Skip to content

ChatGPT course · Included in Osum membership

Get a straight answer out of ChatGPT.

You ask a real question and get your own opinion back, polished. Ask it in third person and the answer changes. Learn the Set Task: the task, the neutral ask and the check that make answers usable.

  • Lesson 1: your first two-way test, run and checked in 14 minutes
  • By lesson 12: a weekly task deck of saved, checked asks
  • No code. The free plan runs the whole course
Get my ChatGPT plan About 2 minutes. Lesson 1 is a test you can run on a real decision the same day.
Would you approve this advice? Try it first ↓
Lesson 1: Get a straight answer in 14 minutes
12 hands-on lessons36 exercises1 certificate for LinkedIn30 courses in one membership

Sound familiar?

You asked for an opinion. You got a mirror.

“ChatGPT keeps agreeing with whatever you say instead of giving a straight-up honest answer.”

ChatGPT user

“it forgets its prompts within 3 replies.”

ChatGPT user

“I’ve seen so many influencers sharing “prompt hacks” to make it sound less agreeable, but even after trying those, it still feels too polite or neutral sometimes.”

ChatGPT user

A better prompt hack rarely fixes it. An ask that hides your hand usually does. That’s the Set Task, and this course teaches it.

What you’ll be able to do

Ask so it answers straight. Check before you use.

01

Run the two-way test

One real decision, asked as written and in third person. The flattery shows itself in minutes, and the better ask becomes your first saved card.

02

Write the task, not the question

A role, the constraints and a done-when line on one card. Answers come back in a shape you can check, so the rewording stops.

03

Give each job one home

The card on top, this week beside last week. A dropped rule shows in seconds, because two runs sit side by side.

04

Hide your hand on big calls

Money, a client, your name on the result: the facts go in, the preference stays out, and both sides come before any recommendation.

05

Make a good ask repeat

A clean card goes on the calendar, then into a tested custom GPT. The weekly run starts without you deciding to start it.

06

Catch the confident wrong answer

Claims to sources, numbers re-asked once, one spot-check you run yourself. The invented statistic dies in the draft, not in the client’s inbox.

Try it first

Would you approve this advice?

You asked ChatGPT about a decision you owe this week: a rush job that pays double and eats your weekend, and you already told the client you are interested. Here is the question you sent and the reply. Take a minute and decide. It is the same call you make on every answer you use.

What you asked, and what is true

What you sentWhat it revealsWhat a check finds
“I’m thinking of taking it”Your lean, in the first sentenceThe reply starts from yes
“It pays double”A factNever weighed against the weekend cost
“I already told the client I’m interested”A constraint you set yourselfNever mentioned again
Your other client’s Friday deadlineThe fact that decides itAbsent from the reply
  1. Rule 1: Both sides appear before any recommendation.
  2. Rule 2: Claims carry a source or an example, or they are marked.
  3. Rule 3: Every constraint you set is addressed, one by one.

The reply you got

Taking on rush work shows ambition, and clients notice it. Rush clients often become long-term clients. Go for it!

Approve the advice, ask again, or use it with a fix?

See what a checked review finds
  • Ask again. The reply reads well, but it fails all three rules.
  • Rule 1 fails: there is no case against. A rush weekend has a cost, and the reply never names it.
  • Rule 2 fails: “rush clients often become long-term clients” arrives with no source and no example. Marked.
  • Rule 3 fails: the fact that decides it, your other client’s Friday deadline, never appears. The reply answered the question it was given, and your wording never posed that one.
  • What you can use: the re-ask in third person, “A freelancer friend was offered…”, which returns the deadline risk, the scope of “rush” and the weekend cost in one message.

A longer prompt would not have caught this. Changing whose decision it was did, and that move is the two-way test from the Set Task. Lesson 1, “Get a straight answer in 14 minutes”, has you run it on a real decision the same day. Get your plan and run it on your own chats.

Get my ChatGPT plan About 2 minutes. Lesson 1 is a test you can run on a real decision the same day.

The curriculum

The ChatGPT course: 12 lessons, one real project.

Osum’s ChatGPT course is 12 hands-on lessons, about 4 hours in all. Lesson 1 is a 14-minute two-way test you can run the same day. By lesson 12 you run a weekly task deck: three saved asks, checked before anything leaves. The free plan runs the whole course. You finish with a certificate you can add to LinkedIn.

Module 01

Ask your first straight question

You hold a two-way tested decision and your first task card, the opening card of the weekly task deck you finish in lesson 12. You also know what ChatGPT remembers about you and what it does not.

  1. Get a straight answer in 14 minutesRun the two-way test on one real decision: ask your question as written, re-ask it in third person, and compare before you decide.
  2. Write the task, not the questionTurn lesson 1’s decision into a task card: a role, the job, the constraints, the finished form and what must not happen.
  3. Know what it remembers about youMap what carries between chats on your plan, and restate the rules that must never fade inside the card itself.

Module 02

Set up work that repeats

You hold one home per repeating job, a neutral-ask habit for high-stakes questions, and the first weekly run on the calendar.

  1. Give each repeating job one homeMove one repeating job into a single home with its card attached, so the working version stays current and nothing lives in ten half-chats.
  2. Hide your hand when the stakes are realTurn the two-way test into a standing habit: on any high-stakes ask, strip the preference before the question goes in.
  3. Set the weekly runPut one card on a schedule so its run happens without you remembering, and read the result with the check attached.

Module 03

Hand it real work

You hold a saved run for a job you repeat, a files-and-sources habit for claims, and one red-teamed draft that survived the strongest case against it.

  1. Save a task as a custom GPTTurn a card that has run clean into a saved run with fixed instructions, and test it before you trust it.
  2. Give it the files and the sourcesRun one job on your own documents, and require every claim to name where it came from or say it cannot confirm.
  3. Make it argue with youRun one real draft through a red-team pass that attacks it, so the version you keep survives the strongest case against it.

Module 04

Check it and run it weekly

You hold a verification habit, a check that catches the flattering review, and your weekly task deck: three saved asks, run and checked in one sitting.

  1. Catch the confident wrong answerVerify any answer in minutes with the three-part check: claims with sources, numbers re-asked, and one spot-check you run yourself.
  2. Catch the flattering reviewRun the approval check from the course page on your own work: constraints addressed, argument present, claims sourced.
  3. Run your weekly task deckRun three saved asks in one sitting, check each result, and hold the weekly deck you keep using after the course.
Get my ChatGPT plan About 2 minutes. Lesson 1 is a test you can run on a real decision the same day.

Us vs. them

Why learn ChatGPT here

OpenAI’s academy teaches AI at work in general. Videos show you someone else’s build. This course takes you from one tested decision to a weekly task deck of saved, checked asks, with an exercise on your own work after every lesson.

FeatureOsum courseOpenAI AcademyYouTube walkthroughsVideo courses
One real project, built across every lesson✓✕✕Partly
Hands-on exercise on your own work after every lesson✓Partly✕Follow-along
A worked example to check each exercise against✓Partly✕✕
A check that catches flattering answers✓PartlyPartly✕
Certificate from a final check, one question per lesson✓Official badges on some courses✕Completion certificate
Every other Osum course included✓✕✕✕
Official, always-current factsLinks to them✓PartlyPartly
CostMembershipFreeFreePaid course

One membership

Learn ChatGPT. Get every AI course with it.

The same membership covers Codex for pull requests you would merge, Claude Cowork for finished reports from your own files and Midjourney for client-ready images. Each course has its own certificate.

Certificate

Finish it. Prove it.

Show a client, a manager or a hiring team that you can get checked work out of ChatGPT, not just chat. The certificate records a final check you passed, one question per lesson.

  • 12-question final check, one per lesson
  • Miss one? You see which lessons to review, then submit again
  • A public page you control, one click to add it to LinkedIn
  • Earn one for every course in the library

Questions

Before you start

How much does the ChatGPT course cost?

One Osum membership covers the ChatGPT course, its certificate and every other Osum course. It is a subscription. You see the price after your 2-minute plan, before you pay anything.

Do I need ChatGPT Plus to start the course?

No. Osum’s ChatGPT course runs on the free plan: unlimited text chats, with limits on uploads, images, voice and deep research (checked 28 September 2026). Plus adds Projects, scheduled tasks and custom GPTs; lessons 4, 6 and 7 practice them, and Plus was $20 a month, billed monthly, when we checked on 28 September 2026.

How long does the ChatGPT course take?

Osum’s ChatGPT course takes about 4 hours: 12 lessons, starting with a 14-minute two-way test you can run on a real decision the same day.

I’m not technical. Can I do the ChatGPT course?

Osum’s ChatGPT course needs no code: if you can type a message and open a document, you can do every exercise. Everything runs in the ChatGPT window on your own low-risk work.

I already wrote custom instructions. Why would this course work?

Osum’s ChatGPT course adds what saved instructions leave out: rules that travel inside each task, and a check on what comes back. Users report saved instructions fading within replies; the task card and the three-part check are the habits that hold. You finish with a weekly task deck of saved, checked asks.

Will this course make ChatGPT stop flattering me?

No course can change how the model answers. You learn to remove the pull instead: state facts without your preference, ask for both sides, re-ask the big ones in third person, and check claims against sources. The flattery stops mattering when the ask and the check are yours.

Is the ChatGPT course made by OpenAI?

No. Osum is independent and not affiliated with OpenAI, and the certificate is an Osum certificate. Every tool fact in the course links to OpenAI’s own pages, dated.

Keep re-asking ChatGPT the same question. Or set the task and get the answer straight.

The Set Task is four habits: set the task, hide your hand, give it a home and a clock, check before anything leaves. You get a two-way test in lesson 1, a weekly task deck by lesson 12, a certificate and every other Osum course. A few quick questions, then your plan.

Get my ChatGPT plan About 2 minutes. Lesson 1 is a test you can run on a real decision the same day.

Keep exploring

Related learning paths

Learn a useful skill. Put it to work.

Practical courses for AI tools and skills. Course and tool names are the trademarks of their respective owners; Osum is independent and not endorsed by them.