[BUG] Dispatch burns tokens unnecessarily! Makes me want to move away from Max Plan! Competitor is better!

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Bug 1: No context compression or session rollover
Long Dispatch sessions accumulate tokens until the limit is hit mid-task with no warning, no compression, and no recovery path. There is no way to summarize and continue. Workflows are interrupted and context is lost.
Bug 2: No model selection in Dispatch
Users on Max 5x ($100/mo) are forced through the heaviest model for every task including trivial ones where Haiku would be sufficient. There is no model selector in the Dispatch UI. This burns limits faster and removes cost/speed control entirely.
Bug 3: No new thread option in Dispatch
There is no "New Thread" button in Dispatch. When a session hits limits or users need a clean context, there is no in-app reset path. This compounds Bug 1 — context doesn't compress and there's no escape valve.
Business impact:
I am a Max plan subscriber at $100/month. These three failures together make Dispatch unusable for sustained development workflows. Competing tools at $20/month are delivering more productive output per dollar. I am actively moving my spend to a competitor as a direct result of these unresolved issues. Anthropic is losing a paying Max plan customer over fixable UX gaps.

Long Dispatch sessions accumulate tokens until the limit is hit mid-task with no warning, no compression, and no recovery path. There’s also no model selector (forces heaviest model even for trivial tasks) and no “New Thread” button to reset context, so workflows get interrupted and context is lost.

What Should Happen?

Context compression/session rollover, model selector, and a New Thread option to continue cleanly.

Error Messages/Logs

Steps to Reproduce

Start a Dispatch session and keep working until token limit is reached.
Observe no compression, no rollover, and no way to summarize and continue.
Try to switch model or start a clean context - no UI controls available.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

1.1617.0

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_