The original playbook of simply scaling parameters and data is now obsolete. Top AI labs have pivoted to heavily designed post-training pipelines, retrieval, tool use, and agent training, acknowledging that raw scaling is insufficient to solve real-world problems.

While frontier models are like supercars for pushing the limits of intelligence, most enterprise tasks (e.g., invoice processing) don't need that power. More efficient, fine-tuned open-weight models are the practical 'Model Y' for everyday business automation, offering better cost-performance for specific workflows.

It's economically rational to use expensive, high-IQ frontier models for functions with unlimited upside, like sales or product development. For functions where the goal is precision rather than unbounded creativity (e.g., accurately closing financial books), cheaper, fine-tuned open-weight models are more efficient.

Decagon found that fine-tuning smaller, specialized open-source models led to better performance, lower latency, and reduced costs compared to using general-purpose frontier models for specific enterprise workflows. This contradicts the common "smart but expensive vs. dumb but cheap" tradeoff.

The most advanced AI models are not universally superior; their capabilities form a "jagged frontier." This means organizations can often use more economical, locally-run open-weight models for tasks where they are "good enough," reserving expensive frontier models for specialized needs.

Decagon uses powerful frontier models not for high-volume core operations, but for complex, open-ended auxiliary tasks like system-wide performance analysis and model improvement. This hybrid approach optimizes for both efficiency in core tasks and intelligence in strategic ones.

The smartest 'AI-pilled' companies adopt a two-tiered model strategy. They use expensive, frontier models for internal, high-leverage tasks like creating new knowledge and optimizing processes. However, they use cheaper, open-weight models in the 'bill of materials' for the customer-facing product to manage costs effectively.

A production AI agent performs tasks of varying difficulty. Forcing all requests through a single, expensive frontier model is inefficient. A better architecture routes tasks to the most appropriate model: small, cheap open models for high-volume, low-difficulty work like retrieval, reserving the costly frontier API only for high-stakes reasoning where it matters.

Fable demonstrates a new capability: acting as an effective "post-trainer" for smaller, specialized AI models. This achieved a more than 10x performance improvement on a specific task, suggesting a path to a world of abundant, affordable, and safer narrow AI agents trained by larger models.