---
title: "Agents Keep Running Out of Context Mid-Task"
canonical: "https://current.tinyfish.ai/issue/10/foundations/article/7224/agents-keep-running-out-of-context-mid-task"
publication: "TinyFish"
publisher: "TinyFish"
language: "en-US"
guide: "https://current.tinyfish.ai/llms.txt"
issue: 10
section: "foundations"
section_name: "Foundations"
published: 2025-12-31
word_count: 109
source_links: 0
---

# Agents Keep Running Out of Context Mid-Task

*December 31, 2025*

[Read on the website](<https://current.tinyfish.ai/issue/10/foundations/article/7224/agents-keep-running-out-of-context-mid-task>)

We're watching companies hit the same wall. Agents start complex tasks, run out of context halfway through, leave work half-finished. Anthropic documented this clearly: even frontier models exhaust their context windows mid-project, forcing the next session to reconstruct what happened.

Here's what matters. Companies are designing agents to work like human developers—continuous sessions, maintained context, marathon coding sprints. Agents need different architecture. They operate in discrete sessions with no memory of what came before.

Look at what's actually working: multi-agent systems, the 89% observability adoption rate. Both point to the same lesson. Agents need explicit session boundaries, clear state management, specialized roles. Not attempts to replicate human work patterns.

## More

- [Foundations — Issue 10](<https://current.tinyfish.ai/issue/10/foundations/markdown>)
- [Issue 10](<https://current.tinyfish.ai/issue/10/markdown>)
- [Publication guide](<https://current.tinyfish.ai/llms.txt>)