Where We Are At With AI — Current State of LLMs

The AI industry’s massive investment in drug discovery has failed to produce a single FDA-approved drug, exposing the dangerous gap between hype-driven narratives and actual scientific outcomes. Large Language Models (LLMs) function as powerful statistical engines rather than precursors to AGI, lacking the causal reasoning necessary for complex, real-world applications. Because LLMs generate plausible rather than correct code, they require rigorous software engineering, such as deterministic verification and targeted, vertical-specific implementations, to be truly effective. Rather than chasing general intelligence, businesses should focus on solving specific, high-value problems using custom AI stacks that prioritize reliability and control. Ultimately, the most successful organizations will be those that integrate AI into solid engineering foundations, treating these tools as infrastructure to be owned and managed rather than rented solutions prone to hallucination and architectural failure.