3. Problem Selection & Expert Depth

You now have 5 candidate problems. You are going to pick one — not the flashiest one, not the most popular one. You are going to pick the one that is specific enough to show deep domain knowledge and that you can build a meaningful MVP of in 3–4 weeks.

Phase 3: Problem Selection & Expert Depth (quick-start)

This is the fast, do-it-now version of the problem selection step. Follow these six steps and you will pick the one problem your project solves, then become the person who understands it best. Use the attached doc for the full process behind each step.

The core idea before you start: this is where you prove you understand a real problem better than the average applicant. Two moves, in order. First, pick the right problem. Then out-research everyone on it.

Step 1: Pull your candidate problems

Open your Phase 2 Signal Log and group the recurring signals into three to five real problems your target companies share. Lean toward problems that showed up at multiple companies, because solving one of those makes your project relevant to a whole segment at once.

Step 2: Judge each problem against the bar

A good problem to build for is:

• Real and specific, not a vague theme. "Companies use AI" is a theme. "Support teams can't find answers buried in their internal docs" is a problem.

• Shared across several of your target companies.

• Buildable in one to two months. If it needs a year, scope it down.

• Revealing of domain understanding, so it shows you get the industry.

• Genuinely interesting to you, because you will live in it for weeks.

Step 3: Pick the un-obvious one, then commit

Do not grab the flashiest trend. The second, third, or fourth problem on your list is often the better pick: less saturated, more buildable, and it proves deeper industry understanding. Remember the bill of lading, an unglamorous, still-on-paper document that pulled serious attention precisely because knowing about it signaled real domain understanding.

Score your candidates against the bar above, pick the one that wins on buildability and domain depth, note a runner-up, and commit. A good problem you execute beats a perfect one you never start.

Gut-check before you lock it: Can you describe it in one sentence? Name three companies that have it? Picture the MVP? Ship it in one to two months? If yes to all, go.

Step 4: Become the expert (one to two week research camp)

Go deep on that one problem. Answer:

• Why does this problem exist, and why has no one solved it well yet?

• What solutions already exist, and where do they fall short?

• How expensive are the current approaches?

• What would an ideal solution look like?

Sources: sign up for and actually use competitor products, read their docs and pricing, dig through industry reports, forums, and communities, and study the companies' engineering blogs and job posts. Get fluent in the vocabulary insiders use, so everything you build and say lands as credible.

Step 5: Talk to real people

This is the highest-leverage move, and it is where your silent network pays off. Message the people you connected with in Phase 2, framed correctly. You are not asking for a job or a referral. You are building this with or without them and you want their expertise.

• Reference how you are connected or something specific about their company.

• "I'm building [X] to tackle [the problem]. I saw [their CEO's talk / your team's work]."

• "Would you be open to 15 minutes to share how you and your team think about [the problem]? Building this with or without help, I just want to learn from people who actually work in it."

• Keep it short and genuine. Do not ask for a job.

This makes your product far better, and it plants a seed. Months later when you apply, they remember you as the person who built that thing and talked to them about it.

Step 6: Write a research brief

Capture it all in a short brief: the problem in your words, why it exists, current solutions and their gaps, what you will do differently, key facts, and the best quotes from your conversations. This becomes the backbone of your MVP and your launch story, and your ammunition in interviews. Do not keep it in your head.

What you should have when you are done

• One problem you can state in a single sentence.

• At least three target companies that clearly have it.

• A research brief covering why it exists, current solutions and gaps, and what an ideal solution looks like.

• Real insights from a few conversations, plus warm connections for later.

• A rough idea of what your MVP needs to do.

Do not move on until you can state your problem in one sentence and explain, from real understanding, why it matters and why no one has solved it well.

Avoid these

• Chasing the flashiest trend. It is the most saturated and least differentiating.

• Picking a problem too big to build. Shrink it until it fits one to two months.

• Choosing a vague theme instead of a specific problem.

• Skipping the expert-depth work. Shallow understanding shows up everywhere.

• Researching entirely alone. You miss the best insights and the referral seeds.

• Framing outreach as "I want a job." Ask for expertise, not employment.

• Never committing. Endless research is just avoiding the build.

Go deeper: the full doc

This is the quick-start. The attached document walks through the full scorecard, the complete research-camp question set, the exact outreach approach, and how to build your research brief. Use it whenever you want the depth.