Lesson 16 of 25

The Seed of a New Idea (Brainstorming)

Brainstorming with an AI only works when you give it somewhere to push against. An open request - "suggest some names" - returns the most average answer in existence, because average is exactly what a prediction engine is built to produce. You get genuinely useful raw material by asking for volume, giving the model a specific brief to generate against, sorting the output into categories you chose, and banning the obvious answers by name.

Reach for it whenWhen you are at the blank-page stage - naming something, solving a problem you have been staring at too long, or hunting for a metaphor - and you need thirty options to react to rather than one polished answer.

Ask for more ideas than you need and sort them into categories you name yourself, because the categories are what stop the list collapsing into twenty variations of the same thought. Then run a second pass on the same list - have it mark which ideas are derivative, which are unworkable, and which three it would actually back - so the filtering happens in the chat rather than in your head. If you brainstorm for the same business repeatedly, the brief itself belongs in Custom Instructions.

Worth knowingThe first ten ideas on any list are usually the ones you would have thought of yourself. That is not a failure of prompting - it is what averaging across text looks like - so treat the top of the list as warm-up and look at items 11 to 20. Do not ask a reasoning model to "think step by step" before brainstorming either; it already deliberates, and the phrase buys you nothing here.
  • Ask for 20, expect to use 2.
  • Name the categories yourself - it stops repetition.
  • Ban the obvious answers explicitly, by name.
  • Second pass: have it critique and rank its own list.
  • Keep the standing brief in Custom Instructions.
Prospecting for a Brand Identity

Instead of

What should be the name for my new coffee shop in Kondapur?

Perplexity version

I am opening an upscale coffee shop in Kondapur, Hyderabad. Customers are IT professionals from the surrounding tech parks - mostly 25 to 40, used to good coffee, in and out quickly on weekdays and lingering on weekends. The vibe is modern and minimalist with a thread of traditional South Indian warmth running through it: filter coffee served properly alongside a pour-over bar.

## WHAT I NEED

Generate 20 brand names, sorted into these three groups:

1. **Modern and tech-adjacent** - names that would sit comfortably next to the companies my customers work for, without being cute about it.
2. **English blended with Telugu or Hyderabadi** - and use real words with real meanings, not transliterated nonsense. Tell me what each word means and how it is pronounced.
3. **Elegant and sophisticated** - short, speakable, no hint of a pun.

## RULES

- No "Brew", "Bean", "Grind", "Roast", "Cup", "Daily", "Co." or "& Co." Those words are on every coffee shop in India and I want the list to start past them.
- No names that are only a word with a vowel removed.
- Each name must survive being said out loud over a phone to a signboard painter.

## THEN, A SECOND PASS

After the list, do three things:

1. **Cull.** Name every entry you would delete and say why - too close to an existing chain, hard to spell, means something unfortunate in Telugu, Hindi or Urdu.
2. **Back three.** Pick the three you would actually open with, and for each give me the one-line story the name tells a customer standing on the pavement.
3. **Tell me what to check.** For your three, say plainly what you cannot verify - trademark status, whether the .in domain is free, whether a shop in Hyderabad already uses it. Do not guess at any of those.

If a name in group 2 only works for someone who speaks Telugu, say so. Half my customers will not.
Open Perplexity 1,831 characters
Finding Unconventional Solutions

Instead of

Solve the traffic problem in Hyderabad.

Perplexity version

The problem: severe congestion on the Outer Ring Road in Hyderabad during peak hours, roughly 8 to 11 AM and 5 to 8 PM, driven heavily by IT commuters moving between the western tech corridor and the rest of the city.

Brainstorm 15 unconventional interventions. Spread them across these four levers, so the list does not become fifteen apps:

1. **Technology** - sensing, routing, pricing, scheduling.
2. **Policy and regulation** - what the state or GHMC could mandate, tax or permit.
3. **Behavioural** - changing when and whether people travel, rather than how.
4. **Employer-side** - what the large tech campuses themselves could be made or paid to do. This lever is the most under-used and the most available, since a small number of employers control a large share of the traffic.

## OFF THE TABLE

Build more flyovers. Expand the metro. "Encourage carpooling" with no mechanism attached. More buses, stated generically. Widen the road. These are the five answers every report already contains and they are not what I am asking for.

## FOR EACH IDEA

Two or three sentences, then one line naming the single thing most likely to kill it - usually a political cost, a cost-recovery problem, or an enforcement problem. An idea with a named obstacle is useful; one without is a slogan.

## THEN

- **The three cheapest.** Which three could be tried for the least money and the least legislation, and what would a pilot on one stretch of the ORR look like?
- **The one that is actually hard.** Which idea on your list is the most likely to work and the least likely to happen, and what makes it politically impossible?
- **Where you are guessing.** Congestion pricing, commuter numbers, employer concentration - if you are describing how Hyderabad works from general knowledge rather than data, say so. I would rather have a flagged guess than a confident invention.

Do not quote traffic volumes, costs or travel-time figures as though they were sourced. If you need a number to make an idea legible, mark it as illustrative.
Open Perplexity 2,030 characters
Generating Creative Raw Material

Instead of

Write a blog post about the importance of saving for retirement.

Perplexity version

I am writing a blog post for young Indian professionals - late twenties, first or second job, earning in rupees, saving irregularly - about why starting retirement savings early matters far more than the amount you start with. I need the raw material, not the post.

Give me 10 analogies for compound interest, each grounded in something an Indian reader already has an intuition for. Draw from different wells so they are not all the same analogy wearing different clothes - for instance: cricket and the way a strike rate compounds across overs; a mango or banyan tree planted by a grandparent; the monsoon filling a tank one night at a time; a chit fund or a recurring deposit; dough rising; a family business passed down; a WhatsApp forward spreading.

## FOR EACH ANALOGY

1. **The analogy** in two or three sentences, written as I could say it out loud.
2. **The part it gets right** - specifically which property of compounding it captures: that time matters more than amount, that the growth is on the growth, that waiting five years is far costlier than it feels.
3. **Where it breaks** - every analogy misleads somewhere, and I would rather know than have a reader catch it. If the mango tree implies you can only plant once, say so.

## THEN

- **The best three**, ranked, for a 27-year-old in Hyderabad who has heard this lecture from a parent and tuned it out.
- **The one to avoid.** Which of your ten is most likely to be misread as a promise of guaranteed returns? That one is a liability in a post about money, and I would rather cut it than explain it.
- **One analogy for the opposite point** - the cost of delay rather than the reward of starting. Loss tends to land harder than gain, and I may want to open the post with it.

No numerical examples. I will build the maths separately, and I do not want invented return percentages in the metaphor.
Open Perplexity 1,867 characters