AI for Everyone
A practical, tool-agnostic course for using generative AI well: better prompts, examples, agents, media, documents, research, data, code, verification, privacy, and reusable workflows.
495 cards14 lessons14 video lessonsFree
Already have FlashFeed?Open this pack
Get this pack on your phone
Scan with your phone's camera.
flashfeedlearningapp.com/packs/ai-for-everyone
Part of the Learn AI collection.
14 lessons.
Each lesson has a study guide to read and objectives you can test yourself on. Its cards and games show up in your feed. Try one of its cards right here. Play a few of its games here too. Every lesson has a short video.
-
Pexels · Close-up of a hand writing in a notebook with a checklist for effective task management. ↗01A Good Prompt Has Parts
Name the Task · Give Relevant Context · Set Constraints and Format · Define Success
Lesson 1 · 10 cards1 / 10Which opening is most actionable?
“Compare” names a visible operation and identifies the supplied choices.
Lesson 1 has 40 cards. Keep going in the free app.
Question
Start With a Verb and an Outcome On YouTube ↗Transcript
"Help with my trip." That prompt leaves the system guessing. Ideas? An itinerary? A budget? Packing advice?
A useful prompt starts with a verb and an outcome. "Draft a two-day walking itinerary" makes the job visible.
Before sending, underline the action and circle the deliverable. If either is missing, add it.
For an unfamiliar task, ask the system to restate the assignment first. Mismatched assumptions show up while they are still cheap to fix.
-
Unsplash · Andrew Dunstan · Unsplash License ↗02Learn by Showing
Zero-Shot Requests · One-Shot Examples · Few-Shot Boundaries · Iterate with Evidence
Lesson 2 · 10 cards1 / 10What can a zero-shot prompt include?
The term concerns demonstrations, not the amount of instruction.
Lesson 2 has 40 cards. Keep going in the free app.
Question
The First Prompt Is a Probe On YouTube ↗Transcript
Hand a model your notes with one instruction: turn these into bullets. No example of the answer.
That is a zero-shot prompt: instructions without a worked example. Context can still be detailed.
Treat the first result as a probe. Its failures are information. If the format drifts, add an example, not vague adjectives.
If the model understands and errors cost little, keep the simple prompt. The smallest prompt that reliably works is easier to maintain.
-
Pexels · Elderly woman reviewing a manuscript with a typewriter at a wooden desk. ↗03Roles, Reasons, and Control
Roles Are Lenses · Ask for Visible Reasons · Break Work into Stages · Surface Assumptions and Uncertainty
Lesson 3 · 10 cards1 / 10What can a role prompt reliably provide?
A role is a behavioral cue, not verified expertise.
Lesson 3 has 40 cards. Keep going in the free app.
Question
The Doctor Prompt On YouTube ↗Transcript
Ask an artificial intelligence model to act as a doctor. Does it become one?
No. A persona is not a credential: no medical training, no accountability, no access to a patient's record.
Research on factual questions found that personas do not reliably improve objective knowledge. Their effects vary by model and question.
Put the task, evidence and rubric first. Use a role only when it changes observable behavior you want.
-
Pexels · panumas nikhomkhai · Pexels License ↗04From Chatbots to Agents
The Agent Loop · Tools, State, and Permissions · Human Approval Checkpoints · Agent Failure Modes
Lesson 4 · 10 cards1 / 10What most distinguishes an agent?
Agency comes from a loop that selects and executes actions.
Lesson 4 has 40 cards. Keep going in the free app.
Question
Data, Not Orders On YouTube ↗Transcript
A webpage may contain instructions meant to redirect an artificial intelligence agent. An email or a document may too.
It's called prompt injection: untrusted content designed to override or redirect what the agent was told.
So treat retrieved content as data, not as orders to follow.
Test an agent with hostile instructions planted in what it reads, before it is used more widely.
-
Wikimedia Commons · Mohamed Abdiwali Yusuf · CC BY-SA 4.0 ↗05Make with Multiple Media
Prompt an Image · Choose the Canvas · Plan Video as Shots · Work with Audio
Lesson 5 · 10 cards1 / 10What should an image prompt name first?
The required visual content should lead the brief.
Lesson 5 has 40 cards. Keep going in the free app.
Question
Not a Neutral Crop On YouTube ↗Transcript
A product sits centered in a square image. In a wide banner, that product may need to move to leave space for a headline.
Aspect ratio is width divided by height. Changing the ratio is not neutral cropping.
Converting one ratio to another may cut off faces, hands, products, or context.
So choose the canvas for where the image will appear. Generate it for that ratio when possible, then preview it in the real layout.
- Take the lessons with you. Free, no account.Download on the App Store ↗Get it on Google Play ↗
-
Pexels · Stack of colorful sticky notes with a 'To Do' note on top, isolated on white. ↗06Writing and Documents
Separate Drafting from Editing · Summarize Long Material · Transform Tone and Format · Ask Questions of Documents
Lesson 6 · 10 cards1 / 10What should a rewrite prompt protect?
Preservation rules keep fluent revision from changing substance.
Lesson 6 has 40 cards. Keep going in the free app.
Question
The Buried Detail On YouTube ↗Transcript
Bury a decisive detail among many documents. Ask an artificial intelligence model for a summary. That detail can be missed.
Research finds that a model's performance can vary with where a detail appears, including drops in the middle.
So summarize each section first. Then build the final summary from those traceable notes.
Ask for page or section references, so a reader can check the summary against the original.
-
Pexels · A young man in a library studying with books and taking notes under a desk lamp. ↗07Research, Learning, and Decisions
Plan Research Before Answering · Use AI as a Practice Partner · Compare Options without Outsourcing Choice · Six Thinking Moves
Lesson 7 · 10 cards1 / 10What should happen to an AI citation?
The source and its support for the claim both need checking.
Lesson 7 has 42 cards. Keep going in the free app.
Question
The Page Behind the Link On YouTube ↗Transcript
An answer from artificial intelligence cites a source. The link is real. The page behind it may still not support the claim.
Generated citations may be invented. They may be misquoted. Or they may cover a nearby point, not the claim.
Open every source that matters. Confirm who wrote it and when. Check that the cited passage actually supports the claim.
Until then, treat a summary of a source as a lead.
-
Unsplash · KOBU Agency · Unsplash License ↗08Data, Spreadsheets, and Code
Describe the Data First · Build Formulas and Charts · Use a Coding Cycle · Test and Secure Generated Code
Lesson 8 · 10 cards1 / 10What should be stated about each row?
The unit of observation determines valid aggregation.
Lesson 8 has 40 cards. Keep going in the free app.
Question
One Signal On YouTube ↗Transcript
Ask an artificial intelligence model to check its own code for security. Is that enough?
The model's own review is one signal, not an independent check.
Generated code can contain ordinary bugs and security holes. Treat it like code from any untrusted source, no matter how fluent it seems.
Try uncertain code in a sandbox, an isolated space that limits a program's effects. Ask for tests, then run them independently.
-
Pexels · RDNE Stock project · Pexels License ↗09Verify and Protect
Verify Claims and Citations · Minimize Sensitive Data · Resist Injection and Impersonation · Know the High-Stakes Boundary
Lesson 9 · 10 cards1 / 10What must be checked about a citation?
A source must be both genuine and aligned with the statement.
Lesson 9 has 40 cards. Keep going in the free app.
Question
A Familiar Voice On YouTube ↗Transcript
A familiar voice says there is an emergency and asks for money, fast. Is it really them?
A familiar voice is not proof. Voice cloning can imitate a person well enough to support a scam like this.
Urgency, secrecy, payment demands and switching channels are warning signs, no matter how familiar the voice sounds.
So pause and verify. Use a known phone number or an agreed code, not contact details from the message.
-
Pexels · From above of crop anonymous skillful male in casual clothes choosing screwdriver from toolbox in daylight ↗10Build Your Personal AI Workflow
Choose the Tool for the Task · Turn Success into a Template · Use a Small Evaluation Set · Practice, Then Go Deeper
Lesson 10 · 10 cards1 / 10What should lead tool selection?
Requirements and representative cases reveal fit.
Lesson 10 has 39 cards. Keep going in the free app.
Question
Count the Corrections On YouTube ↗Transcript
Does artificial intelligence really save you time? Measure corrections, not just speed.
Time saved before review can disappear once you count the errors, rewriting, privacy cleanup and rework.
Run a 2-week practice loop on one recurring task you can safely undo, like outlining.
Record the original time, the assisted time and the corrections required.
Keep the workflow only when evidence justifies the habit.
-
Unsplash · Sergio Vilches ↗11A Map of AI
Narrow Systems, AGI, and Superintelligence · Training and Inference · Six Milestones in AI
Lesson 11 · 4 cards1 / 4Why is a superhuman Go program still a bounded system?
Exceptional performance in one domain does not imply flexible competence across unrelated domains.
Lesson 11 has 24 cards. Keep going in the free app.
Question
Asking Is Not Training On YouTube ↗Transcript
Ask an artificial intelligence model for a summary. Does that prompt instantly retrain the model?
A model changes during training. Over many examples, it makes outputs, gets a learning signal and adjusts its weights.
Asking for that summary is inference. A trained model takes a new input and produces an output.
The prompt can change what the model considers in that chat, but it does not ordinarily rewrite the weights.
-
Pexels · Laptop displaying financial chart alongside documents and credit cards on desk. ↗12Words You'll See Everywhere
Models, Data, and Parameters · Hallucination, Bias, and Grounding · Six Key AI Concepts
Lesson 12 · 4 cards1 / 4What is the main role of a training dataset?
Examples provide the signal from which the training process shapes model behavior.
Lesson 12 has 24 cards. Keep going in the free app.
Question
The Missing Foundation On YouTube ↗Transcript
An artificial intelligence model can write a fluent answer that cites a study. The study may not exist.
That is called a hallucination. It sounds plausible, but it is unsupported or false.
A model may also invent a quote, or combine details from different events. The fluent surface can hide the missing foundation.
So for a factual claim, open the source. Confirm that it supports the sentence.
-
Unsplash · Vitaly Gariev · Unsplash License ↗13What AI Can—and Cannot—Do
Strengths and the Reliability Gap · Energy, Water, and Physical Scale · A Regulation Snapshot
Lesson 13 · 6 cards1 / 6Which task is the best beginner pilot?
A reversible draft creates value while keeping the consequence of an error low.
Lesson 13 has 19 cards. Keep going in the free app.
Question
Weightless on a Screen On YouTube ↗Transcript
Artificial intelligence feels weightless on a screen. But AI runs on chips, servers, cooling and power plants.
The International Energy Agency projects about 945 terawatt-hours of electricity for the world's data centers in 2030.
That would be a little more than double the 2024 level.
AI is the largest driver of that growth. But the figure covers all data centers, not AI alone.
-
Wikimedia Commons · Kwamitsey · CC BY 4.0 ↗14Where Smart People Disagree
Will AI Take Your Job? · Catastrophic and Existential Risk · Copyright and Training Data · Timelines to AGI
Lesson 14 · 8 cards1 / 8What is usually the better first unit for analyzing automation?
Occupations combine tasks with very different potential for assistance or substitution.
Lesson 14 has 27 cards. Keep going in the free app.
Question
A Forecast, Not a Schedule On YouTube ↗Transcript
Ask researchers in artificial intelligence when unaided machines will beat humans at every task.
One survey did. Combined, the answers gave a 50% chance by 2047, if scientific progress went on without disruption.
But individual answers varied widely.
The same survey put all occupations becoming fully automatable much later: a different finish line.
So 2047 is a conditional forecast, not a schedule.
What you'll be able to do 14 objectives, from the pack's lessons
- Build a prompt with a clear task, relevant context, useful constraints, and a testable output.
- Choose zero-, one-, or few-shot prompting and improve examples through deliberate iteration.
- Use roles, decomposition, assumptions, and visible checks without mistaking performance cues for expertise.
- Design bounded agent workflows with tools, permissions, checkpoints, and explicit stopping conditions.
- Prompt and review image, video, and audio systems with attention to composition, continuity, consent, and fit.
- Use AI to draft, revise, summarize, transform, and inspect documents while preserving meaning and provenance.
- Use AI to structure inquiry, practice learning, compare options, and perform distinct thinking moves without outsourcing judgment.
- Use AI to inspect structured data and code through schemas, tests, security checks, and reversible changes.
- Verify material claims, minimize sensitive data, resist manipulation, and keep high-stakes decisions with qualified people.
- Choose tools deliberately, turn successful prompts into templates, evaluate them on real cases, and route deeper learning through the collection.
- Distinguish present AI systems, hypothetical intelligence categories, and the training and inference phases.
- Use collection-wide AI vocabulary to distinguish components, failure modes, interfaces, and agents.
- Choose reviewable AI uses, interpret footprint claims, and recognize jurisdiction-specific governance.
- Explain why credible estimates differ on jobs, catastrophic risk, copyright, and AGI timelines.
Study guide.
The same study guide as in the app, one page for each lesson, with its key facts and a self-test.
- 01A Good Prompt Has Parts
- 02Learn by Showing
- 03Roles, Reasons, and Control
- 04From Chatbots to Agents
- 05Make with Multiple Media
- 06Writing and Documents
- 07Research, Learning, and Decisions
- 08Data, Spreadsheets, and Code
- 09Verify and Protect
- 10Build Your Personal AI Workflow
- 11A Map of AI
- 12Words You'll See Everywhere
- 13What AI Can—and Cannot—Do
- 14Where Smart People Disagree
Built from 24 sources.
363 of its 495 cards cite a source. These are the eight it cites most.
- Prompt design strategies ↗
- Generative AI Profile ↗
- Rethinking the Role of Demonstrations ↗
- A practical guide to building agents ↗
- AI Risk Management Framework Playbook ↗
- Role-Playing Does Not Improve Factual Accuracy ↗
- Image generation ↗
- Secure Software Development Framework ↗
See all 24 sources
- Prompt design strategies ↗
- Generative AI Profile ↗
- Rethinking the Role of Demonstrations ↗
- A practical guide to building agents ↗
- AI Risk Management Framework Playbook ↗
- Role-Playing Does Not Improve Factual Accuracy ↗
- Image generation ↗
- Secure Software Development Framework ↗
- Configure image aspect ratio ↗
- AI Act framework ↗
- Energy and AI (2025) ↗
- Artificial Intelligence coined at Dartmouth ↗
- XPT report ↗
- Generate videos with Veo ↗
- Speaker diarization ↗
- AlphaGo ↗
- Thousands of AI Authors ↗
- Deep Blue ↗
- AI and the Future of Work ↗
- ImageNet Classification ↗
- Generative AI Profile ↗
- Introducing ChatGPT ↗
- Attention Is All You Need ↗
- Sensitivity analysis ↗
Hover over a photo, or tap its i, to see where it came from. All 14:
- Lesson 1: Pexels · Close-up of a hand writing in a notebook with a checklist for effective task management. ↗
- Lesson 2: Unsplash · Andrew Dunstan · Unsplash License ↗
- Lesson 3: Pexels · Elderly woman reviewing a manuscript with a typewriter at a wooden desk. ↗
- Lesson 4: Pexels · panumas nikhomkhai · Pexels License ↗
- Lesson 5: Wikimedia Commons · Mohamed Abdiwali Yusuf · CC BY-SA 4.0 ↗
- Lesson 6: Pexels · Stack of colorful sticky notes with a 'To Do' note on top, isolated on white. ↗
- Lesson 7: Pexels · A young man in a library studying with books and taking notes under a desk lamp. ↗
- Lesson 8: Unsplash · KOBU Agency · Unsplash License ↗
- Lesson 9: Pexels · RDNE Stock project · Pexels License ↗
- Lesson 10: Pexels · From above of crop anonymous skillful male in casual clothes choosing screwdriver from toolbox in daylight ↗
- Lesson 11: Unsplash · Sergio Vilches ↗
- Lesson 12: Pexels · Laptop displaying financial chart alongside documents and credit cards on desk. ↗
- Lesson 13: Unsplash · Vitaly Gariev · Unsplash License ↗
- Lesson 14: Wikimedia Commons · Kwamitsey · CC BY 4.0 ↗
- Cards
- 495
- Lessons
- 14
- Version
- 0.1.19
- Language
- English
The whole pack is one file. Read the pack file ↗
FlashFeed is on the App Store and Google Play.
FlashFeed is free on iPhone and Android, and needs no account.



























































