In partnership with

Keep up to date on the latest products, workflows, apps and models so that you can excel at your work. Curated by Duet.

Stay ahead with the most recent breakthroughs—here’s what’s new and making waves in AI-powered productivity:

Gemini 3.5 Transcribe

Google just gave voice input a serious upgrade. Gemini 3.5 Transcribe turns natural speech into polished text, removing filler words, handling self corrections, recognizing more than 85 languages and identifying speakers. It is already appearing in the Gemini app for macOS and in Rambler on Android, while developers can also use it through Google AI Studio. For everyday users, the interesting part is that you can talk naturally instead of carefully dictating like you are speaking to a machine.

NewFace AI

Creating convincing social videos and product content usually means juggling scripts, images, video editors and a lot of manual production. NewFace AI is taking a different approach. Its new platform lets creators describe what they want in plain language, then uses an AI agent and visual canvas to build the project. It can create user generated content style videos, ads, social content and product visuals, making it particularly interesting for small businesses that need a steady stream of marketing content without a production team.

Plaud One

What if your AI assistant lived in your ears instead of inside another browser tab? Plaud One is a pair of AI earbuds that can record conversations and calls, transcribe them and use the resulting context to help with tasks. Its charging case can also record in person conversations, while built in cellular connectivity means the device can operate without relying on your phone. Plaud says the system can work with tools such as Gmail, Google Calendar, Notion and Slack, turning conversations into follow ups, notes and actions.

Perplexity Portable Computer

Most AI agents send your work to the cloud. Perplexity is experimenting with flipping that model around. Portable Computer runs its agent system locally on compatible NVIDIA hardware, allowing files and much of the actual work to stay on your own machine. If a task needs a more powerful cloud model, the system asks for permission before sending that step out. It is currently aimed at users with powerful NVIDIA hardware, but the bigger idea is compelling for anyone who wants capable AI agents without automatically sending everything to someone else’s servers.

CreatorsJet AI Sponsorship Discovery

For creators, finding brands that actually spend money on sponsorships can be almost as difficult as creating the content itself. CreatorsJet just added an AI powered sponsorship discovery feature that scans public creator content to identify brands actively working with creators. Instead of starting with a giant list of companies and guessing who might be interested, creators can research brands based on actual sponsorship activity and then decide which opportunities are worth pursuing. That makes it especially relevant for creators and small businesses trying to turn an audience into a business.

Gemini 3.5 Transcribe: Your Voice Just Became a Much Better Keyboard

Typing is still the default way most of us communicate with a computer, but it is not always the best way to get an idea out of your head. Gemini 3.5 Transcribe is Google's newest speech to text model, designed to understand natural speech rather than simply convert every sound into literal text. It can recognize more than 85 languages, identify speakers, understand specialized vocabulary and produce timestamps for recorded audio.

The biggest improvement for everyday users is Smart transcription. Instead of preserving every “um,” false start and verbal correction, it can clean up conversational speech and turn it into structured writing. If you say, “Let’s meet Tuesday, actually no, Wednesday at two,” the system can understand the correction and produce the useful version rather than faithfully documenting the mistake. It can also automatically organize spoken thoughts into paragraphs, lists, dates and other formatted text.

That becomes much more interesting when it is combined with the Gemini app on macOS. Gemini 3.5 Transcribe can use voice together with screen context to perform more complex workflows, including analyzing local files, repurposing text across applications, generating images and searching for information. In other words, voice is no longer just an alternative way to type a prompt. It can become the interface for telling your computer what you want done.

For consumers and small businesses, the practical benefit is simple: you can think out loud without first translating your thoughts into carefully written prompts. Dictate an email while walking around the room, talk through ideas for a social post, explain what you want changed in a document or brainstorm a project without touching the keyboard. The technology is still evolving, but this is one of those AI improvements that could quietly change how people interact with their computers every day.

Your AI Can Buy Things Now

Here’s a question worth answering honestly before reading any further: would you give an AI assistant your credit card and let it spend money on your behalf?

For most people, the answer is an immediate no. Not a considered no — a flinch. Something about handing purchasing authority to software that occasionally invents facts feels like a category error, and no amount of enthusiasm from the tech industry has made that feeling go away.

That instinct deserves more respect than it usually gets. It also turns out to describe the shape of the entire market.

The most-hyped version of this got quietly walked back

In September 2025, OpenAI and Stripe launched Instant Checkout in ChatGPT. Buy directly in the chat window — no tabs, no forms, no leaving the conversation. Etsy sellers first, with “over a million Shopify merchants” promised right behind them.

In March 2026, OpenAI changed course. In its own words: the initial version of Instant Checkout “did not offer the level of flexibility that we aspire to provide, so we’re allowing merchants to use their own checkout experiences while we focus our efforts on product discovery.”

That’s the company’s framing, and it’s about product design rather than shopper reluctance. But the reporting around it found something more pointed: people were researching products in ChatGPT and then completing the purchase somewhere else. Very few merchants had switched the native checkout on at all. One industry analyst quoted at the time offered an explanation worth sitting with — that shoppers still default to the checkout flows they already trust, like Apple Pay, Google Wallet, and Amazon’s one-click. Not because those are technically superior, but because people have years of evidence that money put into them comes out correctly on the other side.

That’s an analyst’s read, not a proven cause. But it matches what everyone reading this already feels.

One correction to the version of this story you may have heard: Instant Checkout wasn’t switched off entirely. OpenAI’s current help documentation still says that for some eligible products and merchants, ChatGPT “may also show an Instant Checkout option that lets you complete checkout in ChatGPT instead of leaving for the merchant’s site.” It stopped being the main event. It didn’t disappear.

The versions that work ask you to trust them with less

It would be tidy to say the other companies watched OpenAI stumble and redesigned around the hesitation. That’s not what happened — the timing doesn’t allow it. Amazon launched its version in April 2025, five months before Instant Checkout existed. Google’s arrived in November 2025, four months before OpenAI changed course.

These were parallel bets, not reactions. But the ones that stuck share a quality, and it’s the useful lens for everything below: they ask you to hand over as little new trust as possible.

So the question isn’t “can AI buy things?” It’s “how much am I actually authorizing, and what happens if it’s wrong?”

Three very different asks

“AI shopping” covers three levels that require wildly different amounts of faith.

Level one: it finds things. You describe what you want in plain language and it produces candidates. “A rain jacket that packs down small, under $150, that isn’t loud orange.” This asks nothing of you financially, and it’s genuinely the most useful of the three.

Level two: it fills the cart. It assembles the order — size, color, quantity, shipping — and hands you a finished checkout page to approve. You still press the button.

Level three: it pays without asking again.

Here’s the thing worth knowing: almost nothing mainstream is actually level three. The headlines suggest otherwise. The products themselves are considerably more cautious than the marketing around them.

Google’s version: you set the conditions, you still approve

Google’s approach is attached to something that already existed and that people already liked: price tracking. You flag an item — specific size, specific color — and name the price you’d pay.

The new part: if the merchant is eligible, a “Buy for me with GPay” option appears. But read what actually happens when your price hits, because it is not what most coverage implies. Google’s own help page: “When the desired price point is reached, you’ll get a notification and email to confirm the purchase, payment method, and shipping address.” And more plainly still: “Before any transaction proceeds, you’ll get a notification to review and approve all details.”

So setting a price target authorizes nothing. You are not pre-approving a charge. Google watches, Google tells you, and you decide — then it handles the checkout mechanics on the merchant’s site.

That’s a smaller feature than “AI buys things for you,” and it’s a much better one. The machine does the waiting, which humans are genuinely bad at, because we forget to check and then buy at the wrong moment out of impatience. The deciding stays where it belongs.

A related one where no money moves at all: search for certain products “near me” and you may see a Let Google Call option. Google phones nearby stores, asks whether they have the thing and what it costs, and texts or emails you the answers. Anyone who has spent a Saturday morning on hold with three sporting goods stores will see the point. Real limits worth knowing: it’s English-only, requires you to be signed in, doesn’t cover most apparel or grocery categories, and is unavailable in Indiana, Louisiana, Minnesota, Montana, and Nebraska.

What’s expanding: at its developer conference in May, Google introduced Universal Cart, a cart that follows you across Search, Gemini, YouTube, and Gmail, rolling out in the US across Search and the Gemini app. Google named Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and Shopify merchants including Fenty and Steve Madden — though the language is hedged, promising these checkout features “soon.” Note also that a retailer adopting Google’s Universal Commerce Protocol doesn’t automatically get a buy button: Google’s merchant documentation says the feature “is available for select merchants at this time,” through an early-access program, in the US, Canada, and Australia only.

Amazon’s version: nothing new to hand over

Amazon solves the trust problem differently — by not asking for anything you haven’t already given them.

In May 2026, Amazon combined Rufus and Alexa+ into a single assistant called Alexa for Shopping, reachable from the main search bar and from its own chat window.

Its agentic feature is Buy for Me, and the notable thing is what it buys: products from other retailers’ websites. Search for a brand Amazon doesn’t carry and you’ll see results labeled Shop Direct — as of Amazon’s most recent figures, more than 100 million products from over 400,000 merchants, with tens of millions of those eligible for Buy for Me.

Crucially, this isn’t hands-off either. You confirm your order details on Amazon’s own checkout page — delivery address, taxes, shipping fees, payment method — before anything is purchased. And know who owns the aftermath: the outside merchant handles delivery, returns, exchanges, and customer service, not Amazon.

Whatever you think about Amazon having your card on file, you made that decision years ago. This feature doesn’t ask you to make it again — which is the whole trust argument, and probably why it worked where the chat-window version didn’t.

Sit with the strangeness for a second, though: Amazon built a feature whose main job is helping you buy things that aren’t on Amazon. That tells you how badly the company wants to be where shopping starts, regardless of where the money lands.

(Amazon also has a separate “Auto Buy” feature for buying at a target price — but that one applies within Amazon’s own store, not to outside merchants. Different tool, easy to confuse.)

Two more that are actually live

Google and Amazon get the attention, but they aren’t the only options a US shopper can use today.

Copilot Checkout began rolling out in January 2026 on Copilot.com and has since reached the Copilot mobile app, with merchants including Urban Outfitters, Anthropologie, Ashley Furniture, and Etsy sellers. It works — the honest caveat is that far fewer people are in Copilot to begin with.

Perplexity’s Instant Buy lets US users buy from merchants directly inside Perplexity, with PayPal as the payment partner. It’s available on the free tier — no subscription required.

Where your money actually goes

The most common fear here is “the AI will have my credit card number.” That is mostly not what happens.

Google uses Google Pay. Amazon uses the payment method already in your account. And for the ChatGPT-style flow, Stripe built something called a Shared Payment Token — a credential that OpenAI’s payment specification describes as “single-use and constrained,” restricted by a maximum amount, a specific merchant, and an expiry time. (The broader standard underneath, the Agentic Commerce Protocol, was co-developed by OpenAI and Stripe; the token itself is Stripe’s.)

The design principle across all of them is the same: keep reusable card credentials away from the AI model.

That’s genuinely reassuring, and it’s worth being precise about what it does and doesn’t solve. It sharply reduces credential exposure. It does not eliminate account takeover, a purchase you approved too quickly, or a merchant behaving badly. The nightmare scenario most people picture — an AI loose on the internet with your card number — isn’t the actual risk. The duller risks below are.

The fears that are justified

It buys the wrong thing, and the return is your problem. The purchase takes seconds. The return takes forty minutes, a printer, and a trip — and with external purchases it’s the merchant’s process, not your assistant’s. That asymmetry should make you slower, not faster.

Friction was protecting you. The pause between wanting something and entering a card number is where a lot of purchases die, and many of them deserved to. Removing that pause is sold as convenience. For some people it will function as a spending problem with a nicer interface. Know which kind of shopper you are before automating anything.

The agent has no opinion about whether you should buy at all. Ask it to find the best option and it will find one, every time, whether or not “none of these, wait six months” was the right answer. Purchase-or-don’t isn’t a question it’s built to ask. Keep asking it yourself.

Know which results are paid — and the answer differs by platform. This is where it pays to be specific rather than cynical. OpenAI states that ChatGPT’s “product results are selected independently by ChatGPT and are not ads, nor influenced by any OpenAI partnerships,” with ads kept separate. Google, by contrast, has confirmed that ads can appear inside an AI Mode recommendation list as a “Highlighted Answer,” labeled “Sponsored.” And Amazon has said existing sponsored-ad campaigns are automatically eligible to appear within Alexa for Shopping conversations. None of that is disqualifying. It does mean you should look for the label, and compare anything expensive somewhere else.

Read the retailer’s confirmation email, not the assistant’s summary of what it did. Quantity, size, address, delivery date. Agents make ordinary mistakes, and the confirmation is where you catch them while fixing it is still cheap.

And keep the approval step. Every mainstream version currently asks you to confirm before charging you. If a tool ever offers to skip that to save you a tap, decline. It’s one of the most important safeguards you’ve got.

How to try it without handing over anything you’d miss

Start where no money moves. Use ChatGPT or Gemini for the deciding, not the buying — comparison, narrowing, “what’s the actual difference between these two.” Nearly all the time savings have been hiding here anyway.

Skip the errand, not the decision. Next time you need something local and urgent, try Let Google Call instead of phoning four stores.

Then a small real test. Set up price tracking on something you want but don’t need, with a target meaningfully below the current price. When the alert arrives, you’ll review the total, payment method, and address before anything happens — which is a useful way to see the whole flow with your hand still on the wheel.

If you try Buy for Me, use something under $30 from a brand you already know, and remember the outside merchant owns the return.

What’s coming, so you’re not surprised

In June, Visa and OpenAI announced infrastructure for agent-initiated payments using tokenized Visa credentials, operating within “spending limits, merchant categories or required approvals.” No consumer launch date was given, and the language throughout is forward-looking. But it’s a signal: the confirmation-every-time model is the current default, not a permanent one. When something offers you a standing budget instead of a per-purchase approval, that’s the moment to re-read this section.

The part worth remembering

Buying was never the hard part. Entering a card number takes fifteen seconds, and nobody was lying awake wishing those fifteen seconds were shorter.

The hard part is deciding — sorting a thousand options down to three, knowing what actually differs, judging whether now is the moment. That’s where these tools earn their keep, and it’s the part they’ll do all day without ever touching your money.

Hand off the searching. Hand off the waiting. Hand off the phone calls.

Then keep the last approval for yourself. It costs you almost no time, and it’s the point where a mistake stops being a suggestion and becomes a charge.

Partner Spotlight: Periscope VPN by Duet Display

Your Internet connection should not require you to hand your privacy to someone else.

Periscope by Duet Display is an easy to set up VPN designed around privacy and security. It uses state of the art end to end encryption to protect your connection while keeping privacy at the center of the experience.

There is no complicated configuration and absolutely no logging. Periscope is built to give individuals, families and small businesses a straightforward way to secure their Internet connection without giving up control of their data.

Learn more at getPeriscope.com

Holiday Creator Calendars Are Filling Up. Q4 Panic Is Optional.

Creators lock in their holiday content calendars 90 days out, before most ecommerce brands finalize their commission strategy and way before Black Friday and October deal events.

Get ahead of the seasonal rush with The 90-Day Holiday Sprint, a practical guide for brands that want creators driving holiday demand while competitors are still recruiting:

• Structure commissions by lifetime value, not just first-order margin

• Lead with the right products so creators promote with confidence

• Recruit and onboard creators with a day-by-day plan for the first 30 days

• Read performance early and pull program levers by Day 60

• Brief creators with a holiday checklist before calendars fill up

Your 90-day countdown starts now.

Get the Free Guide

Stay productive, stay curious—see you next week with more AI breakthroughs!