DoorDash opens a waitlist to order food by text with an AI agent

DoorDash is betting the next app you order from won't be an app. On Tuesday it opened a US waitlist for ordering through an Apple Messages thread, plus a developer waitlist for a connector that lets AI assistants place corporate orders — the first time a major delivery platform has said out loud that agents, not screens, are where it expects to be found.

DoorDash opened a US waitlist for an AI agent that takes orders inside Apple Messages — 20,000 users are already in, and it is iOS-only for now. You text a dish, a craving or a photo of your fridge; the agent searches nearby restaurants, reads menus, proposes a cart, and checks out inside the thread using the payment method on your DoorDash account. It will also message you first — a nudge about your usual coffee, based on habit — and it handles a group order where one person is vegetarian and another wants Thai. DoorDash calls it a pilot for select US account holders and puts a warning in its own help docs: check the cart before paying, because the agent can be wrong. That honesty matters. The 20,000 live users came through an earlier private round; the waitlist at doordash.com/text is how everyone else gets in, with Android "coming soon."

The same announcement carried a B2B connector, and it is the part developers should read. DoorDash published a Model Context Protocol connector — the standard agent-to-tool plumbing — that lets an assistant find items, build a cart, place an order and track arrival. Access is a limited beta aimed at companies ordering for their own employees, with a waitlist opening at developer.doordash.com/mcp; SpaceXAI, Vercel, Cognition, Tempo and Mercor are in the early pool. A Slack bot that collects lunch requests in a thread and places the order is DoorDash's own illustrative example, not a shipped integration — that distinction is worth keeping straight while the agent-race coverage runs hot.

The agent layer is not replacing the storefront. DoorDash also put Macy's, Anthropologie, The North Face, Vans and Timberland on its marketplace, with Macy's and Anthropologie promising delivery in as little as an hour, and added a returns service at a flat $7.99 with pickup starting in 30 minutes. Retailers are not the losers here yet; they are the inventory agents will shop.

Why it matters: today's personal agents — OpenAI's Dots, Meta's Muse — all hit the same wall. They can plan anything and buy almost nothing, because the transaction layer was never built for them. DoorDash is racing to be that layer for food and retail before the assistants pick a winner without asking. Executed well, that is the most durable position in the stack: an agent's default merchant is worth far more than an app icon, and switching costs get set once, by whoever shows up with a working connector first.

We covered the assistant side of this race yesterday — OpenAI's Dots give Pro users an agent with its own computer — and DoorDash is now testing whether the boring half, the part that actually hands over your money, is where the moat is. One caveat before you queue up: DoorDash has not said when the doors open, how it picks who is next, or how often the agent gets an order wrong.

What to watch: whether Apple's Messages platform gives a third-party agent the same standing as a human contact — and whether any restaurant chain demands a say in how an agent reorders on its behalf.

Would you let an AI agent buy your dinner without seeing the cart first? Tell us in the comments.

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Microsoft finished pre-training its biggest in-house model yet — then used it to make a point about the company it bankrolls. Microsoft AI completed pre-training of MAI-Base-2, a mixture-of-experts foundation model with 80 billion active and 2.8 trillion total parameters, and says the recipe is 4.2–12.3x more pre-training FLOPs-efficient than DeepSeek-V4-Pro-Base. The run itself is the number worth keeping: 16,000 Nvidia GB200 GPUs for 38 days across August and September, at 21% average model

Amazon's new Bee devices listen all day — for $11.99 a month

Amazon turned its always-listening Bee line into a two-device family on Thursday, and NVIDIA opened a 2.8-trillion-token Indic corpus recipe. The connecting thread: both are bets that the next AI money is in memory — yours, or your model's. Amazon put two new always-listening devices on sale, and both now lead to a subscription. Bee Frontier, the next-generation wristband that also clips to clothing, ships now at $74.99 in Onyx, Lavender and Sage — up from the $50 predecessor — with seven-day

Perplexity open-sources its agent sandbox after 118 failed escapes

Perplexity handed its agent containment layer to the open-source world, and Washington's AI task force stopped reading briefings and started working. Perplexity open-sourced SPACE on Thursday — the sandbox it built to keep its own AI agents from getting out — and says nine frontier models failed to break out of it across 118 attempts. Each agent runs inside a small virtual machine with its own operating system, with credentials and network access managed outside the sandbox so an agent can re

Read more

Brain organoids classified Braille at 83% — in a trio, not alone

Suleyman says Microsoft's base model beat OpenAI's on a quarter compute

Amazon's new Bee devices listen all day — for $11.99 a month

Perplexity open-sources its agent sandbox after 118 failed escapes