Don't just sprinkle AI features onto your existing product ('AI at the edge'). Transformative companies rethink workflows and shrink their old codebase, making the LLM a core part of the solution. This is about re-architecting the solution from the ground up, not just enhancing it.

Unlike traditional APIs, LLMs are hard to abstract away. Users develop a preference for a specific model's 'personality' and performance (e.g., GPT-4 vs. 3.5), making it difficult for applications to swap out the underlying model without user notice and pushback.

LLMs operate autoregressively, making one decision (token) at a time without seeing the full problem space. This can lead to hallucinations or dead ends. EBMs are non-autoregressive, allowing them to see all possible routes simultaneously and select an optimal path, much like having a bird's-eye view of a map to avoid a hole in the road.

Despite concerns about the limits of Large Language Models, Microsoft AI's CEO is confident the current transformer architecture is sufficient for achieving superintelligence. Future leaps will come from new methods built on top of LLMs—like advanced reasoning, memory, and recurrency—rather than a fundamental architectural shift.

Enterprises will shift from relying on a single large language model to using orchestration platforms. These platforms will allow them to 'hot swap' various models—including smaller, specialized ones—for different tasks within a single system, optimizing for performance, cost, and use case without being locked into one provider.

To avoid being made obsolete by the next foundation model (e.g., GPT-5), entrepreneurs must build products that anticipate model evolution. This involves creating strategic "scaffolding" (unique workflows and integrations) or combining LLMs with proprietary data, like knowledge graphs, to create a defensible business.

Despite its age, the Transformer architecture is likely here to stay on the path to AGI. A massive ecosystem of optimizers, hardware, and techniques has been built around it, creating a powerful "local minimum" that makes it more practical to iterate on Transformers than to replace them entirely.

The rapid, step-change improvements in LLMs are likely slowing down. This is because models have already been trained on most of the available internet, and the compute budget required for each incremental improvement is increasing exponentially to an unsustainable degree. A new architectural breakthrough, not just more data and compute, is needed for the next leap.

Powerful AI products are built with LLMs as a core architectural primitive, not as a retrofitted feature. This "native AI" approach creates a deep technical moat that is difficult for incumbents with legacy architectures to replicate, similar to the on-prem to cloud-native shift.