-
Open Reasonix and start a coding/debugging task on an existing production project.
-
Ask Reasonix to diagnose and fix a production coding issue.
-
Let Reasonix inspect multiple project files and continue debugging across the codebase.
-
During the same session, Reasonix repeatedly searched/read files, inspected configuration, checked application logic, reviewed API-related code, and continued diagnosing the issue.
-
Keep the coding/debugging session running for about one hour.
-
Check DeepSeek billing/token usage after the session.
Actual result:
Reasonix consumed approximately USD 20 in about one hour. This is far above my normal DeepSeek mass-coding usage, which is usually around USD 10 per month.
Expected result:
Reasonix should not consume such a large amount in one hour under normal coding/debugging usage. It should keep DeepSeek cache usage stable, avoid repeatedly resending huge context, show cost/context/cache-hit information clearly, warn before high-cost actions, and allow a hard cost limit per session.
Open Reasonix and start a coding/debugging task on an existing production project.
Ask Reasonix to diagnose and fix a production coding issue.
Let Reasonix inspect multiple project files and continue debugging across the codebase.
During the same session, Reasonix repeatedly searched/read files, inspected configuration, checked application logic, reviewed API-related code, and continued diagnosing the issue.
Keep the coding/debugging session running for about one hour.
Check DeepSeek billing/token usage after the session.
Actual result:
Reasonix consumed approximately USD 20 in about one hour. This is far above my normal DeepSeek mass-coding usage, which is usually around USD 10 per month.
Expected result:
Reasonix should not consume such a large amount in one hour under normal coding/debugging usage. It should keep DeepSeek cache usage stable, avoid repeatedly resending huge context, show cost/context/cache-hit information clearly, warn before high-cost actions, and allow a hard cost limit per session.