Claude Code picked up a handful of end-of-week changes, headlined by a new “maximum effort” reasoning mode that’s explicitly designed to trade usage for deeper work. In a set of updates posted by Thariq on X, Claude Code now supports setting effort to max, which “reasons for longer and uses as many tokens as needed,” while warning that it will burn through usage limits faster and therefore must be enabled per session. The feature is available via the /effort command (and can be tested immediately), as shared in the thread at x.com/trq212/status/2032632596572811575.
“Effort: max” is a new per-session dial for token spend
The key behavioral change is straightforward: effort max pushes Claude to spend more tokens on average, with the intent of prioritizing the best possible outcome over thrift. In a reply clarifying the intent, Thariq described it as, essentially, “try your absolute best don’t worry about tokens.”
The max reasoning setting only applies to Opus models.
A few practical notes from the same thread:
- It’s opt-in per session, explicitly to avoid surprise usage spikes.
- Switching levels requires typing the level (i.e., it’s not a toggle).
- For CLI and automated flows,
--effort maxalso works, addressing concerns that agents and scripts can’t easily run interactive slash commands.
Predictably, some of the discussion immediately turned to how this relates to “ultrathink” and whether the two stack, but the thread itself doesn’t spell out an interaction model—only that “max” effort is now available and documented.
Remote Control can start new sessions
Another quality-of-life update: new sessions can now be started from Remote Control. It’s a small change, but it suggests the Remote Control surface is moving closer to being a first-class session manager rather than just a “drive what’s already running” feature.
Voice mode reaches 100%, including Claude code Desktop
Thariq also noted that voice mode is now rolled out to 100% of users, explicitly including Claude code Desktop. For teams that bounce between keyboard-driven coding and spoken prompts (especially during reviews or debugging walkthroughs), this is the kind of input-mode parity that tends to matter over time.
Setup scripts arrive on web and desktop
Claude Code can now run a setup script on both the web and desktop apps. The post doesn’t elaborate on the scripting interface or guardrails, but the directional intent is clear: reducing the friction between “open a session” and “have the environment prepared.”
Smaller workflow changes: naming, color, hooks
A final batch of smaller features rounds out the update:
claude --name <NAME>names a session at start.- After plan mode, Claude will auto-name the session.
/colorchanges the color of the prompt input.- There’s also mention of a postcompact hook.
For the full details, Thariq pointed readers to the product documentation and changelog here: https://t.co/jcaYzwGMdv.
Original source: https://x.com/trq212/status/2032632596572811575
