The watchlist
48 products, watched for one question: did the interaction change?
A slot here isn't earned by popularity — it's earned by pattern density. A weekly pipeline diffs app-store release notes and changelogs; a human decides what's worth documenting. Missing something? Tell us.
Assistants & chat
The default benchmark — agent mode, memory, voice, apps-in-chat.
Artifacts, projects, desktop agent, connector permissions.
OS-level embedding and the consumer companion pivot.
Visible reasoning that set a global pattern; radical minimalism.
~260M MAU consumer assistant — voice, companion surfaces, in-chat shopping.
Cloud agent, Live mode, deep Workspace and Search integration.
Personality-forward UX and social-platform embedding.
Long-context reading UX and research-report agents.
Consumer surface of the Qwen models; meeting and doc AI lineage.
Search-integrated assistant with digital-human personas.
Messaging-embedded AI inside an existing social graph.
Search & browsers
Browser-native AI — tab context and inline actions.
Citation-first answers, Comet browser, Mac agent.
AI-first search results, camera input, document tools.
Agents
An agent that operates real phone UIs — consent and interruption patterns.
Plan approval, live workspace viewport, session replay, credit metering.
Coding & building
Terminal-agent canon: plan mode, permission tiers, subagents.
Cloud coding agent with task queue and PR-centric review.
IDE agents; watching for post-acquisition UX shifts.
Non-developer app building — scoped edits and publish flow.
End-to-end app agent with checkpoints and rollback.
Prompt-to-UI generation with code/design toggle.
Creative
Voice cloning consent flows and dubbing studio.
Video and voice generation, multimodal composer.
Template-first creation and one-line-to-film agent mode.
Credit and queue economics; community feed as onboarding.
Web editor, personalization loops, room-based creation.
Pro timeline meets generation; reference workflows.
Remix lineage, stems, and persona-based music creation.
Productivity & knowledge
Prompt-to-design inside the canvas.
Ambient meeting capture that enhances rather than replaces notes.
Source-grounded QA and audio overviews — the grounding benchmark.
Embedded-AI canon: inline generation, database autofill.
OS-level entry point; command bar as AI surface.
Newsletter
New patterns and product teardowns from across the AI landscape — what to steal, what to skip, from someone who actually uses them.