The 4 Future Opportunities in AI

AI is advancing fast, but the biggest opportunities still lie ahead. Four core problems remain unsolved—each one will define a new class of generational companies. Here’s what they are.

1. The Hallucination Killer

LLMs still make things up. That’s the hallucination problem. Solve this, and you unlock the full power of AI in high-stakes domains—medicine, law, research, finance. The company that cracks this will likely combine retrieval, fact-checking, and reasoning at scale. Once models stop hallucinating, trust—and use cases—explode.

2. The Subgoal Synthesizer

Once hallucinations are solved, agentic AI becomes viable. But agents are only as good as their planning. Today’s agents struggle to break complex tasks into clean, reliable subgoals. The next breakthrough will be in structuring long-range tasks: understanding dependencies, adapting plans, and recovering from failure. Think of it as giving AI a real executive function.

3. The Invention Engine

Current LLMs predict the most likely next word. That’s why they’re boring. They sound smart but rarely say anything new. Invention requires models that don’t just echo what’s probable—but surface what’s nonobvious but true. The company that solves this invent phase will unlock real AI creativity: art, story, insight. Less autocomplete, more inspiration.

4. The Proxy Company

Eventually, AI will make decisions for you—not just give options. This is the proxy stage. Book the trip. Cancel the meeting. Say no to that invite. Today, you’d trust a chief of staff. Soon, you might trust an AI proxy. This leap requires not just intelligence, but judgment and alignment. Whoever builds that trust will own the final interface between humans and machines.

Each of these stages is a bet on the future. But they’re not just technical milestones—they’re blueprints for the next billion-dollar companies.