Transcript of Dia AI browser is seriously impressive. Hands on: how you can use this at work.
Department of Product
0:00So, it's been a few days since the0:01browser company officially launched its0:03first AI browser called DIA. I've had a0:05few days to play around with it and I'm0:06seriously impressed. In this video, I0:09want to explain why and to share some0:10hands-on use cases that you can use at0:13work for things like market research,0:14accessibility checks, and more. And as0:16always, if you enjoy the channel, hit0:18the subscribe and the like button. So,0:20last week, I installed the beta version0:21of the browser company's new product,0:23Deer. And as someone who's never really0:25managed to click with their previous0:27browser, Ark, despite trying very hard,0:29I was a little bit skeptical. But just a0:32few days later, I'm pretty much now sold0:34on this concept of AI browsers and find0:36myself using the AI features far more0:38than I had imagined. Using an AI browser0:41has pretty much transformed how I0:43interact with many of the pages that I0:44visit. And I've realized that when you0:46give your browser contextual awareness,0:48it suddenly opens up a whole new bunch0:50of ways to use a browser that I hadn't0:52really considered before. Now, as well0:54as deer, perplexity is working on its0:56own upcoming AI browser, Comet. Chrome0:58is gradually adding more AI features and1:00Opera has launched an AI browser of its1:02own, too. So, before we take a look at1:04some of the ways you can use these1:05browsers, let's first get a snapshot of1:08some of the core technologies that are1:09used in these browsers and why they have1:11the potential to transform how we use1:13the web. So, here's a snapshot of some1:16of the core technologies that are used1:17in an AI browser. And the first thing1:19you notice is this concept of contextual1:21awareness. Now, contextual awareness is1:24at the heart of these AI browsers. And1:26just a few days ago, Perplexity CEO made1:28reference to this in relation to their1:30upcoming new browser with a simple tweet1:32that read, "Context is all you need." In1:35AI browsers, contextual awareness refers1:37to the information that the browser has1:39access to that will allow users to1:40transform everyday workflows. So, for1:42example, this means that you can ask AI1:44browsers to reference multiple tabs at1:46the same time when performing some1:48analysis or reference a set of brand1:50guidelines when crafting a release note1:52or an email about an upcoming feature.1:54Contextually aware browsers use a1:57combination of multiple sources like1:59open tabs, cookies, APIs, and the DOM or2:02the document object model. So, in a2:03sense, they can see the pages that you2:06view. And with mission, they will use2:07locally stored browsing histories to2:10personalize your experience. On device,2:12machine learning helps the browser learn2:14from your behaviors and fine-tune the2:16user experience by suggesting relevant2:17websites based on your previous2:19behaviors. And the final piece of the2:21puzzle is access to AI models, typically2:23in the side panel, which users can2:25interact with and chat with. And in many2:27cases, these include a combination of2:29proprietary models built upon the likes2:30of OpenAI, Gemini, Meta, and so on. Dear2:33CEO says that its AI models will become2:35more personalized over time and says2:37that with every tab that you open, it2:39should feel like this AI model is2:41getting more and more personalized to2:42you. So that at the end of a week of2:44browsing, a month of browsing, let alone2:46a year, it's going to know you as well2:47as your closest friends or colleagues.2:49Now, that does sound a little bit2:50disturbing. So long as most of the2:52processing takes place in the browser,2:54it's a trade-off that many users may be2:56willing to make. So let's take a look at2:58some use cases that you can put into3:00practice if you're interested in AI3:02browsers. First up is the example of3:04conducting market research. We can use3:06the browser's ability to reference tabs3:08to create a comparison between two3:11different products. So in this example,3:12I've got two windows open. The first is3:14for intercom and the second is for3:17decagon. And we're going to imagine that3:18we are working at Zenesk and we want to3:21conduct a comparison of these two3:22products. So in our AI browser, we can3:25simply reference both of these windows3:27and ask it to create a comparison table.3:30So all I do is I say create a table3:33outlining the core features of these two3:36products. And then since the browser is3:38contextually aware, it will use both of3:40those tabs to create the table that3:42we've asked it to do. So this gives us a3:44nice overview of the different features3:46on offer by each of these two different3:48companies. And if you wanted to, we3:50could then take it a step further by3:52asking you to reference our own product,3:53in this case, Zenesk, to give us an3:56understanding of how each of these3:57products compare and to identify any4:00potential opportunities for feature4:02differentiation. De also comes with this4:04power called skills. So skills4:06essentially allow you to quickly run4:08prompts that you pre-program and then4:10run them whenever you need them. So, in4:11this example, I'm going to create a4:13prompt that asks the browser to search4:15Reddit for threads on a specific topic4:18relevant to my product and to find4:20problems, solutions, and uploaded4:21comments. So, if I go ahead and save4:24that in the browser, with this skill now4:26created, I can invoke it whenever I want4:28to for any industry. So, in this4:30example, you can see how I can use this4:32for industries like pet insurance or for4:34SAS customer service products. And4:36within a few seconds, you'll get an4:38up-to-date view of what users are saying4:40on Reddit about your industry just by4:42using the keyword in your browser. So,4:44that's a really powerful way of using4:46the skills feature. I can also take this4:48a step further as part of my competitor4:50analysis. And I can create a skill that4:52asks dear to find all recent appearances4:55of competitors leadership teams. So,4:58let's say I want to keep on top of what4:59kinds of marketing competitors are doing5:01through podcast appearances and so on. I5:03can create a skill that is called comp5:05pod and I can ask it to find any podcast5:08appearances from my competitors and I5:10can either hardcode my competitors in5:12there or I could leave that flexible so5:14that I can use that as part of my query5:16then tell the skill which competitors to5:18focus on. So that's a really handy way5:20of doing competitor analysis as well.5:22You can also use this for things like5:24accessibility checks. So in this5:25example, I'm going to visit a website5:27that outlines accessibility checks that5:30designers might want to refer to and5:31then ask the browser to cross-ch checkck5:33this against the UI that I upload. So in5:36this case, I'm going to upload a snippet5:37from Apple's new divisive liquid glass5:39on Apple Music and I'm going to see if5:42it passes the criteria outlined. All I5:44need to do is simply open up a new tab5:47and then in that tab ask it to reference5:49the checklist and say does this UI5:52attached meet the accessibility criteria5:54that is outlined in this document. And5:58as you can see unfortunately for Apple6:00that UI has failed the test. But you can6:02imagine that this is super helpful for6:04any instances really where you've got6:06multiple windows open and you want to6:08cross reference information from one tab6:10into another. So, for example, let's say6:12you're in linear or you're in Jira and6:14you want to get a snapshot of some of6:16the user stories that are currently in6:18development and then use that6:19information into an email. You can do6:20just that by referencing the tab and6:22then pasting that information directly6:24into an email. So, it makes things like6:26status updates really easy as well. And6:28the final use case that I wanted to show6:30was the ability to create your own6:32personal industry news or research6:33digest. So in this example, you might6:35want to keep on top of industry news in6:37your category. Or if you're interested6:39in keeping up to date with the latest AI6:41research, you could create a skill that6:43asks deer to find the most recent and6:45interesting AI research that is6:47specifically focused on your area. So6:49this skill asks deer to summarize the6:51main findings, the authors and research6:53groups involved, and the potential6:54impact of each study. But I can also6:56specify how I want the format to look.6:58So I can say I want a list of new6:59studies with a header and two to three7:01bullet points underneath along with the7:03links to the studies. And I can make7:05this customizable by saying allow me to7:07specify the focus area using the keyword7:10query. So let's say for example I want7:12to focus specifically on AI research7:15relating to social media. All I need to7:17do to invoke this is to use the res7:19keyword and then type in social media7:21and the browser will do the rest. The7:23browser will then go and find all of the7:25relevant AI research linked to social7:27media. So I've got examples here from a7:29new paper published by researchers at7:31UCLA which talks about content7:33dissemination and regulation in multi-7:35aent simulations. And because DIA comes7:37with the ability to chat, I can then7:39open up the chat side panel and begin to7:41ask questions or get a summary of this7:43specific research model which is really7:45helpful. So, as you know, I'm not really7:47one for hyperbolic, overdramatic hype7:49for the sake of it, but based on my7:51couple of days experience working with7:52this browser, I'm seriously impressed.7:54And I really do think that this will7:56transform the way we use browsers. And7:58now, I'm really just looking forward to7:59seeing what Plexity and other companies8:01cook up. If you're interested in digging8:03into some other use cases for how to use8:04this AI browser at work, then head over8:06to Substack where I cover more use8:08cases, including things like writing8:10PRDs and understanding your technical8:11documentation and analyzing spreadsheets8:14from multiple data sources. So, thanks8:16very much for watching and I'll see you8:17in the next one.
1,757 words · 250 lines
