All guides
ChatGPTPromptsProductivityGuide8 min read29 Sep 2026

100 ChatGPT prompt shortcuts

From Adam

Hey, it's Adam. These are one-word shortcuts that change how ChatGPT answers. Type /tldr and you get the short version. Type /steelman and it argues the side you disagree with. Type /quizme and it turns into a tutor. There are 100 here, in ten groups, plus the trick that makes them work without explaining them every time.

how they work

How they work 🧠

These aren't hidden features. Typing a slash doesn't unlock anything. ChatGPT reads language, and a short code with a clear meaning is a fast way to tell it what you want.

That gives you two ways to use them:

1. Paste the code with its meaning. That always works:

Always works
/steelman — build the strongest version of the opposing argument.My view: [your position]

2. Teach it the codes once, then use the code on its own. Go to Settings → Personalization → Custom instructions, or a Project's instructions, and paste the 20 to 30 you use most, one per line, like this:

Custom instructions setup
When I start a message with one of these codes, follow it:/tldr = the shortest useful summary, 3 lines max/steelman = the strongest version of the opposing argument/quizme = quiz me one question at a time and wait for my answer

After that, /tldr followed by your text is all you need to type. Custom instructions allow up to 5,000 characters, so the full 100 won't fit — but your favourites will. To use all 100, put them in a Project's instructions.

the 100 shortcuts

The 100 shortcuts 🗂

Type the code, add your text or topic, and send. Codes stack: two or three in one message combine.

✍️ Writing & tone

01/human
02/rewrite
03/tighten
04/expand
05/plain
06/formal
07/casual
08/punchy
09/warm
10/proofread

💡 Explain it to me

11/eli5
12/eli12
13/expert
14/analogy
15/why
16/firstprinciples
17/deeper
18/examples
19/jargon
20/mentalmodel

📋 Format it

21/tldr
22/bullets
23/table
24/checklist
25/steps
26/outline
27/keypoints
28/onepager
29/faq
30/timeline

🎓 Learn & study

31/quizme
32/flashcards
33/feynman
34/socratic
35/teachme
36/misconceptions
37/studyplan
38/explainback
39/mockexam
40/recall

🚀 Ideas

41/ideas10
42/alt3
43/hooks10
44/scamper
45/remix
46/wild
47/angles
48/names
49/titles
50/worstidea

⚖️ Decide

51/proscons
52/swot
53/compare
54/decide
55/pareto
56/risks
57/secondorder
58/premortem
59/regret
60/nextstep

🥊 Challenge me

61/assumptions
62/counter
63/steelman
64/devilsadvocate
65/redteam
66/blindspots
67/factcheck
68/sources
69/confidence
70/whatsmissing

📱 Content & social

71/script
72/thread
73/caption
74/carousel
75/hookrewrite
76/cta
77/repurpose
78/newsletter
79/linkedin
80/storymode

💼 Work & business

81/email
82/reply
83/meetingnotes
84/agenda
85/proposal
86/pitch
87/persona
88/pricing
89/okr
90/sop

🛠 Utility & code

91/translate
92/convert
93/compress
94/reframe
95/coach
96/critique
97/betterprompt
98/debug
99/explaincode
100/optimize
three combos to try first

Three combos to try first 🔑

Making a decision
/assumptions then /premortem then /nextstepHere's my situation: [describe the decision]
Making your writing better
/human then /tighten. Keep my meaning and personality:[paste your text]
Actually learning something
/quizme on [topic]. One question at a time — wait for myanswer, tell me what I got right or wrong, then ask the next.
make the code work harder

Make the code work harder 💪

A code gets you started. The context makes the answer good. Add three things:

For /factcheck, ask it to search the web. For anything predictive, give it your assumptions and ask for several possible outcomes, not one.

honest bit

The honest bit ⚠️

These aren't "secret codes." There's no hidden feature and nothing to unlock. They're shorthand, and they work because ChatGPT reads language. The meaning is what does the work, which is why the "teach it once" trick above matters.

They're interpreted, not fixed. The same code twice gives you similar answers, not identical ones. If an answer misses, add a line of context rather than switching codes.

They work in other AIs too. Claude, Gemini and anything else that reads instructions.