Create a Claude Project and add three files. Generic answers come from generic inputs, so this is where the quality is decided.
Read income.md, costs.md, skills.md and constraints.md.
Find where my existing skills already meet paid demand. For each candidate:
1. Name the specific service or offer, phrased the way a buyer would search for it, not as a job title.
2. Who exactly pays for this, and what triggers them to go looking.
3. What evidence would confirm or kill this in under a week, and exactly where I'd go to look for it (specific marketplaces, job boards, communities, search terms).
4. What I'd realistically charge, expressed as a range, plus what makes the difference between the low and high end.
5. The honest reason this might not work for me specifically, given my constraints.
Rank by how fast I could get a first real signal, not by how big the upside looks. Do not invent market data. Where you're inferring rather than knowing, label it clearly as an assumption I need to verify.The last two lines do the heavy lifting. Without them you get a confident-sounding market analysis assembled from nothing, which is the single most common way this kind of exercise wastes a month.
Take the top candidates and design exactly 3 experiments I can run in the next 30 days alongside my current commitments.
For each experiment give me:
- The single question it answers (one sentence, phrased so the answer can be yes or no).
- The smallest version that still produces a real signal. Assume I should not build anything before someone shows interest.
- Total hours required, and cash required. If either is above [your limit], redesign it smaller.
- The success threshold, decided NOW: what result means continue, and what result means stop. Be specific and numeric.
- The first three actions, each small enough to do in one sitting.
Make the three experiments genuinely different from each other, so I learn three separate things rather than testing one idea three ways.Ask Claude to generate a tracker you'll actually fill in. Small and boring beats elaborate.
Build me a simple tracker for these 3 experiments as a markdown table I can paste into Notion. Columns: experiment, the question it answers, success threshold, hours spent, cash spent, what actually happened, verdict (continue / adjust / stop), and what I learned that I didn't know before.
Then write me a short weekly check-in prompt I can paste back to you every Sunday with my results, so you can update the playbook rather than starting over.A playbook written once is just a document. What makes it worth anything is that it gets rewritten by contact with reality.
Week [n] results: [paste what happened, including the boring and the disappointing bits].
1. For each experiment, is the evidence pointing to continue, adjust or stop against the threshold we set? Hold me to the original threshold, don't move it for me.
2. What did this week teach us that changes an assumption in the playbook?
3. What's the single highest-information thing I could do next week? Highest information, not highest effort.
4. What should I stop doing?The squeeze in the reel is real and worth stating with actual numbers rather than vibes. According to the Australian Institute of Health and Welfare, Sydney's median house price went from around $680,000 in the first quarter of 2014 to around $1.4 million at the end of 2024, roughly doubling in a decade. Over a longer window, housing costs in Sydney rose from 13.4% of disposable income in 1994 to 22.2% in 2024 to 2025.
That second number is the one that tells the story. If incomes had kept pace with prices, the share of income going to housing would have stayed flat. It didn't, and that gap is what people are actually reacting to when they reach for a new worldview. Building a playbook is just the less dramatic response.