When AI models achieve superhuman performance on specific benchmarks like coding challenges, it doesn't solve real-world problems. This is because we implicitly optimize for the benchmark itself, creating "peaky" performance rather than broad, generalizable intelligence.

METR's research reveals a consistent, exponential trend in AI capabilities over the last five years. When measured by the length of tasks an AI can complete (based on human completion time), this 'time horizon' has been doubling approximately every seven months, providing a single, robust metric for tracking progress.

Andrej Karpathy's 'Software 2.0' framework posits that AI automates tasks that are easily *verifiable*. This explains the 'jagged frontier' of AI progress: fields like math and code, where correctness is verifiable, advance rapidly. In contrast, creative and strategic tasks, where success is subjective and hard to verify, lag significantly behind.

AI's capabilities are highly uneven. Models are already superhuman in specific domains like speaking 150 languages or possessing encyclopedic knowledge. However, they still fail at tasks typical humans find easy, such as continual learning or nuanced visual reasoning like understanding perspective in a photo.

The advancement of AI is not linear. While the industry anticipated a "year of agents" for practical assistance, the most significant recent progress has been in specialized, academic fields like competitive mathematics. This highlights the unpredictable nature of AI development.

While language models are becoming incrementally better at conversation, the next significant leap in AI is defined by multimodal understanding and the ability to perform tasks, such as navigating websites. This shift from conversational prowess to agentic action marks the new frontier for a true "step change" in AI capabilities.

AI progress was expected to stall in 2024-2025 due to hardware limitations on pre-training scaling laws. However, breakthroughs in post-training techniques like reasoning and test-time compute provided a new vector for improvement, bridging the gap until next-generation chips like NVIDIA's Blackwell arrived.

The perceived plateau in AI model performance is specific to consumer applications, where GPT-4 level reasoning is sufficient. The real future gains are in enterprise and code generation, which still have a massive runway for improvement. Consumer AI needs better integration, not just stronger models.

The media portrays AI development as volatile, with huge breakthroughs and sudden plateaus. The reality inside labs like OpenAI is a steady, continuous process of experimentation, stacking small wins, and consistent scaling. The internal experience is one of "chugging along."