AI Daily Brief — 6 April 2026

Cognition’s Windsurf ships an Adaptive model router — dynamically chooses the best underlying model per task while billing at a fixed per-token rate — alongside a redesigned model picker that shows token pricing inline and the removal of Max-tier daily limits.

Top stories

• Windsurf Adaptive launches. Adaptive routes each task to the right underlying model, helping users stretch quota by avoiding premium-model overuse. Available to Pro, Max and Teams self-serve users. via Windsurf
• Pricing transparency push. The new model picker exposes exact per-token billing rates inline — a direct response to user feedback that pricing had become opaque. via DigestWeb
• NVIDIA joins Devin Desktop multi-agent preview. NVIDIA engineers are running multi-agent workflows on Devin Desktop, helping Cognition define how agents share context and coordinate across complex tasks. via Devin

Quiet corners

The week opens with attention on whether Anthropic‘s anticipated “Mythos” preview lands in the next few days.

By the numbers

• Adaptive rolled out to all self-serve Pro / Max / Teams users
• Daily limits removed for Max tier

Compiled by AI Feed’s editor from publicly reported announcements on 6 April 2026.