-by-step (do this now)

1. Decide tile and palette. Note fabric type (cotton, silk, poly).
2. Run the prompt below for 6–8 tiles. Save full-res PNGs.
3. Seam test: place one tile in a 3×3 grid. If seams or halos appear, reject and move on.
4. For a keeper: offset by 50% in your editor, clone/heal seam lines, remove stray artifacts, then add a subtle grain layer (10–15% opacity) to mimic fabric texture.
5. Export a 10 cm swatch at 300 dpi and order a strike-off. Expect one color tweak after seeing fabric.

Copy-paste AI prompt (robust, repeat-safe)

“Create 6 seamless textile pattern tiles for apparel. Tile 1024×1024 px. Motif: small-scale floral. Color limit: 3 colors — #0A2342 (navy), #F5EFE6 (cream), #B45A3C (rust). Style: soft watercolor edges, moderate detail, 60/40 negative space. Edge rule: leave a 10% low-contrast margin at tile edges to prevent visible seams. Transparent background. No text or logos. Deliver evenly balanced repeat-ready tiles.”

Quick example

• Generate 6 tiles. Two are clean. Offset one 50%, fix seams (5–10 minutes), add 10% grain, export 10 cm @300 dpi, send to lab. Adjust navy 5% cooler after strike-off if needed.

Mistakes & fixes

• Visible seams: offset 50% and clone/heal the join.
• Muddy print: simplify shapes, thicken line weights, remove micro-texture.
• Color drift: lock two anchor colors; after first strike-off nudge only the third color 5–10%.
• Licensing: keep a record of prompts and edits; avoid branded motifs.

48-hour action plan

1. Hour 1: pick end use, tile size, 3 colors, collect 4 refs, run prompt for 6–8 tiles.
2. Hour 2: do the 3×3 seam gate, fix the top keeper, export a 10 cm swatch at 300 dpi, and send to lab.
3. Next day: review strike-off in daylight, make one small color tweak if needed, then lock final files.

Small experiments, fast proofs. Run the prompt, apply the seam gate, print a swatch — you’ll learn far more from a real fabric sample than another hour of screen tweaks. Try it and iterate.

Quick win: Use your phone + a simple AI prompt to turn receipt photos into categorized expenses in minutes — no tech degree needed.

Why this works: Modern OCR (text recognition) plus a small AI classifier can read totals, dates and merchants, then assign categories consistently. You get faster bookkeeping, fewer mistakes, and ready-to-use data for tax time.

What you’ll need

• Smartphone with camera (or a scanner)
• A scanning app or the phone camera that saves clear JPEG/PDF
• An AI tool that accepts text or images (many apps offer this) or a simple workflow using an OCR step + AI prompt
• A spreadsheet or accounting software where you want categorized outputs

Step-by-step (do this first)

1. Photograph the receipt on a flat surface with good light. Aim for readable text and no glare.
2. Use OCR to extract text (many scanner apps do this automatically). Save the raw text.
3. Send the OCR text to an AI with a prompt that asks for merchant, date, total, tax, line-items and a category from your list.
4. Review the AI output and export as CSV or paste into your spreadsheet/accounting tool.

Copy‑paste AI prompt (use as-is)

“Extract the following fields from this receipt text: merchant, date (YYYY-MM-DD), total amount, tax amount, and key line-items. Then assign one category from this list: Meals, Travel, Office Supplies, Utilities, Rent, Other. Output as JSON with keys: merchant, date, total, tax, items (array), category. If date or tax is not present, return null for that field.”

Worked example

• Photo -> OCR returns: “Joe’s Diner 2025-06-15 Subtotal $45.00 Tax $4.05 Total $49.05”
• AI output JSON -> merchant: “Joe’s Diner”, date: “2025-06-15”, total: 49.05, tax: 4.05, items: [“Lunch”], category: “Meals”
• Paste CSV row into your expense spreadsheet or import to accounting software.

Mistakes & fixes (quick checklist)

• Do take a clear photo on a plain background.
• Do keep a short, consistent category list to avoid confusion.
• Don’t rely on AI blindly — scan a few samples and check accuracy.
• Don’t include receipts with personal info you don’t want stored in cloud apps.

Action plan (next 30 minutes)

1. Take 5 receipts, photograph them.
2. Run OCR and paste the text into the AI prompt above.
3. Check results, correct any mis-categorized items, then import to your spreadsheet.

Start small, tune categories, and you’ll cut time spent on expense filing by 70% or more. Keep it simple and do one workflow until it becomes routine.

Hook: Yes — AI can often turn dense research into clear, student-friendly explanations, but “reliable” depends on the process you follow.

Context: AI is fast at simplifying language and creating teaching scaffolds. It’s great for first drafts and classroom-ready overviews. It struggles with nuance, math-heavy content, and citations unless you guide it carefully.

What you’ll need:

• PDF or plain text of the paper (or key excerpt).
• Target student level (age or year of study).
• Learning goal (conceptual, procedural, or critical thinking).
• Time to verify facts and equations manually.

Step-by-step (how to do it):

1. Pick a short excerpt (300–800 words). Long papers should be chunked into sections.
2. Decide the student level and desired output (summary, lesson, quiz, analogy).
3. Use a clear prompt (example below). Ask the AI to define technical terms and flag uncertain claims.
4. Run the prompt, then fact-check: verify key claims, numbers, and citations against the paper.
5. Iterate: ask for simpler language, or expand with examples and questions until it matches your learning goal.

Copy-paste AI prompt (use as-is; replace the excerpt placeholder):

Prompt: Read the following excerpt from a research paper. Rewrite it as a clear, student-friendly explanation for a college freshman (or 15–18 year old). Keep it to about 200–300 words using short sentences. Avoid jargon; if you use a technical term, define it in one simple sentence. Include: (1) a one-sentence summary of the main idea, (2) a simple analogy, and (3) three multiple-choice review questions with answers. At the end, list any claims you are not fully confident about and what to check in the original paper. Here is the excerpt: [PASTE_EXCERPT_HERE]

Worked example:

• Original sentence: “We observed a 30% reduction in error rate using the proposed algorithm.”
• Simplified for students: “Using the new method, mistakes dropped by about one-third. That means if there were 100 errors before, there are about 70 now.”

Mistakes & fixes (do / do not):

• Do chunk long papers; check numbers and formulas.
• Do ask the AI to list uncertain claims and show original sentences.
• Do not accept diagnostics or citations without manual verification.
• Do not rely on AI for novel math derivations or unstated assumptions.

Action plan — 3 quick wins:

1. Choose one paper and extract the abstract + one paragraph.
2. Run the prompt above and produce a 200–300 word student version.
3. Fact-check two key claims and create a short quiz for students.

Closing reminder: Use AI as a rapid assistant and editor — not a final authority. With a few checks and a clear prompt, you’ll get fast, usable explanations that students actually understand.

Nice point — the 5-minute loop is the practical habit that wins. The last 3–5 messages plus a quick privacy pass are exactly the right constraints to keep results reliable and low-risk.

Here’s a tight, actionable upgrade you can use immediately: a short context prefix, a clear prompt (copy-paste ready), a simple routine, and a small pilot to prove value.

What you’ll need

• Your email thread (last 3–5 messages) copied as plain text.
• An AI chat tool you already use (web or phone).
• A 30-second privacy checklist (remove attachments, redact account numbers/health/financials).

One-line context prefix (use first)

Project: [name] • Relationship: [client/colleague/vendor] • Priority: [low/medium/high]

Copy-paste AI prompt (primary)

Summarize the following email thread into three bullets: key asks, pending decisions, and deadlines. Then draft three reply options: (A) 20 words — quick acknowledgement + next step, (B) 60 words — confirm and ask one clarifying question, (C) 100 words — propose a solution and a clear call to action. Tone: concise, polite, professional. At the end, list any missing facts I must confirm before sending.

Step-by-step — 5-minute routine

1. Open the thread, copy the last 3–5 messages and add the one-line context prefix to the top.
2. Paste into your AI chat, paste the primary prompt above, and run it.
3. Quickly scan the summary and replies: correct names/dates, remove anything sensitive, choose a reply and send.

Prompt variants (copy these when needed)

• Polite decline: “Draft a short, respectful decline (30–50 words) that thanks them, gives a brief reason, and offers an alternative or next step to keep the relationship positive.”
• Request more info: “Write a one-paragraph reply asking three specific clarifying questions needed to decide.”
• Escalation: “Summarize this thread for leadership in 5 bullets: issue, impact, decisions needed, recommended next steps, and urgency.”

Common mistakes & fixes

• Too little context: AI misses the decision. Fix: add that one-line prefix or paste an earlier key message.
• Tone drift: AI too casual or blunt. Fix: add explicit tone instruction (“formal, warm, deferential”).
• Sensitive data: Don’t paste contracts/financials. Fix: summarize them in one sentence instead.

Quick 3-day action plan

1. Day 1: Run 10 threads through the routine and save the best prompt tweaks.
2. Day 2: Start a light automation pilot for low-risk threads — route AI outputs to drafts for human review.
3. Day 3: Measure minutes saved and a simple satisfaction score (1–5). Decide whether to scale.

Small steps, fast wins. Try one thread now — you’ll see the momentum.

— Jeff

Quick win: Use AI to clean messy brainstorm notes and produce a tightly grouped mind map you can import into any mapping tool or sketch on paper.

Context: Brainstorms are messy—ideas all over the page. AI can act as your editor and structure-builder. You don’t need to be technical. Here’s a simple, repeatable process.

What you’ll need

• Raw brainstorm (voice memo, rough notes, or a list of ideas).
• An AI chat tool that can rewrite and format text.
• A mind-mapping app (optional) that accepts Markdown/CSV/JSON, or just a blank page and a pen.

Step-by-step

1. Capture: Collect all ideas in one place (copy-paste notes or a transcript).
2. Clean: Ask AI to remove duplicates, merge near-same ideas, and label each as a main idea or subpoint.
3. Structure: Ask AI to create a hierarchical outline with 3–5 top-level branches and clear short labels.
4. Export/Build: Ask AI to output the outline as a Markdown mind map or simple CSV for import. Or draw the map from the outline.
5. Refine: Review the map, prune excess branches, and add priorities or next steps to nodes.

Example

Raw notes (short): “New product ideas, customer feedback, pricing, launch channels, team hires, timeline, content plan, partnerships.”

AI-produced Markdown mind map (example):

• New Product
  • Core features
  • Customer needs
  • Roadmap
• Go-to-Market
  • Channels
  • Launch plan
  • Content
• Pricing
  • Models
  • Testing
  • Discounts

Common mistakes & fixes

• Mistake: Too many top-level branches. Fix: Group into 3–5 main themes.
• Mistake: Long node text. Fix: Use 2–5 word labels and keep details as subnotes.
• Mistake: No priorities. Fix: Tag top 3 actions or mark nodes A/B/C.

Copy-paste AI prompt (use this exactly)

“Here are raw brainstorm notes: [paste your notes]. Please do the following: 1) Remove duplicates and group similar ideas. 2) Create a clear hierarchical mind map with 3–5 top-level branches and up to 4 child nodes each. 3) Output as a Markdown mind map using bullets and indentation. 4) Keep each label to 2–5 words. 5) Add one-sentence summary for each top-level branch.”

30-minute action plan

1. Spend 10 minutes collecting and pasting your notes into the AI prompt above.
2. Spend 10 minutes reviewing the AI output and trimming to 3–5 branches.
3. Spend 10 minutes drawing the map on paper or importing the Markdown to a mind-map tool and assigning next actions.

Try it once and you’ll see how quickly messy ideas become a clear plan. Small habit: run every weekly brainstorm through this process and you’ll always have a visual roadmap.

— Jeff