Agents Go Shopping Before Black Friday: Shopify’s Quarter, Dots’ Second Brain and a Checklist for Your Wallet

Black Friday is still three weeks away, and the shoppers have already changed. Some of them no longer have thumbs. On Tuesday, Shopify reported a quarter built on what its president called “every agent,” OpenAI’s Dots learned to pick its own brain for each errand, and a new open model posted startling scores on the web-browsing tests that matter most for an assistant sent out to buy things. Here is what happened, and what to do about it before your own software starts filling carts.

Shopify: Record Volume, a Rare Miss and the Word “Agent”

In its third-quarter release, Shopify reported revenue up 32 percent and a free cash flow margin of 22 percent. President Harley Finkelstein put the thesis in a single sentence: “We powered $119 billion in GMV, while continuing to build commerce into every surface, every agent, and increasingly, everywhere.”

Wall Street wanted a little more. Yahoo Finance’s Inés Ferré reported that shares fell about 8 percent in morning trading after revenue of 3.74 billion dollars came in just under the 3.75 billion consensus, the company’s first revenue miss since 2022. Subscription revenue (836 million dollars) and gross merchandise volume (119.15 billion) also trailed estimates, while adjusted earnings per share edged past them. The chief financial officer, Jeff Hoffmeister, held the line: “Delivering these results at our scale speaks to the strength of our model, our multiple growth drivers, and our expanding addressable market.”

Why it matters to you: when one of the largest commerce platforms in the world tells investors its next surface is “every agent,” the checkout page you are used to is about to become optional. Merchants are being built to sell to software that shops for you.

RelatedSuperintelligence Week: Gemini 4 Carbon Tops the Board, Terra Halves the Price, Pika Opens Its Video Model

Dots Gets a Second Brain

Android Authority’s Shimul Sood reports that OpenAI’s Thibault Sottiaux said on X that the models powering ChatGPT Dots now include GPT-6 Astra and GPT-6.1 Terra. Terra is described as faster with a friendlier personality and is meant to bring Dots to more people. Dots will choose the model automatically and can switch models in the middle of a single task, and users can message a Dot from ordinary text-messaging apps.

Why it matters to you: the assistant that compares three toasters and the one that pays for the winner may now be two different minds inside one conversation. That is efficient. It also means the moment of payment deserves your full attention, because it may not be handled by the model you were chatting with.

LongCat-2.5 and the Browsing Test

The open-model news of the week came from the LongCat team’s Nov. 1 release of LongCat-2.5, a natively multimodal model built for long-horizon agentic work. The team describes 1.6 trillion parameters with about 48 billion active per token, reports a BrowseComp score of 91.9 and an MCP Atlas score of 83.1, and says four-bit inference on domestic accelerators sped up prefill by 43.3 percent and decoding by 42.77 percent. Those are the team’s own figures; independent replication will tell us how they hold.

Why it matters to you: BrowseComp measures how well a model can hunt down hard-to-find information on the open web, which is precisely the skill you want in a deal hunter. When open models score this high, shopping agents stop being a premium feature and start being a default.

The Plumbing: Checkout Buttons Built for Software

The rails are already being laid. In late September, Stripe wrote that agentic purchases made with its Link wallet had grown 38 times in the previous month. Stripe also announced in September that Meta’s Muse assistant shops with Link, and CNBC reported in October that Amazon had blocked Muse from its store. On the browser side, the WebMCP draft, a community-group proposal for letting websites expose actions directly to assistants, is moving through a Web Machine Learning community group. Not every store wants a robot at the register, and the fights over who gets to let one in have started.

Your Pre-Black Friday Agent Checklist

• Cap it. If your assistant can pay, give it a wallet or virtual card with a hard limit, not your everyday card.
• Keep the approval step on. Require a confirmation before every purchase, and read the merchant name, total and shipping on that card, not in the chat.
• Ask for the receipt trail. Have the assistant list the store, the price it saw and the link it used, so you can check the deal yourself.
• Check the return policy. An agent can find a lower price at a store with a worse return window. Price is not the only term.
• Watch for blocks. If a retailer refuses assistant traffic, buy there by hand rather than letting software route you to an unfamiliar reseller.
• Compare against a human list. Run your agent’s best find against a curated deal page before you buy; our Deal Desk publishes one most days.

The season’s first lesson is simple. The machines are happy to shop. You are still the one who pays.

Sources

1. Shopify Investors: Shopify sustains rapid growth with revenue up 32% and free cash flow margin of 22%
2. Yahoo Finance: Shopify stock tanks as Q3 revenue, total order value miss Wall Street estimates
3. Android Authority: ChatGPT Dots update adds GPT-6 Astra and GPT-6.1 Terra
4. LongCat blog: LongCat-2.5 release
5. Stripe blog: Helping personal agents shop more intelligently and reliably with Link
6. Stripe newsroom: Stripe helps Meta Muse shop with Link
7. CNBC: Meta’s Muse shopping agent and the Amazon block
8. W3C Web Machine Learning Community Group: WebMCP draft report

Nightlife Report is funded by readers. Back the reporting at patreon.com/aperlas.