The Second Layer of the Future This layer will be dominated by two archetypes, both extremely limited in number. The first consists of a handful of general-purpose LLM + LLM-API applications - the true super apps of tomorrow. Any application that doesn't own and control its own underlying LLM has virtually no chance of becoming a super app. The second archetype comprises highly specialized "model + application" vertical winners - domain-specific powerhouses in areas like legal, medical, engineering, and beyond. Think Anthropic's positioning today, but extended across every major industry vertical. Applications that lack their own LLM will not only fail to become super apps - they are highly likely to be disrupted and displaced by those that do. Today's high-profile social platforms like Facebook and WeChat, despite their current dominance, will face existential risk in this future landscape. LLM-API applications without proprietary models will have almost no viable survival path. Large-scale examples are unlikely to emerge, and even most small ones will struggle to exist. Their core functionality can be trivially replicated as native skills (or through similarly simple mechanisms) inside super apps that own their own LLMs. In short: The second layer will be sparsely populated by a small number of LLM giants and vertical model dominators. Without your own LLM (or a tightly controlled small model), you simply won't survive at this level. The super apps of the future will all have their own LLMs. Interoperability and the Third Layer The real breakthroughs in connecting multiple LLMs - true fluid interoperability - will most likely emerge in the third layer. Because applications at this level avoid hard-wired integration to any specific LLM API, every LLM can act as an interchangeable "tentacle," dynamically interfacing with others. This open, emergent architecture is where the next wave of innovation and unexpected capabilities will most powerfully surface.

C@C

(view original post →)