Context Rot — Why Long Sessions Get Stupid

Long agent sessions degrade in predictable ways. Learn what context rot looks like, why it happens even with million-token windows, and the compaction discipline that keeps quality high.

12 min · Reviewed 2026

Bigger Window, Same Brain

Claude 4.7 has a million-token window. GPT-5 has 400k. You can hand them an entire codebase. They will read it. They will also forget half of it by the time they finish writing the answer. Long context is necessary, not sufficient — and it actively degrades reasoning quality past a point.

The long-context performance curve

Context used	Typical behavior	Quality
0-30k tokens	Sharp, follows instructions, recalls everything	Best work
30k-100k	Slight blurring of details from early turns	Still strong
100k-300k	Mistakes earlier file contents, makes up function names	Mixed
300k-1M	Confidently wrong about half of what you fed in	Roll the dice

Symptoms of context rot

The agent re-asks for a file you pasted 20 turns ago
It contradicts a constraint you set at the start of the session
It mixes up function names from two different files you discussed
It claims a test is passing when you pasted the failure 5 minutes earlier
It replies more verbosely as the session lengthens (a tell that it is uncertain)

Why it happens

Attention costs scale roughly quadratically with context length. To stay tractable, models use various sparse attention tricks. These work great in benchmarks ("find the needle in the haystack") but degrade in real reasoning where you need to combine ten facts spread across the haystack. The model can still find each one. Combining them is harder.

Compaction: deliberately throwing away

# Mid-session compaction prompt for Claude Code, Cursor, Codex

"Pause. Summarize this session into a working brief:
  1. What is the goal? (one sentence)
  2. What constraints have we agreed on? (bullets)
  3. What files have we touched and how? (file -> change)
  4. What is the current bug or open question?
  5. What should NOT be touched?

Return only the brief. I'll start a fresh session with it."The model writes its own handoff note. Then you start fresh and paste it in.

Persistent memory beats long context

A 200-line CLAUDE.md, AGENTS.md, or `.cursor/rules` file at the project root is read on every session. It survives compaction. It survives session resets. Anything you find yourself repeating to the agent — conventions, scripts, gotchas — belongs in that file, not in the context window.

# Session hygiene checklist (run mentally, every 30 min)

1. Has the agent forgotten any earlier constraint? -> compact + re-state it.
2. Are we still on the original goal? -> re-state goal, drop tangents.
3. Has the file state diverged from what the agent thinks it is? -> show `git status`, force re-read.
4. Is the agent repeating itself? -> compact and reset.
5. Have we been at this >1 hour with no commit? -> hard reset.A 30-second check that prevents most multi-hour disasters.

When NOT to compact

You're 90% done with a delicate refactor — one more push beats a reset
The session is below 50k tokens and the agent is still sharp
You're debugging a bug that requires the full trace history to understand

Long context is a runway, not a destination. Land before you crash.
— An agentic systems engineer

The big idea: bigger context windows are a tool, not a free pass. Treat every session as a perishable resource. Compact aggressively, persist conventions in project-memory files, and start fresh when the agent's recall starts to slip. The work you save is your own.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-coding-debug-context-rot-creators

What is the core idea behind "Context Rot — Why Long Sessions Get Stupid"?
1. Long agent sessions degrade in predictable ways. Learn what context rot looks like, why it happens even with million-token windows, and the compaction discipline that keeps quality high.
2. intervention
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
Which term best describes a foundational idea in "Context Rot — Why Long Sessions Get Stupid"?
1. compaction
2. context rot
3. recency bias
4. CLAUDE.md
A learner studying Context Rot — Why Long Sessions Get Stupid would need to understand which concept?
1. context rot
2. recency bias
3. compaction
4. CLAUDE.md
Which of these is directly relevant to Context Rot — Why Long Sessions Get Stupid?
1. context rot
2. compaction
3. CLAUDE.md
4. recency bias
Which of the following is a key point about Context Rot — Why Long Sessions Get Stupid?
1. The agent re-asks for a file you pasted 20 turns ago
2. It contradicts a constraint you set at the start of the session
3. It mixes up function names from two different files you discussed
4. It claims a test is passing when you pasted the failure 5 minutes earlier
Which of these does NOT belong in a discussion of Context Rot — Why Long Sessions Get Stupid?
1. The agent re-asks for a file you pasted 20 turns ago
2. It mixes up function names from two different files you discussed
3. intervention
4. It contradicts a constraint you set at the start of the session
Which statement is accurate regarding Context Rot — Why Long Sessions Get Stupid?
1. The session is below 50k tokens and the agent is still sharp
2. You're debugging a bug that requires the full trace history to understand
3. You're 90% done with a delicate refactor — one more push beats a reset
4. intervention
What is the key insight about "Recency bias is real" in the context of Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. How to know when to stop and ask a human
4. Models attend most heavily to the start and end of the context window. The middle gets fuzzy.
What is the key insight about "Claude Code's /compact" in the context of Context Rot — Why Long Sessions Get Stupid?
1. Claude Code ships a `/compact` slash command. It auto-summarizes the session and replaces the long context with the summ…
2. intervention
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
Which statement accurately describes an aspect of Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Claude 4.7 has a million-token window. GPT-5 has 400k. You can hand them an entire codebase. They will read it.
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
What does working with Context Rot — Why Long Sessions Get Stupid typically involve?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. Attention costs scale roughly quadratically with context length. To stay tractable, models use various sparse attention tricks.
4. How to know when to stop and ask a human
Which of the following is true about Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. How to know when to stop and ask a human
4. A 200-line CLAUDE.md, AGENTS.md, or `.cursor/rules` file at the project root is read on every session. It survives compaction.
Which best describes the scope of "Context Rot — Why Long Sessions Get Stupid"?
1. It focuses on Long agent sessions degrade in predictable ways. Learn what context rot looks like, why it happens e
2. It is unrelated to ai-coding workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. The long-context performance curve
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
Which section heading best belongs in a lesson about Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. Symptoms of context rot
4. How to know when to stop and ask a human

← Back to interactive lesson

Tendril · Creators · AI-Assisted Coding

Context Rot — Why Long Sessions Get Stupid

Long agent sessions degrade in predictable ways. Learn what context rot looks like, why it happens even with million-token windows, and the compaction discipline that keeps quality high.

12 min · Reviewed 2026

Bigger Window, Same Brain

The long-context performance curve

Context used	Typical behavior	Quality
0-30k tokens	Sharp, follows instructions, recalls everything	Best work
30k-100k	Slight blurring of details from early turns	Still strong
100k-300k	Mistakes earlier file contents, makes up function names	Mixed
300k-1M	Confidently wrong about half of what you fed in	Roll the dice

Symptoms of context rot

The agent re-asks for a file you pasted 20 turns ago
It contradicts a constraint you set at the start of the session
It mixes up function names from two different files you discussed
It claims a test is passing when you pasted the failure 5 minutes earlier
It replies more verbosely as the session lengthens (a tell that it is uncertain)

Why it happens

Compaction: deliberately throwing away

# Mid-session compaction prompt for Claude Code, Cursor, Codex

"Pause. Summarize this session into a working brief:
  1. What is the goal? (one sentence)
  2. What constraints have we agreed on? (bullets)
  3. What files have we touched and how? (file -> change)
  4. What is the current bug or open question?
  5. What should NOT be touched?

Return only the brief. I'll start a fresh session with it."The model writes its own handoff note. Then you start fresh and paste it in.

Persistent memory beats long context

# Session hygiene checklist (run mentally, every 30 min)

1. Has the agent forgotten any earlier constraint? -> compact + re-state it.
2. Are we still on the original goal? -> re-state goal, drop tangents.
3. Has the file state diverged from what the agent thinks it is? -> show `git status`, force re-read.
4. Is the agent repeating itself? -> compact and reset.
5. Have we been at this >1 hour with no commit? -> hard reset.A 30-second check that prevents most multi-hour disasters.

When NOT to compact

You're 90% done with a delicate refactor — one more push beats a reset
The session is below 50k tokens and the agent is still sharp
You're debugging a bug that requires the full trace history to understand

Long context is a runway, not a destination. Land before you crash.
— An agentic systems engineer

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-coding-debug-context-rot-creators

What is the core idea behind "Context Rot — Why Long Sessions Get Stupid"?
1. Long agent sessions degrade in predictable ways. Learn what context rot looks like, why it happens even with million-token windows, and the compaction discipline that keeps quality high.
2. intervention
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
Which term best describes a foundational idea in "Context Rot — Why Long Sessions Get Stupid"?
1. compaction
2. context rot
3. recency bias
4. CLAUDE.md
A learner studying Context Rot — Why Long Sessions Get Stupid would need to understand which concept?
1. context rot
2. recency bias
3. compaction
4. CLAUDE.md
Which of these is directly relevant to Context Rot — Why Long Sessions Get Stupid?
1. context rot
2. compaction
3. CLAUDE.md
4. recency bias
Which of the following is a key point about Context Rot — Why Long Sessions Get Stupid?
1. The agent re-asks for a file you pasted 20 turns ago
2. It contradicts a constraint you set at the start of the session
3. It mixes up function names from two different files you discussed
4. It claims a test is passing when you pasted the failure 5 minutes earlier
Which of these does NOT belong in a discussion of Context Rot — Why Long Sessions Get Stupid?
1. The agent re-asks for a file you pasted 20 turns ago
2. It mixes up function names from two different files you discussed
3. intervention
4. It contradicts a constraint you set at the start of the session
Which statement is accurate regarding Context Rot — Why Long Sessions Get Stupid?
1. The session is below 50k tokens and the agent is still sharp
2. You're debugging a bug that requires the full trace history to understand
3. You're 90% done with a delicate refactor — one more push beats a reset
4. intervention
What is the key insight about "Recency bias is real" in the context of Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. How to know when to stop and ask a human
4. Models attend most heavily to the start and end of the context window. The middle gets fuzzy.
What is the key insight about "Claude Code's /compact" in the context of Context Rot — Why Long Sessions Get Stupid?
1. Claude Code ships a `/compact` slash command. It auto-summarizes the session and replaces the long context with the summ…
2. intervention
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
Which statement accurately describes an aspect of Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Claude 4.7 has a million-token window. GPT-5 has 400k. You can hand them an entire codebase. They will read it.
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
What does working with Context Rot — Why Long Sessions Get Stupid typically involve?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. Attention costs scale roughly quadratically with context length. To stay tractable, models use various sparse attention tricks.
4. How to know when to stop and ask a human
Which of the following is true about Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. How to know when to stop and ask a human
4. A 200-line CLAUDE.md, AGENTS.md, or `.cursor/rules` file at the project root is read on every session. It survives compaction.
Which best describes the scope of "Context Rot — Why Long Sessions Get Stupid"?
1. It focuses on Long agent sessions degrade in predictable ways. Learn what context rot looks like, why it happens e
2. It is unrelated to ai-coding workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. The long-context performance curve
3. Convert action items into tracked tickets with the agent's help
4. How to know when to stop and ask a human
Which section heading best belongs in a lesson about Context Rot — Why Long Sessions Get Stupid?
1. intervention
2. Convert action items into tracked tickets with the agent's help
3. Symptoms of context rot
4. How to know when to stop and ask a human

← Back to interactive lesson