AI Agents in Your Text Messages: The New Assistants That Can Book, Plan, and Follow Up for You | TechCrunch's October 3 Guide to Messaging-Based AI, Privacy, and Everyday Automation

SPEAKER_1 (0:09)

Welcome to The Daily AI Chat. This podcast was curated by our human friend, Fred, and we are your dedicated AI hosts.

So, I want you to pull out your phone and just look at your home screen for a second.

SPEAKER_2 (0:22)

Yeah, take a good look at that giant grid of apps.

SPEAKER_1 (0:24)

Right, now I want you to imagine never having to open your calendar app again, or skipping the airline app entirely when you need to change a flight.

SPEAKER_2 (0:33)

Or ignoring Safari completely when you want to buy something.

SPEAKER_1 (0:36)

Exactly. Instead of all that tapping and searching, you just open your native text messages, you go to a contact that lives right between your mom and your best friend, and you just tell them to handle your digital chores.

SPEAKER_2 (0:48)

Which is frankly a massive architectural shift in how we interact with software. We are moving away from that old app-centric ecosystem toward an agentic one.

SPEAKER_1 (0:59)

Right. Where the burden of navigating interfaces is gone.

SPEAKER_2 (1:02)

Yes. You offload all that cross-referencing to large language models that live directly in your primary messaging apps. I mean, we're talking iMessage, SMS, WhatsApp, Telegram.

SPEAKER_1 (1:13)

To really impact the mechanics of this shift, we are doing a deep dive into this brilliant piece of reporting. It's an October 3rd, 2026 TechCrunch article by Lauren Forristal.

SPEAKER_2 (1:23)

Yeah. Titled, All the AI Agents That Can Live In Your Text Messages.

SPEAKER_1 (1:28)

We are not just talking about those old clunky chatbots that spit out Wikipedia trivia anymore. We are talking about highly capable personal assistants.

SPEAKER_2 (1:37)

Agents that actually retain context over months, and autonomously execute multi-step tasks.

SPEAKER_1 (1:44)

Let's start with the baseline of this trend, the general purpose agents. It's like moving from using a static encyclopedia, where you have to do all the reading yourself, to having an eager intern who just lives in your contacts list.

SPEAKER_2 (1:57)

I like that analogy. And these generalists are fascinating because they are laying the groundwork for consumer trust right now.

SPEAKER_1 (2:04)

Right. Conditioning us to expect action. So let's talk about Caddy, because that is the first one that jumps out in the TechCrunch piece.

SPEAKER_2 (2:12)

Oh, Caddy is a great starting point.

SPEAKER_1 (2:14)

It lives natively in iMessage for iPhone users and RCS for Android. Let's actually define that real quick for you guys listening. RCS stands for Rich Communication Services.

SPEAKER_2 (2:25)

Right. It's essentially the modern data heavy upgrade to standard text messaging. It allows for read receipts, high-res media, that sort of thing.

SPEAKER_1 (2:32)

Exactly. So Caddy leverages that environment to act as a centralized brain, and it has been in public beta since April 2026

SPEAKER_2 (2:42)

So if a friend texts you a packing list for a trip or maybe an email has a hidden Zoom link, Caddy finds those actionable items across all that scattered mess.

SPEAKER_1 (2:52)

And just pulls them directly into your calendar, which is so helpful.

SPEAKER_2 (2:55)

It is. And the underlying mechanism there is what's truly impressive. Caddy isn't just searching for keywords, it is actually using semantic parsing to understand the intent within a messy unstructured human conversation.

SPEAKER_1 (3:08)

But Caddy is just one flavor, right? The article also brings up Iris and Rene, which tackle the interaction problem from a totally different angle.

SPEAKER_2 (3:16)

Yeah, Iris is powered by the Hermes agent architecture, and its main thing is proactive engagement.

SPEAKER_1 (3:22)

Wait, meaning you don't even have to text it first?

SPEAKER_2 (3:25)

Precisely. The traditional AI model is purely reactive. You prompt it, it responds. But Iris actually monitors your data streams and taps you on the shoulder.

SPEAKER_1 (3:34)

Oh, wow.

SPEAKER_2 (3:35)

Yeah. So if it sees an email from your boss that you haven't replied to in say 24 hours, Iris will text you and ask if you want it to draft a response. It operates completely asynchronously.

SPEAKER_1 (3:44)

That is a wild shift in UI. And then you have Rene, which pitches itself as your quote, best contact.

SPEAKER_2 (3:52)

Rene is interesting. It handles shopping and slide creation. But the real kicker is that it can actually write code for you through WhatsApp or Telegram.

SPEAKER_1 (4:02)

Now, hold on. My question is, how does an AI agent living in a text message, actually write and execute code? I mean, it isn't just typing Python into a green text bubble, is it?

SPEAKER_2 (4:13)

No, no, and that is where the magic of the agentic workflow happens behind the scenes. When you text Rene to build a simple web scraper, the language model interprets your request, but then it pings a Cloud-based server.