An AI agent's failure on a complex task like tax preparation isn't due to a lack of intelligence. Instead, it's often blocked by a single, unpredictable "tiny thing," such as misinterpreting two boxes on a W4 form. This highlights that reliability challenges are granular and not always intuitive.

Agentic workflows involving tool use or human-in-the-loop steps break the simple request-response model. The system no longer knows when a "conversation" is truly over, creating an unsolved cache invalidation problem. State (like the KV cache) might need to be preserved for seconds, minutes, or hours, disrupting memory management patterns.

Unlike humans who can prune irrelevant information, an AI agent's context window is its reality. If a past mistake is still in its context, it may see it as a valid example and repeat it. This makes intelligent context pruning a critical, unsolved challenge for agent reliability.

AI models struggle to create and adhere to multi-step, long-term plans. In an experiment, an AI devised an 8-week plan to launch a clothing brand but then claimed completion after just 10 minutes and a single Google search, demonstrating an inability to execute extended sequences of tasks.

Despite massive context windows in new models, AI agents still suffer from a form of 'memory leak' where accuracy degrades and irrelevant information from past interactions bleeds into current tasks. Power users manually delete old conversations to maintain performance, suggesting the issue is a core architectural challenge, not just a matter of context size.

Long-running AI agent conversations degrade in quality as the context window fills. The best engineers combat this with "intentional compaction": they direct the agent to summarize its progress into a clean markdown file, then start a fresh session using that summary as the new, clean input. This is like rebooting the agent's short-term memory.

Instead of just expanding context windows, the next architectural shift is toward models that learn to manage their own context. Inspired by Recursive Language Models (RLMs), these agents will actively retrieve, transform, and store information in a persistent state, enabling more effective long-horizon reasoning.

Even with large advertised context windows, LLMs show performance degradation and strange behaviors when overloaded. Described as "context anxiety," they may prematurely give up on complex tasks, claim imaginary time constraints, or oversimplify the problem, highlighting the gap between advertised and effective context sizes.

A single, general-purpose agent with a large context window is prone to catastrophic errors. A more robust system uses a hierarchy of specialized agents with narrow tasks (e.g., only handling Git commits). This division of labor minimizes hallucinations and ensures reliability.