Now Supporting iPhone Duo

Triangulate mode — a fourth AI mode where Claude, GPT-4.1, and Gemini each answer independently, then a synthesis layer maps where they agree, where they differ, and why Self-deleting chats — set a chat to auto-delete after 1 hour, 24 hours, or 7 days when you start it; a countdown badge shows in the drawer until it disappears Deep research toggle — tap the research button before sending in Standard mode to trigger an extended research pass before the response Mermaid diagrams — flowcharts and sequence diagrams in AI responses now render as actual visuals instead of raw code blocks Personality picker — choose the tone the AI uses across every chat (in your profile) Default model preference — set which model opens new Standard-mode chats by default (in your profile)

- **Copy bubbles** — long-press any user or assistant message to copy its text to clipboard - **Real web search for all models** — Claude searches via Anthropic's server-side tool; GPT-4.1 via OpenAI Responses API; Gemini via Google Search grounding (previously none of them actually searched the web) - **Mode descriptions** — each mode (Standard, Parallel, Bash, Triangulate) now shows a subtitle explaining what it does - **Standard mode flicker fix** — no more brief flash of Claude before the routed provider appears - **Smart scroll** — screen auto-scrolls as AI types, but if you scroll up to read it locks in place and stops chasing the bottom - **Simplified long-press menu** — cleaner action set; "Remove from project" only appears when a thread is actually in a project - **Bulk-move fix** — "Remove from project" in multi-select is gated to project views only - **New Project modal keyboard fix** — keyboard no longer covers the text field when creating a new project on iOS - **GPT identity fix** — GPT-4.1 no longer misidentifies itself as "GPT-4" or references other models' codenames

• Real model names in-app: Claude, GPT-4o, Gemini • New Triangulate mode — runs all three models then maps where they agree, disagree, and why • Per-chat model lock + default model preference in profile • New blue-glass app icon • Refreshed onboarding tour with the four modes, personality presets, and tools/attachments overview • Faster chat surface — animated waveform removed, cleaner composer • Bug fixes: Gemini responses no longer truncate; HEIC photo uploads convert server-side; lock picker now actually routes to the selected model