Ep 773: New ChatGPT default model, Copilot Cowork goes mobile, Codex heads to Chrome and 7 more AI upgrades worth checking out

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life.

You know what most companies pitching AI agents aren't actually doing? Deploying AI agents. Sage just did the opposite at Sage Future. They expanded their AWS collaboration to put real AI agents into real finance workflows built on Amazon Bedrock Agent Core and available in the AWS Marketplace. So that means enterprise-grade AI for your finance team without the IT headache. Check it out at sage.com. Everything got more powerful this week. Even though we didn't get a new frontier large language model over the past few days, what we can accomplish we accomplish with our daily driver devices really did change. I mean, this week your browser, your iPhone, your Google Docs, and your physical computer all got a little more powerful because of some under-the-radar AI updates that you probably missed. So if you don't have hours every single day to check what's new and understand what new capabilities that you have, well, Don't worry, that's what today's show in our new-ish Friday Features show is all about, where we bring you the new features in AI that you can actually start using today that are going to change your workflow. So on today's show, here's what you're going to learn. And well, stick around and you're going to not only have a bunch of new workflows, but you're going to understand why AI text slop could be on the decline. Bless up. You're going to find out about the new way that you can copilot on the go and the more secure and easier way to use an OpenClaw-esque agentic system on your computer that now millions of more people are going to have access to. All right, if you're ready to get into it, into it, I am too. Let's go. My name is Jordan. Welcome to Everyday AI. This is your daily livestream podcast and free daily newsletter helping everyday people like you and me, not just keep up with all these updates, but how to actually use them and to grow your company and your career. So it starts here with the unedited, unscripted livestream podcast. But to be the smartest person in AI in your department, at your company, whatever it may be, our website is your new best friend, BFF. All right. YourEverydayAI.com. Make sure you go there, sign up for the free daily newsletter. We're gonna be recapping all 7 tools from today's show, uh, all those updates as well as all the other AI news that you need to know to stay ahead. All right, let's get to it. So first, let's get right into it, uh, this one Chrome extension that I've personally been waiting for for a very long time, and I think it's going to maybe open up a lot of people to Codex, because if you've been sleeping on Codex, you are missing out. So here's what's new. OpenAI released a Codex plugin for Google Chrome that lets the Codex coding agents operate inside of the actual Google Chrome profile that you use. So this was announced just yesterday, and it can work in parallel across tabs in the background while the user keeps using the browser or read and act on signed-in sites. Like Gmail, LinkedIn, Salesforce, whatever. So technically this is not new capability because Codex actually has a built-in browser. And to even rewind a little bit more, because you might be wondering like, what is Codex, right? If you've heard of Claude's desktop tools, right, they have Claude Chat, Claude Cowork, Claude Code. All great and all, except those three things don't talk to each other. They have no clue that any of them exist. They don't carry context over. Codex is a unified one platform Does it all, in my opinion, way better benchmarks for the most part, you know, hit or miss, right? Depending. But I love Codex and this makes it even better because yes, before Codex did have an in-app browser, which is bonkers. So good. So useful. Right. But sometimes I have all these Chrome extensions that I like to use. And, you know, one of the downsides of having an in-app browser browser inside Codex, right? The good thing is, well, it's way faster because it's all contained kind of within the harness of Codex. But the downside, aside from all of those, you know, Chrome extensions that you don't necessarily get access to inside of Codex, is just being able to stay logged in across sessions. So if you do, you know, close Codex as an example and go, you know, open up that in Codex browser, it doesn't stay logged in. So a lot of cool use cases, but I'm actually going to kind of scroll through here on the Twitter thread that Codex, that OpenAI put out. So here's what they said. They said Codex now works directly in Chrome on macOS and Windows. It's even better at working with apps and sites in Chrome and now works in parallel cross tabs in the background without taking over your browser. That's the big thing and what separates it from anything you know, sorry, anything Claude, because Claude unfortunately does take over your browser. So it says with the new Chrome extension, Codex could quickly move through repetitive browser work like navigating structured pages and complex data entry flows. Under the hood, it writes and runs code to navigate and complete tasks. If a task needs multiple tools, Codex chooses the best one for each step. It uses plugins when they can handle the job. Chrome when it needs a logged-in website and combines approaches as needed. To me, that's the coolest part, right? So it's not always going to use, right? Like if you're having it carry context, let's say from your LinkedIn account to Twitter to your email, whatever, right? I'm just, you know, naming out tools like a lot of you all probably use on a day-to-day basis. It's not going to go and use your gmail.com if you have a Gmail connector. Inside of Codex, right? So it's smart enough where it's going to handle the job. Another big advantage here, right? And sometimes the advantages, you really only know them when you're comparing, you know, this new process or improved process versus its main competitor, which in this case is Claude via either computer use on the desktop program or in the Claude Chrome extension. So that's the other thing, the Claude Chrome extension. Doesn't talk to the Claude desktop app. Two different things. And when you close the, uh, the Claude Chrome extension, uh, it just loses history of whatever you were previously working on. Uh, I was actually using it, um, yesterday before this new update came out. And when this new update in Codex came out, I was like, yes, finally. Uh, so this is a big one. So who has access? It's available, like I said, right now on macOS and Windows. You just— and install the plugin from the Codex app and then you connect the Chrome extension. So Codex itself, it's not a separate plan. It's already included if you have any paid plan. So ChatGPT Plus, Pro, Business, Edu, and Enterprise. Here's why it's useful and who's going to find it valuable. So it removes the gap between Codex and tasks that require a logged-in browser. Or if you use Chrome extensions like I do and rely on them heavy— heavily, I rely on my Chrome extensions very heavily. So that's great. So, you know, you can go and also for those, you know, apps that you use on an ongoing basis or, you know, platforms that you use that don't have a connector and maybe they don't even have an MCP access, this is a great way that you can kind of juggle those tools, right? Because the Codecs itself, being able to talk to your Chrome browser can decide when it's going to use the tools that it already has. not just the connectors but the actual tools, right? Running Python, you know, APIs, etc., but also then navigating your browser as it needs. All right, our next big, uh, announcement, and I'll say this one big— some of these are under the radar, um, but I think this one is actually a pretty sizable one. That's because, uh, Copilot Cowork has gone mobile. Uh, so if you don't know what Copilot Cowork is, right? Microsoft is a big investor in Anthropic, and Anthropic's Cowork has obviously been a very popular tool in 2026 so far. So this is essentially like Microsoft's version, literally, of Anthropic's Cowork. They are using Anthropic's kind of Cowork technology, but all of their, um, infrastructure on top of it. So the new Microsoft Cowork Agent is now in the Microsoft 365 Copilot mobile app. So this was announced alongside reusable co-work skills and a plugger— plugins connectors expansion. So here's who has access. You do have to be in the Frontier program in Microsoft, which I know most enterprise companies are. So if you're an enterprise company, you probably already have access to this. And also Microsoft says broader availability is targeted around the July Inspire conference. So you do obviously have to be in the Frontier program and obviously have a Microsoft 365 Copilot Premium license. So the actual mobile app that runs on iOS and Android via the existing Microsoft 365 Copilot app. So essentially, if you are already a Copilot app user, Essentially, this just brings in a bunch of new kind of co-work capabilities within, within that app. So here is kind of what Microsoft says about it. So they said, work doesn't just happen at your desk. Bringing co-work to iOS and Android mobile, it says a key part of our Copilot vision is bringing AI into the flow of work wherever that work happens. Co-work already runs in the cloud, so you don't have to worry about closing your laptop or if your PC is running. Now we're introducing Copilot Cowork on iOS and Android so you can delegate work the moment you think of it on your commute, between meetings, or away from your desk and come back to a finished outcome. Instead of completing those tasks, you can hand them off and keep going while the work progresses in the background. And then obviously they added support for skills as well inside of Copilot Cowork. So Who— here's why it's useful. Well, now you can do a lot of what you were accomplishing in Copilot Cowork on the go. That's the biggest thing, right? And if you haven't checked out Copilot Cowork, we did cover that a couple of weeks ago in the Friday Features section. It was actually one of our first weeks that we covered Copilot Cowork after we kicked off this new weekly session. So I mean, anyone that's going to find it valuable, right? So if you're a Microsoft 365 shop and you have access to Copilot Cowork, I mean, this is an instant upgrade in what you can accomplish on the go. You know, I don't know about everyone else, but I actually have multiple computers that I use. And this is one thing I always struggle with is, well, I don't have one, you know, at least on my phone, I can't, you know, connect to different computers. And that's the cool thing. That's a little bit different about the Microsoft Copilot COWORK version versus the, you know, Anthropic COWORK version. The Anthropic COWORK version, it requires your local machine. The Copilot COWORK version is a little bit different. It works in Microsoft's cloud. So think of it as in this way, Copilot COWORK, if you haven't used it, think of it as kind of like a virtual machine, right, that has access to all of your Microsoft 365 data that you can just— it can run and do work in the background. So it's more like a virtual machine running Copilot COWORK, but now you have a remote connection to it with the iOS app. So This one, it's kind of confusing because you first have to, you know, explain, well, this is different than normal Copilot, but it's in the 365 app, but it uses Anthropic technology, but now it's on the go. So I actually think, right, this is better for people like me because I was even giving a presentation last week, a keynote on Copilot, or sorry, on Claude Cowork, and I couldn't really showcase certain, you know, phone to computer capabilities because it was originally synced to a different computer. So I'd have to unlink it and do all these other things. So, you know, pretty cool, especially if you're a multiple computer user. But regardless, if you are a Microsoft 365 organization, if you're in the Frontier program, Copilot Cowork, I actually think it's a really good setup. And now you can, you know, always be connected. One more reason to keep working, right? Sometimes I don't know about you guys, sometimes all these tools And then it just, to me it's like, oh man, well now I have to be more productive because all of a sudden, you know, I can talk to my Apple Watch and, you know, it's gonna start my computer and do all my work for me. Right? So it's almost like, it's kind of like this concept of FOMAT, right? Fear of missing agent time that I talked about a couple of days ago. Right? But now it's just the ability to control all of these work processes is getting easier and easier and more agentic and coming to more platforms and more devices. All right, our next feature that you might have missed. Claude has some new finance agent templates. And yes, we're covering templates this week because if you work in any of these kind of verticals, you are definitely going to want to use these. I think this is one area where Anthropic has really carved out a niche. All right. People always say like, oh, you're hard on Anthropic. No, I'm not. I just tell the truth, right? Like, as an example, Codex is just way better than anything Claude CodeCloud Co-Work. And I think finally the rest of the world has been figuring that out, even though I've been saying it since the first week of February. But one area where I think Anthropic is continuing to clean up is just in these niche verticals. And that's why, you know, sometimes when they, you know, especially when they release the plugins, some of the original skills, right? You saw some legacy SaaS companies really get hit on the stock market. So it'll be interesting, you know, these have only been out for a couple of days. It'll be interesting if we see a similar, reaction here as well. But let's talk about what's new. Anthropic released 10 pre-built agent templates aimed at financial services. So the 10 templates are Pitch Builder, Meeting Preparer, Earnings Reviewer, Model Builder, Market Researcher, Valuation Reviewer, General Ledger Reconciler, uh, Month-End Closer, Statement Auditor, and KYC, Know Your Customer Screener. So each template packages 3 things: skills, uh, right? So that's kind of your workflow instructions, paired with your domain knowledge. So skills, connectors, and subagents. So who has access? So anyone, if you're on a paid Claude plan through Anthropic's Financial Services Marketplace. So not your standard, you know, Claude plan, but you have to go into the Anthropic Financial Services Marketplace on GitHub. So there's two deployment paths. So you can deploy these as plugins inside of Claude Cowork or Claude Code, or you can do them as cookbooks for Claude-managed agents, which are available on the Claude platform in beta. So Anthropic recommends pairing the templates with Claude Opus 4.7, you know, because why not? Now you can get through your— here, I had to make a joke about this one. Now you can get through your weekly limits, right? So let me actually just pick a random bone here since we're talking about like, oh, of course, Claude is, or Anthropic's gonna recommend you use Opus 4.7. So one thing they actually did, which I found humorous, Anthropic had their Dev Day this week and you know, the big headline that came out of it was, oh my gosh, Anthropic doubled their limits, right? Which I think still even at the doubling limit rate, I don't, I still don't even think the, maybe the $100 plan at double, quote unquote, true double limits, but we didn't get it one second, still you might get what you get on the $20 a month ChatGPT plan, but I still think the ChatGPT plan gets more weekly limits than the $100 a month Anthropic plan. Anyways, you didn't actually get double limits. You got double 5-hour windows, but the weekly limit is the same. So kind of the joke I made, it's like they're, they're like, all of a sudden it's like, oh, let's just say you're on a diet and you are on a 2,000-calorie diet, right? And they're like, okay, Now you can eat whatever you want at any meal, but the, you know, the $2,000 calorie, it doesn't change, right? So it's like, oh, double the rate limits over a 5-hour period. It actually doesn't matter. So, because your limits don't actually go up, it's just a way for people to pay more money. It's a sneaky way, right? All right. Anyways, these agents though are actually really cool. And here's why it's useful because it compresses kind of the standup time for a finance agent. From many weeks to, well, maybe days or hours since the workflow logic, data connectors, and approval flows are already pre-wired for you. So each template— here's the thing that is important to keep in mind— they can be customized to your firm's own modeling conventions, risk policies, and approval routes. This also does pair with the brand new Cloud add-ins for Microsoft 365, you know, so that's already out for Excel, PowerPoint, Word, and then Outlook, I believe, is in beta. So then the context queries from a model into Excel, into a deck in PowerPoint without having to re-explain it. So I think that's probably the big kind of small print there. So, you know, especially if you're a heavy 365 organization, obviously I think it was a smart move for Microsoft to start offering both Anthropic models and OpenAI models in their different products and services. But this right here, being able to use these 10 different agents inside the actual platform where a lot of this work gets done, is a big step forward. So here is what Anthropic said about their release. So they said, we're releasing 10 ready-to-run agent templates for the most time-consuming work in financial services: building pitchbooks, screening KYC files, and closing the books at month end. Uh, each one ships as a plugin in Claude Cowork and Claude Code and as a cookbook for Claude Manage Agents, so a team can put Claude on real financial work in days rather than months. Claude also now works across, uh, Microsoft Excel, PowerPoint, Word, and Outlook, coming soon through the Claude add-ins for Microsoft 365. Once the add-ins are installed, context carries automatically between applications, so work that starts in a model can end in a deck without re-explaining anything in between. So pretty big news, especially if you work in the finance space, uh, because I think this one is again gonna be one that is fairly popular, uh, for people that work in those spaces. All right, next, kind of a very small one that even I almost missed, but I actually ran into this. I didn't even see the announcement from Google. I just showed it pop up in my Google Docs and I'm like, okay, if you're a heavy Google Docs user like me, you've gotta know this one. And I almost missed it even though I'm constantly looking out for these things. So here's what's new. Sounds small. I'm loving it so far. It's custom instructions for Gemini in Google Docs. So here's what it means. Google just added persistent custom instructions to Gemini inside Google Docs so users can set rules once and then have Gemini apply them to every interaction in Google Docs. Not just every, you know, you don't just do it once per document. You set this up once and then it is applied for any and every single doc that you work. So this has been rolled out or will be rolled out gradually over the next few weeks. So instructions kick in on Gemini's existing doc tasks. So drafting, rewriting, summarizing, and the new kind of help me create kind of prompt that you may see pop up at the bottom of Google Docs. So who has access right now? This is US only and English only for now, and it is available to business enterprise and education accounts, and you have to have the Gemini Alpha features enabled, which I think most companies do, but make sure you have that, uh, and you do need a paid account for now. Quick pause. Want to know where AI ROI actually lives? It's not the flashy demo. It's the work nobody wants to talk about. At Sage Future, Sage tackled the painful side of finance system rollouts, implementation, financial data migration, chart of accounts mapping, configuration, the work that stalls real transformation projects. Sage is using applied AI to make that faster, more accurate, and auditable with humans staying in control. That's responsible AI doing real work that actually moves time to value. If your finance team is staring down a migration, check out Sage at sage.com. So here's why it's useful. Well, if you have used Gemini inside Google Docs, it's actually gotten really good. I think when they first rolled it out like a year and a half ago, it was pretty bad. It just didn't work. I actually did a bunch of like video reviews back when I was, you know, doing more kind of of little snippet reviews on YouTube. And I was like, this thing just doesn't work. And now Google, to their credit, has fixed it and it's actually really, really good. So anyone who's a heavy user of Google Docs and you constantly find yourself like, ah, I don't want to leave to jump over to, you know, Gemini or Claude or ChatGPT or Copilot and then bring all this back in there because I know I need to have it ultimately live as a doc. It's a big plus. Although, So, you know, Google Gemini last week, which we covered, finally, finally, finally has the ability now to create actual docs within Google Gemini. So that takes away one workaround. But for me, I still spend a lot of my time in Google Docs, whether it's writing things up for clients, whether it's, you know, sometimes I just like to brainstorm in, right? Sounds crazy, right? Like, oh my gosh, this Jordan guy's weird. He's using a blank Google Doc. Yeah, sometimes I'd like to look at a blank page. So it's nice to have this in there. So yeah, if anyone who's drafting client work in Google Docs and you kind of want that consistent voice, you know, certain words that you need to use, words you don't want to use, I think it's just going to help you get locked in a little bit quicker and still be able to use Google Gemini and kind of collaborate with yourself and Google Gemini at the same time in your voice. All right. I like that one. Might be a small one, but all right, next one could be big. Well, we'll see how the credits go, but Perplexity Personal Computer is now available to all Mac users. You know what's crazy to me is when these things first get announced, like, I don't understand a lot of the companies, like, go to market with some of these splashy new features, right? Because when Perplexity released Personal Computer, right? They put all these resources behind it, right? Um, but it was only to Max subscribers, which is, I'm guessing, a very small portion of their audience. And then they only released it via a waitlist. So now all of that is gone. This is now Personal Computer that's open to all paid Mac users. No waitlist. You don't have to be on the $200 a month. So if you're on the base $20 a month plan, but it was like one little blog post and then I think one tweet, right? That's all where the original, like, I don't, I don't get that yet. Like, I don't know if, if, if companies just want to create all the hype and then people just forget about it when you can actually use it unless you're, well, reading our newsletter every day or listening to our Friday features update. So you would have missed it, but now you can go check it out. So here's what Perplexity Personal Computer is, but now it's available. So we actually covered this one. So I'm going to go through this one a little quicker, but it's essentially Perplexity's local cloud hybrid AI agent. So think of it like an OpenClaw, but a little safer, right? Because it is more refined and it can run locally on a computer and it can control your computer more or less. Right. So this is now just available in the Perplexity Mac app. So that's it's it's just going under the general Perplexity Mac app. So if you have that, make sure you update it and then you can check it out. So, yeah, this was announced originally a couple of months ago, but like I said, it was only for those on the $200 a month and you had to go on a waitlist on top of that. So Perplexity Computer essentially just extends— sorry, Personal Computer essentially just extends Perplexity Computer onto the user's own machine. Let me like rewind and tell you what that actually means. Yeah, kind of confusing because they're named kind of the same. So Perplexity has something called Computer, um, that is confusing because it's online, right? It's their web version of something like OpenClaw, an autonomous agent that I think uses 19 different models, and it's actually really, really good, right? So that's Perplexity Computer, it's a web version. And then Personal Computer is the version that runs on your machine and can run your machine. So that's what we're talking about here. But essentially personal computer extends Perplexity computer onto the user's own machine, and then it has access to all the local files, native Mac apps, connectors, the web, Perplexity servers, all that good stuff, right? So yeah, like I said, if you are on a paid Perplexity plan and you're on the Mac, you can download it now and, you know, give your computer agent capabilities if you want, right? It also brings that, you know, kind of that multi- model setup to your actual computer as well. But the downside of that is, well, you're gonna hit your limits on your Perplexity $20 a month plan very, very quickly. So my guess is Perplexity's rolling this out to everyone because it is very powerful. It is very safe. It's very smooth, right? The big downside with OpenClaw is you could spend 20 hours on it a week and still be like, ah, it's broken again. Or it's, it's not really right. Don't be wrong. It's amazing. It's open source. I set it up. But for me, I don't have that 5, 10, 15 hours a week to continually be tweaking and trying to improve it. So something to me, you know, Perplexity's personal computer is appealing except the cost, right? So it's, you kind of have to choose if you, you know, are concerned about costs. Well, you can go the OpenClaw route because you can use your own models. You can choose which model to use. You can have open source models, right? I have a beefy Mac Studio, so I can use those. But if you use personal computer, you don't. Personal computer chooses for you and you're going to run out of your plan pretty quickly and it's probably going to encourage you to upgrade. All right. So it's kind of— you got to pick, pick your, pick your path there. All right. Two more for you. This one, another one that almost slipped under my radar, but it's actually really cool. So this was announced previously, but it is now starting to slowly roll out. I am talking about Microsoft Copilot Studio Agentic Workflow Builder. Talk about a mouthful. All right, so anyway, yeah, like these last ones have been hard to like say the names because I'm like getting confused between Personal Computer and Perplexity Computer and Microsoft Copilot Studio Agentic Workflow Builder. All right, here's what it is. Well, I mean, visually it just looks like something like a Zapier or a workflow builder, but obviously built inside of Copilot Studio. So this is the new Copilot-powered authoring experience inside of Microsoft Copilot Studio that lets makers build workflows by describing them in natural language. So that's the cool thing is you can just in natural language type out some type of workflow that you might want in your head, and then it's actually going to build it for you, right? Build all the nodes and the connectors and these kind of conditional statements powered by Copilot. right? And then you can go in there and tweak them as you may. So this essentially translates user intent into structured workflow logic and auto-configured actions, triggers, and other components. So this is rolling to general availability later this month, but it has already started rolling out. So make sure you check your existing Copilot Studio environment if you work inside of there. So this sits alongside the existing Visual Flow Designer and the broader Workflows public preview. So here's why it's useful. Well, it reduces the need for manual setup after you do get the technical configuration piece down the first time. You know, if you do want to build those automation flows across different Microsoft products. So it does kind of compress the gap between, you know, describing what you want, you know, so as an example, when an email from my manager arrives, post the subject to the announcements channel and, you know, draft up a press release and save that press release in OneNote and then go into Teams and, you know, tag my colleagues to review it. Something like that, right? You can just describe that in natural language. It will then build out this workflow for you and then you can go tweak the workflow, change different aspects. So Really cool. So it's kind of like a combination between just, well, like an autonomous agent that can go off and do tasks for you, but with something like a Zapier, right? Like an N8N, like a traditional workflow builder. So kind of something in between. All right. Last but definitely not least, are we going to have less AI text slop? Well, maybe. That's because we did get a new GPT-5.5 Instant model. So what does it mean? Why is it important? Well, this is the new default model. So, you know, ChatGPT, the, the big AI, the original OG chatbot that has nearly a billion weekly active users, right? I don't know the exact percent that use the default model, but, you know, Sam Altman last summer did say It was something like 93%. So the majority of people use just the default model. They're not going to go through the model picker. And that's really important. And that's why this update, even though it technically didn't really grab any headlines, it's actually a big deal. GPT-5.5 came out last month, right? But it's the default version was still running what was called GPT-5.3 Instant. Confusing, I know, because before that we had the GPT-5.4 series. So the Instant series was still on 5.3. So the 5.3 model Instant was not really good. All right. And I said that when it came out and I'm like, yeah, you just don't use it at all. And I still will always tell people and advise people always use the thinking mode, even if you put it on light thinking or standard thinking, depending on your plan. Uh, right. I, I don't use instant models except to test them. But one thing I have noticed about GPT-5.5 Instant, uh, Instant, yes, it did get much better benchmarks compared to 5.3, but the writing is actually really good, right? Which by default, you don't really see a lot. And people will always have their opinions like, oh, Gemini is great by default. You know, Claude is great by default. Not really. Right. They all kind of sniff of, you know, yeah, this is just AI, right? Just random amounts of sycophancy. And, you know, you're absolutely right. Right. You still get those AI-isms a lot. GPT-5.5 Instant. And I still think, you know, with 5.3 Instant, you still got that. You don't really get it. So I don't know, at least if you go look at all the kind of default non-thinking versions of all the different models and I don't know, just test it on writing in general. I think 5.5 Instant is actually the best now. Anyways, here's kind of the details of what's actually new. So the new default ChatGPT model is rolled out and it replaced 5.3 Instant. So it does sit alongside its Thinking and Pro counterparts that were released in April. And then if you're looking for the API, it corresponds to chat latest. So this is already rolled out to free users, Plus users, Pro users, anyone. So it's already live as the new default model. And they do have, which is really cool, there's new memory and personalization features that have been rolled out alongside. So now it can pull context from your past chats, files, and your connected Gmail. So that's pretty cool. And you can go and modify, modify the way it actually goes and retrieves that context as well, which is a new feature that they kind of rolled out with 5.5. Okay, so here's why it's useful. Well, OpenAI says that it produces 52% fewer hallucinated claims than Phi-3 instance. So that's the big one. So they said on that happened on medicine, law, and finance, and then 37% fewer inaccurate claims in conversations users had flagged for factual errors. So it just gets things right a lot more often, which is refreshing, right, without having to think. And that's one of the downsides of using a non-thinking model, right? Sometimes they're just not always that good because they're usually just trying to as quickly as possible and as token efficiently as possible be a helpful assistant for you for whatever you are asking them. So, you know, accuracy for any non-thinking default model is always going to be not as good as the thinking counterpart. So it's good to see some of those benchmarks improve. And it's also smarter, OpenAI says, at deciding when to search past chats, files, or your connected Gmail to personalize an answer instead of asking for the same context twice. That's the thing. In my limited testing of GPT-5.5 instance, like I said, instance, I still usually always use the thinking version. It's like, okay, well, what do you do for that extra 20 seconds? I read the chain of thought and think, right? Anyways, I do see that it pulls a lot more when I want it to pull without even asking it and from other chat you know, history. So some of the times, especially I think Claude is really not that good at this, at knowing or inferring when you do want it to go and look at your Gmail without saying, look at my Gmail, right? Or, you know, we were chatting about this yesterday in a different conversation. You should just know. So sometimes with Claude, you have to be a little more explicit. It does a good job when you do it that way. But the new GPT-5.5 Instant, I found not only does it just have a, I think, a better you know, writing tone that's pleasant to read and is not like the rest of, you know, the default kind of AI slop that's out there. But it does just do a better job of knowing when and how to traverse kind of your files, your past chat history in your Gmail, which for me is refreshing. So who knows, maybe I will, you know, use GPT-5.5 Instant a little bit more, but probably not for things that I would normally go in ChatGPT for. I would probably start using this for things where I would normally use like you know, AI Overviews. So, which isn't a lot because I usually use AI Mode, not AI Overviews. But essentially it's like this might just be my new quick Google replacement, right? Whereas before that was AI Mode. But AI Mode, I use it all day, right? And even on a paid plan, I actually run out of the, the higher version of AI Mode. So maybe this is just going to be my next thing I kind of move inside of ChatGPT. Is, and maybe I'll use AI mode less. I don't know. So that's it. Those are the 7 updates that you might have missed. I hope this episode was helpful. Like I said, these might sound like small things, but, you know, if you're constantly using the default mode in ChatGPT, this is big, right? If you're in Google Docs all the time, this is big. If you're a Microsoft 365 Copilot user and you're on the go, having Copilot co-work work on your mobile, that's big. So yeah, even though these, you know, weren't the headline-grabbing features that you might expect, they're actually AI upgrades definitely worth checking out this week. All right, I hope this was helpful. If so, if you're listening on the podcast, appreciate it as always. You can always see the video versions or just go sign up for our newsletter at youreverydayai.com. So thank you for tuning in. We hope to see you back next week and every day for more Everyday AI. Thanks, y'all. Every AI vendor in your inbox is promising to revolutionize finance. Most of them have never set foot inside of a controller's office. Sage has. At Sage Future, they double down on AI you can trust for finance. AI that gives your team confidence with results they can explain and verify. Control so people stay in charge of the decisions that matter, and accountability so so every action can be traced. Plus, a faster way off legacy desktop tools and into the cloud alongside AWS. That's how you actually adopt AI in mission-critical work. Check out sage.com for more.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.