How I pick problems

45 ideas. Three areas. Two lenses.

I keep every idea in one inbox and write down the same things for each: who has the problem, what they do today, and the riskiest assumption. This page sorts all of them by what they are for, when their market is ready, and which human needs they serve. The placements are my judgment, not data.

My take on vertical AI

I researched 46 industries, one by one. Each looked different. Looked at together, the same five problems kept repeating, and only about ten of the 46 actually have a lead-generation problem. So vertical AI is not a chatbot per industry. It is software that joins the money spent to the outcome that matters, inside each industry's own system of record.

AThe join problemThe outcome lives in a system the ad tools can't see: the dental practice system, the gym's member system, the mortgage platform.

BCapacity caps revenueTechnicians, chairs, provider hours. More leads to a shop that can't staff them is waste that looks like success.

CThe calendar confoundsPromo weeks, tax season, term dates, weather. Benchmarks need a season-aligned week.

DThe wrong conversionSign-ups instead of members kept 90 days. Leads instead of funded loans. Agencies get paid on the first; businesses live on the second.

ESomeone owns the middleMarketplaces, delivery apps and lead aggregators. How much of the outcome did you rent rather than own?

The three areas

Area 1 · 14 ideasVerticalAI, martech and revenue tech for specific industries

Area 2 · 16 ideasConsumerKids, productivity, and helping people get good at AI or start something

Area 3 · 15 ideasAI and data infraContext, models, evals and trust for agents

Lens 1 · Time

What changes in one year, and in three

VerticalIn 1 year

Agencies in-house their reporting first. Private-equity roll-ups turn single shops into multi-location buyers. AI answers take clicks away from local media.

In 3 years

Agents plan and buy media. Vertical AI winners own the join between the ad stack and the system of record. Generic dashboards are free.

ConsumerIn 1 year

AI tutors are cheap and everywhere. Parents want judgment, not answers. Solo builders multiply.

In 3 years

Trust, memory and real human connection are the scarce things. Your personal context becomes the product.

AI and data infraIn 1 year

Evals, cost and context are the pains teams pay for now: did the change help, what did the answer cost, and what did the agent know.

In 3 years

Agent identity, authority and multi-agent correctness become standard infrastructure. Middle layers keep getting bought (OpenRouter, Hugging Face bid).

Now

The pain is here today

In 1 year

The market is forming

In 3 years

A bet on where things go

Vertical

Data-layer interop checker

Finds broken joins in a marketing stack

StackSignal

Company name to outreach brief from its tech stack

UTM link governance

Stops broken campaign tracking at the source

AI readiness grader

Pilot-to-production quiz for marketing teams

Category collision watcher

Spots when a neighbouring category moves into yours

Home services franchise economics

FDD Item 19 unit economics

Viao AI

Vertical AI for local businesses. 28 roofers, 0 sales

Bhasha

Indian-language customer messages for NBFCs (RBI mandate)

SaaSMatchup

Stack recommendations for B2B buyers

Agentic ad-ops audit trail

What an agent changed in a campaign, and why

MarTech agent-readiness scanner

Scores a platform for agent use

Creator ops interop agent

Connects the creator-marketing tool stack

CreatorStackSignal

StackSignal for creator-marketing tools

India influence platform

Regional athletes and creators meet brands

None yet

Consumer

Life lessons site

One interactive page per life lesson, for 10 to 14

AI Fellows

Free six-week fellowship with practitioner mentors

Threshold

Diaspora mentors for students in small-town India

Blackmere

Plague-doctor game about deciding with missing information

AI Bridge

Four-week paid-skill apprenticeship

First Light

Family habit tracker for a parent and child

Reflex

Practice prompts tied to what you read

Bengali literary project

Digitize and translate Bengali books

RescueTime intelligence brief

Planned vs actual time, weekly

Google Workspace add-ons

Small paid add-ons found through Marketplace search

LifeBook

Remember everyone from each chapter of your life

AI teaching platform

AI tutors tied to a real outcome

Structured speech coach

Practice pause and structure in speech

Personal intelligence system

One radar that feeds idea validation and writing

Solopreneur platform

Pre-wired agents for one-person businesses

CrowdCapital

Community-backed funding for founders

AI and data infra

Meter (AI unit economics)

What one AI-delivered outcome really costs

Personal context router

Picks which files to load for each question

Skill Home

One versioned home for your agent skills

prompt-bench

Did this prompt change help, or was it luck?

ObsGap

Finds AI observability gaps in your repos

promptopt

Automated prompt optimisation loop

Model router

Cheapest model that passes, per request

Canon / MEASUREMENT.md

Metric definitions agents must obey, with a query gate

Context lakehouse

Layered, indexed context agents must read first

Warrant

Scoped, revocable authority for agents

Agent ledger

Audit trail of agent actions

AgentReady

Make SaaS usable by agents

Quorum

Correctness for fleets of agents

Passport (agent registry)

Certifies which software agents can use

RSI Lab

Two honest self-improvement experiments

Lens 2 · Max-Neef's nine needs

Which human needs does each area serve?

Manfred Max-Neef argued that human needs are few and the same everywhere; only the ways we meet them change. The best products meet several needs at once. Each cell counts the ideas in an area that touch a need.

Subsistencemoney, a living

Protectionsafety, trust, rules

Affectionlove, belonging

Understandinglearning, knowing

Participationhaving a say

Leisureplay, rest

Creationmaking things

Identitywho I am

Freedomchoice, time

Vertical

Consumer

AI and data infra

What the two lenses say together

The vertical and infrastructure ideas serve money, safety and understanding. None of them touch affection or play. The consumer ideas reach belonging, identity and freedom, and earn the least. The life-lessons site, AI Fellows and Threshold each meet four needs at once.

The gap I keep coming back to

One idea that is both a business and meaningful would sit where the consumer needs meet the infrastructure skills: AI that helps a person or a family, built on context they can trust.