AI Company Deep Well

AI company deep well is Li Jigang / 李继刚’s contrast between internet companies and AI-era companies in E45 孟岩对话李继刚：人何以自处. Internet companies are like nets: they create value by connecting people, information, goods, vehicles, meals, or content after the internet removes spatial distance. AI companies may instead be like wells: narrower on the surface, but deeper in their understanding of a user, domain, memory, context, and values.

The practical implication is that a defensible AI application cannot rely on a shallow model wrapper. If a model upgrade can cover the surface behavior, the company needs another form of depth: user context, trusted recommendation, value stance, durable memory, domain-specific judgment, workflow embedding, or emotional relationship.

The episode applies this to financial advice. An AI investment product is valuable only if it understands a person’s life, values, risk tolerance, market situation, and long-term principles well enough to produce a uniquely fitting strategy. That turns Trust As Business Asset, Financial AI Agents, and Context Engineering into business-model foundations rather than nice-to-have personalization.

Key Claims

• Internet companies often scale by connecting many nodes; AI companies may scale value by understanding fewer users more deeply.
• Deep wells depend on context, memory, and value alignment rather than only generic model intelligence.
• Advertising is risky in high-trust AI products because the user may expect the assistant’s answer to be uniquely for them.
• “The best recommendation” can become person-specific: a product that is second for one user may be first for another.
• A thin wrapper is vulnerable when model providers improve; a deep well needs context, workflow, trust, or domain grounding that is not trivially copied.

Connections

• AI As Time Compression — the broader AI-world shift that makes new company forms possible.
• Context Engineering, Persistent Agent Memory, and Context Flywheel — mechanisms that can make the well deeper.
• Trust As Business Asset — trust is the fragile asset in assistant-style recommendations.
• Financial AI Agents and AI Investment Research — investing example from the episode.
• AI Native Product Design and AI Application Layer Moat — product-design and defensibility implications.
• Model Provider Tool Competition and AI Commercialization Pressure — pressure on shallow application-layer wrappers.