Are You Solving the Right Problem?

In partnership with

Hey there,

A Reddit comment caught my attention this week: "So much of our economy is now AI service companies paying AI service companies for AI services."

That hit different.

It came from a discussion about OpenAI's supposedly leaked top 30 customers who burned through over 1 trillion tokens. 

While I can't verify the leak's authenticity, the discussions reveal something important about where we're headed.

The Circular AI Economy

We're building a system where AI companies mostly serve other AI companies, and it's bigger than you think.

Consider these recent developments:

Reflection AI just raised $2 billion at an $8 billion valuation to compete with DeepSeek using other AI models to train their AI models.

AutoBE got 100% compilation success using Qwen3 models to generate backend code, AI building tools for humans to build more AI.

The macro picture: Economists now say that without AI investments, US GDP growth would be just 0.1%. Tech companies are projected to spend nearly $400 billion on AI infrastructure in 2025 alone.

And now there's this: MIT just released an updated version of SEAL, a technique that lets models generate their own training data. It’s basically AI that can rewrite its own code to get smarter. 

The reality? A model that went from 33.5% to 47% accuracy by figuring out what examples help it learn better, then making more of those examples.

Not sentient. Not self-aware. Just more efficient at the work of optimization.

The Sustainability Question

Here's what the circular economy looks like in practice:

• Company A builds an AI model

• Company B uses that model to train their AI

• Company C pays Company B's AI to generate content

• Company D analyzes that content with their AI

• Company E optimizes Company D's AI performance

Where's the actual human value creation happening?

Where Real Value Is Being Created

We're witnessing the birth of the first truly artificial economy, not AI serving humans, but AI serving AI, with humans increasingly becoming middlemen collecting transaction fees.

Some call it progress. Others call it a bubble waiting to burst.

I call it the greatest business opportunity and risk of our lifetime.

While everyone's building chatbots to optimize other chatbots' prompts, there's a massive gap for AI applications that solve actual human problems.

Look at what's working:

• Scientists using AI to develop cancer vaccines with 88% prevention rates in mice

• Researchers creating AI that can detect ADHD through visual processing patterns

• Engineers building AI systems that train robots using human motion data

These aren't AIs serving AIs. These are AIs serving humanity.

Notice the pattern? Each solves a specific, measurable problem for real people. They're not optimizing prompts or generating synthetic training data for other models, they're delivering tangible outcomes.

How to Find Your Human Problem

Here's the framework that's actually working right now:

1. Start with inefficiency, not technology

Don't ask "how can I use AI?" Ask "what takes 10 hours that can take 10 minutes with AI?"

Real example: A legal tech company noticed lawyers spent 60% of their time on document review. They built an AI agent that handles initial review, flagging items that need human judgment. Not revolutionary tech, just AI applied to an obvious pain point. They're now processing millions in contracts monthly.

2. Look for repetitive expertise

AI excels when expert knowledge gets applied repeatedly to slightly different situations. Find professionals doing the same cognitive task over and over.

Where to look:

• Healthcare: Prior authorization reviews, medical coding, initial diagnostic screening

• Finance: Credit analysis, fraud detection, regulatory compliance checks

• Manufacturing: Quality control inspection, predictive maintenance scheduling

• Legal: Contract analysis, due diligence, regulatory research

The opportunity isn't replacing experts. It's giving them leverage.

3. Target industries slow to adopt

Construction, agriculture, logistics, manufacturing, these sectors are drowning in data but starving for implementation. They have problems AI can solve today, but lack the technical resources.

Your advantage: Understanding their workflows matters more than having AI knowledge. A logistics coordinator who learns prompt engineering is more valuable than an AI engineer who doesn't understand supply chains.

4. Build for outcomes, not features

Nobody cares that you're using GPT-4 or Claude or Llama. They care about results.

Bad pitch: "We use advanced language models with RAG architecture..." Good pitch: "We reduce your customer support costs by 40% while improving response time."

The companies winning right now aren't the ones with the most sophisticated AI. They're the ones solving clear problems with measurable ROI.

What You Can Actually Build This Week

Here are three approaches you can start today:

• For consultants/agencies: Offer "AI implementation audits" to local businesses. Most companies don't need custom models, they need someone to show them how to use existing tools effectively. Charge $2-5K to analyze their workflows and implement practical AI solutions using Claude, ChatGPT, or open-source alternatives.

• For domain experts: Take your industry knowledge and build a specialized AI agent with detailed context about your field. Share it with colleagues. If it's valuable, package it as a subscription tool.

• For developers: Pick an industry, interview 5 people who work in it, find their repetitive tasks, build a simple tool. Don't build a platform. Don't build for scale. Build one specific solution for one specific problem. If it works, you'll know what to build next.

When the AI circular economy eventually finds its equilibrium (or crashes), the entrepreneurs who survive will be the ones who built real value for real people.

The companies winning right now share one trait: They started with a human problem and worked backwards to the technology, not the other way around.

So here's my challenge: This week, talk to three people in an industry you understand. Ask them what takes too long, what's frustrating, and what they wish was automated. Don't pitch them anything. Just listen.

The answer to "what human problem are you solving?" might be sitting in one of those conversations.

What do you think? Are we building the future or the world's most expensive house of cards?

Hit reply and let me know what human problem you're tackling—or what industry you think needs AI solutions most.

P.S. If you're already working on something that uses AI to solve actual human problems, I'd love to hear about it. Reply with what you're building and the specific problem it solves. 

- Aashish

200+ AI Side Hustles to Start Right Now

From prompt engineering to AI apps, there are countless ways to profit from AI now. Our guide reveals 200+ actionable AI business models, from no-code solutions to advanced applications. Learn how people are earning $500-$10,000 monthly with tools that didn't exist last year. Sign up for The Hustle to get the guide and daily insights.

Get the guide