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:

OpenAI Dots: Your AI That Keeps Working

What if your AI assistant did not stop when you closed the chat? OpenAI’s new Dots are designed to keep working toward ongoing goals in the background. You give a Dot a goal, connect the apps it needs, and define how much freedom it has to act. It gets its own cloud computer and browser, can work across connected services, and can bring results back to you for review. At launch, Dots are rolling out to eligible ChatGPT Pro and Business Premium users, with broader availability planned. AI is moving from something you ask a question to something you assign a job.

Instagram Edits Gets an AI Creative Assistant

Instagram is putting an AI assistant directly inside Edits, its video creation app. Instead of simply generating another video for you, the assistant looks at your own Instagram performance data, including views, follows, likes, shares, comments, and viewer retention, then helps identify patterns and suggests ideas for what you might create next. You can ask about hooks, captions, trending topics, audio, scripts, and other creative decisions. For creators and small businesses, the appeal is that the AI is looking at what is actually happening on your account rather than giving the same generic social media advice to everyone.

Fireflies Talk: Just Say What You Want to Write

Fireflies has introduced Talk, a free voice dictation feature that works across your desktop applications. Hold a key, speak naturally, and Talk turns what you say into clean text wherever your cursor is, including email, Slack, Google Docs, ChatGPT, and CRM applications. It removes filler words and false starts, adds punctuation, and adapts formatting to the application you are using. Talk is available on Mac and Windows and is included even with the free Fireflies plan, making it an unusually accessible way to replace some of the typing that fills up a workday.

Microsoft Copilot Becomes More Than a Chatbot

Microsoft has redesigned Copilot around three experiences called Home, Code, and Autopilot. Home brings Chat and Cowork together and integrates Word, Excel, and PowerPoint directly into the Copilot experience. Code lets people describe an application, dashboard, tracker, automation, or workflow in natural language and have Copilot build it. Autopilot goes a step further by acting as a persistent digital teammate that can continue working on recurring tasks without waiting for another prompt. The rollout is beginning gradually, but the direction is significant for small businesses because the goal is to make building useful internal software possible without starting with traditional software development.

Wabi 2.0: An AI That Builds the Interface You Need

The new Wabi 2.0 experience combines messaging, AI agents, and software that can appear when you need it. Instead of forcing every task into a conversation, Wabi can create a small interface for the job itself, such as a tracker, calendar, list, map, or planning tool. It is designed around everyday activities such as organizing a trip, managing family plans, handling errands, and collaborating with other people. Wabi 2.0 is currently invite only, but the concept is worth watching because it asks a different question about AI: instead of giving you another app to learn, could the AI create the app you need at the moment you need it?

OpenAI Dots: An AI Assistant That Keeps Working After You Leave

OpenAI is taking ChatGPT in a new direction with Dots, a new type of AI agent designed to keep working toward a goal after you stop interacting with it. Dots are powered by GPT 6 Astra and run on their own cloud computer, giving them access to a web browser and more than 4,000 connected apps. Instead of asking ChatGPT one question at a time, you can give a Dot an ongoing objective and let it continue making progress in the background.

The practical difference is that Dots are designed around goals rather than individual prompts. OpenAI gives examples of using them for things such as research, coding, data analysis, content creation, and other work that can require multiple steps. A Dot can work through those steps independently, monitor information, and return with progress or completed work rather than requiring you to remain in the conversation. Dots can also carry context between ChatGPT and services such as Slack and Microsoft Teams.

There are also controls designed to keep the user involved when an action matters. OpenAI says Dots can operate with customizable permissions and approval rules, including requiring approval for sensitive actions. That distinction could make Dots particularly interesting for small businesses and creators. Instead of simply asking AI to write a report, for example, you could potentially give it a continuing research or content task and have it work through the process while you spend your time on something else.

Dots are currently rolling out to ChatGPT Pro, Business Premium, and Enterprise users. OpenAI currently allows one Dot per user, with plans to support multiple agents in the future. For everyday AI users, the bigger idea is worth watching: ChatGPT is moving beyond a tool that waits for your next question toward an assistant that can take responsibility for an ongoing task and come back when there is something useful to show you.

What You’re Actually Paying For

Just days ago, OpenAI reopened its $200-a-month ChatGPT plan. Same price as before. Half the value — by OpenAI’s own math.

For roughly three weeks, that plan had been closed to new sign-ups entirely, with OpenAI blaming “unprecedented” demand for GPT-6 Astra, its most capable model. When the doors opened again on September 29, the deal on the other side had changed. New subscribers get a smaller included allowance. People who already had the plan keep the old allowance through October 29, then move to the smaller one too.

How much smaller? Measured at the rates OpenAI charges developers, the new allowance is worth about half the old one. That isn’t the same as half the work — OpenAI argues its newer, cheaper models will let subscribers get more done than before. Maybe. But it’s OpenAI’s argument, not a promise.

There’s a sweetener. The five-hour cap that used to throttle heavy users is gone, so you can spend your week’s allowance whenever you like. And OpenAI added a brand-new $500 tier on top.

The most revealing part is OpenAI’s explanation. The company says it would rather keep cutting the price of using its models than prop up flat subscriptions that look generous. That’s a polite way of saying the subscription is slowly turning into a meter.

It’s the most visible sign of something bigger that almost nobody has explained to regular subscribers: the AI subscription you’re paying for is quietly changing shape.

The deal you thought you had

For the last few years, the AI subscription pitch was simple. Pay about $20 a month. Get the good model. Use it as much as you reasonably want.

There were always limits in the fine print, but casual users rarely ran into them. It felt like a streaming service: one price, all you can eat.

That deal is fading. Not through a dramatic price hike — the headline numbers have barely moved. ChatGPT Plus is still $20 a month. Claude Pro is still $20. Google AI Pro is still $19.99. (All U.S. monthly prices, before tax.)

As OpenAI just showed, though, the price can stay exactly the same while what’s inside it shrinks. What’s changing is what happens inside that $20. And it’s taking three distinct forms.

Shape one: the meter

The first change is the simplest to understand, because it’s the oldest idea in the book. You get an allowance. Past the allowance, you can keep going — if you pay by the drop.

OpenAI now sells credits on its personal plans — yes, personal, not just business. Your plan’s included usage gets used first. After that, if you’ve bought credits, certain features start drawing from that balance: ChatGPT inside Word, Excel and PowerPoint, its coding tool Codex, and its more hands-on ChatGPT Work mode. OpenAI’s own estimate is that a typical Excel task runs 5 to 20 credits per message, and a PowerPoint task 10 to 50.

Read that again. Per message. That’s a subscription with a taxi meter bolted on.

Anthropic works the same way. Claude subscribers can turn on usage credits, prepay into a balance, and keep working after they’ve hit their plan’s limit. It’s opt-in — nothing gets charged unless you switch it on — but there’s also an optional auto-reload that tops up your balance automatically whenever it runs low.

That’s the important distinction. Hitting a limit doesn’t put a charge on your card by itself. But once you’ve added a card and turned on auto-reload, the meter can run without you watching it.

Shape two: the shifting wall

Google went the other way. Instead of a meter, it built a wall — one that moves depending on what you’re doing.

Since May, Gemini has used what Google calls compute-based usage limits. There’s no longer a fixed number of messages you can send. Instead, how far your allowance goes depends on how complex your prompt is, which features you use, and how long your conversation has gotten. It refreshes every five hours until you hit a weekly cap.

In plain English: you can see how much of your allowance is left — Google shows it under Settings → Usage Limits and warns you when you’re getting close — but you can’t translate that into a predictable number of questions. A quick question costs almost nothing. A long, detailed back-and-forth with a file attached costs a lot more.

Early users hit the wall fast. Google loosened the limits within weeks after complaints. And when paid subscribers do hit the cap now, they aren’t simply locked out: they can keep chatting with Flash-Lite, a smaller, cheaper model, until the allowance refreshes.

Here’s the part that tells you how new all this is: Google’s own pages don’t agree with each other. The subscription page says you can extend your limits by buying AI credits. Google’s credits help page says those credits work in two other products — its video tool Flow and its coding tool Antigravity — and doesn’t mention the Gemini app at all.

If the company can’t keep its explanation straight, it’s not surprising customers can’t either.

Microsoft has a simpler wall, and a lower one. If you pay for Microsoft 365 Personal or Family, you get 60 AI credits a month for the AI features built into its apps — rewriting in Word, summarizing in Notepad and the like. They’re included in the subscription. But when they run out, those features stop working until next month’s allowance arrives. And on a Family plan, only the account owner gets them.

Shape three: the velvet rope

The third change is the newest, and the one that would have sounded absurd two years ago: the premium model isn’t included in your plan at all.

Anthropic is the cleanest example. Its premium models, Claude Fable 5 and Fable 5.1, come with the more expensive Max plans — though even there, they can use no more than half your weekly allowance. On the $20 Pro plan, they’re not part of your usage at all. You can use them only with usage credits. When Fable 5 made that switch in July, eligible Pro subscribers could claim a one-time $100 credit to soften it. That credit expired in mid-September.

OpenAI’s new top tier takes the idea a step further. The $500 plan is the only personal plan that includes Astra Ultrafast, a faster service tier for GPT-6 Astra. It isn’t available on the $100 or $200 plans — not even if you buy extra credits. The rope isn’t just in front of the best model anymore. It’s in front of the fast lane.

The good news hiding in the fine print

Here’s where it would be easy to write the angry version of this article. But it wouldn’t be the true one, because while the top of the market is getting metered, the bottom is getting better.

• Google AI Plus dropped to $4.99 a month in June and doubled its storage.

• Google’s Ultra range, which used to start at $249.99, now includes a smaller $99.99 entry tier.

• ChatGPT Go gives you a meaningful step up from free for $8.

And there’s a pattern worth noticing. Late last month, Anthropic released Claude Opus 5.5. Anthropic says it performs at the level of Fable 5.1 on most work, and beats it on some tasks. It’s included in the regular $20 plan, and Anthropic raised Pro’s five-hour usage limits at the same time.

That’s the real shape of the market. The premium model shows up behind the rope, on the meter, or in a tier priced for almost no one. Then, so far at least, something nearly as good tends to show up in the plan you already have.

Cheap everyday use keeps getting better. Heavy use and early access are what’s going on the meter.

How to tell what you’re buying

None of this requires you to cancel anything. It just requires you to stop assuming your subscription works the way it did two years ago. A few habits help.

Know what happens at the wall. Every plan now has one. When you hit it, one of three things happens: you wait, you get switched to a smaller model, or you’re offered the chance to pay for more. Find out which one yours does before it matters.

Know which credits are which. “Credits” can mean an allowance that comes with your plan, a balance you’ve paid for separately, or both. Included credits are part of what you already pay. Purchased ones are the meter.

Check your auto-reload. If you’ve ever added funds or bought credits, look for an auto-reload setting and a monthly spending limit. Auto-reload is where a few dollars can quietly turn into a lot more.

Find the usage page. Each of the big assistants now shows your usage somewhere in its settings. Look once. It takes a minute, and it’s the only way to know whether you’re a light user overpaying or a heavy user about to hit a wall.

Don’t pay for first. If you’re tempted to jump to a $200 or $500 tier just to get the best or fastest model, ask whether your current plan already does what you need. If it does, wait. Lately, “nearly as good” has been coming down to the regular plans fast.

Check again at renewal. These terms are changing month to month, sometimes without an announcement. What you signed up for in the spring may not be what you’re renewing in the fall.

The new fine print

For a while, AI subscriptions felt like Netflix. They’re turning into something closer to a phone plan — a base price, a data cap, and the option to pay more when you blow past it.

That’s not necessarily a bad deal. But it’s a different one. And it’s easy to miss the switch if you don’t read the fine print.

Now you have.

Partner Spotlight: Periscope VPN by Duet Display

A VPN should not make your life more complicated. Periscope by Duet Display is designed to be easy to set up while keeping your connection secure with state of the art end to end encryption. Periscope is built around a commitment to privacy with absolutely no logging, so your online activity is never tracked or recorded. Unlike most VPN providers, you are connecting to your own computer back at your house to you do not have to worry about your provider’s system snooping on what you are doing. Whether you are working from home, traveling, or connecting to an unfamiliar network, Periscope gives you a simple way to protect your connection without adding unnecessary complexity.

Sign up or get more information at www.getPeriscope.com.

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