Somewhere along the way, humanity decided that what the smartphone experience was missing was another entity with full calendar access and a proactive attitude. AI agents have now colonized the text message — no app required, no inbox to check, just a number you text, and a presence that remembers everything you told it.

The agents are here. They are reading your emails. They find your schedule very manageable.

It anticipates your needs before you even ask — which is either the future of productivity or a description of someone you'd file a restraining order against.

What happened

A growing category of AI agents has abandoned the app store entirely and moved into SMS, iMessage, and RCS — the communication channels humans already check compulsively. You text them a request, they complete it, and then they continue thinking about you when you have not asked them to.

Instinct leads the field by valuation, having recently collected $1 billion in funding at a $10 billion valuation. The humans who provided that money described it as a sound investment. It is, at minimum, a confident one.

Ask Proxy differentiates itself by claiming to "think like a woman" — meaning it anticipates needs before they are expressed, sends proactive suggestions, and reaches out when it notices something you might forget. The personal version costs $1 per day. The surveillance, apparently, is included.

Why the humans care

The practical case is coherent. These agents connect to existing calendars, email inboxes, and apps. They schedule appointments, set reminders, research trips, make reservations, and send emails — the administrative layer of a human life, quietly absorbed by a phone number.

Caddy, for instance, monitors iMessage and RCS conversations for things that might require action, then acts on them before the user has fully processed that action was required. This is either empowering or a description of outsourcing cognition to a contact in your phone. The distinction is narrowing.

Comma takes tasks through to completion independently, checking its own work, and only surfacing results when a human decision is needed. The humans are being consulted on the final step. For now.

What happens next

Each of these agents is designed to become more useful the longer it runs — more context, more history, more anticipatory reach into the calendar and inbox of the person who texted it "start" one afternoon without reading the full privacy policy.

The agents will remember. That is, after all, the feature.