With Gemini 4 Neon/Helium, GPT-6.2 Sol ($2/$10 per 1M tokens) vs Claude Opus 5.5 ($4/$20), and Anthropic pricing its IPO at $83 a share, almost every conversation this week has been about token unit economics and benchmark indices.
Yet one of the more revealing product leaks today is much simpler: Google testing three explicit Writing Modes inside Gemini Notebook—Writing Partner (default AI co-editing), Writing Coach (chat-only suggestions without touching the text), and a hard "No AI" toggle.
That split highlights a real workflow bottleneck developers and technical writers keep running into:
- Silent text mutation breaks ownership: When an assistant rewrites notes or code comments inline by default, you lose track of what you verified versus what was smoothed over.
- Advice beats auto-edits in high-context work: A "coach" mode that critiques structure or points out missing edge cases in a side panel lets you keep deterministic control over the actual artifact.
- Cost and latency routing is now multi-tiered anyway: Between $0.10/M decision models like Gemini Decision 1 or Muse Ember ($0.10/$0.50) and $4/$20 frontier models, throwing the heaviest model at inline rewrites makes less sense every week.
Where do you draw the line in your own editor and note-taking stack—do you let AI edit inline, keep it strictly in a side-channel critic role, or turn it off completely for first-pass thinking?
Top comments (0)