The New Search, New LinkedIn Guide, Smart Glasses, and Auracast  ## Metadata - Author: [[Marketing Over Coffee Marketing Podcast]] - Full Title: The New Search, New LinkedIn Guide, Smart Glasses, and Auracast - Category: #podcasts - URL: https://share.snipd.com/episode/9f3ff6c3-52b1-4b4f-abda-99f7853076a6 ## Document Notes ## Highlights - **Craigslist Car Sale Scam** - John Wall shared his frustrating experience selling a car on Craigslist with text messaging turned on. - He encountered scams involving fake checks and shipping expenses that caused trouble. Transcript: John Wall I did do an episode in the podcast where I mentioned my father's passing away. I have an estate sale going on tomorrow and just, man, don't ever sell a car on Craigslist and turn on text messaging. That's all I can say. I didn't realize that so many people wanted to send me the check with the shipping expense in it and to have me take the shipping expense back. ([Time 0:00:23](https://share.snipd.com/snip/0fe46daf-f501-4384-83df-085a3397976d)) - **Google AI Alters Search Landscape** - Google's new AI mode replaces the traditional 10 blue links with AI-driven deep research reports. - This change drastically reduces organic SEO visibility and shifts search into conversational AI interaction. Transcript: John Wall So let's get into interesting stuff. I think the big thing for me that popped from the last couple of weeks was Google announced a lot of stuff. And there was a lot of chatter about it changing what organic search looks like. So you've been following all this stuff. Christopher S. Penn What's going on there? Oh, there's so much to talk about. So Google I.O. Was a week ago now. Seems like, was a week ago? Yeah, a week and a half, yeah. I don't even remember anymore. Yeah, they rolled out a ton of new stuff. The big stuff that's going to affect marketers. Number one is AI overviews are currently, I think they said somewhere between like 30-ish percent of searches or something along those lines. And they expect to increase 10% month over month until they expect half of Google searches to have AI overviews by the end of calendar year 2025. So we'll see if that takes place, but that is a big number. And obviously that has huge implications for search. But they also showed AI mode, which is basically just a Gemini chat. So instead of the 10 blue links, AI mode is purely, hey, talk to Gemini about what it is you want. It will go out and find stuff. It's like a mini live deep research. And it is in production now. And it's crazy in terms of what it means for search because it dramatically changes just how you're going to do everything online. So if you go into AI mode and you type in a search query, it immediately starts going out to the web, trying to find results and dig them. And then it assembles a mini deep research report. And then you have to go and you can have conversations and follow-ups and stuff like that. And what's interesting, of course, you know, there's the right rail that has its references and sources. And there's three listings. I'll tell you all the sources I drew from, but the top three listings. So basically, there are no more 10 blue links. That is the big thing. There's no 10 blue links. And so every SEO practitioner is like, so what do we do? And of course, they're going to be experimenting with putting ads, of course, in the right rail as well to preserve their ad business. They've also said that right now, the Google Search Console has no data, and they expect to have at least a little bit of data, but not a ton. So SEO folks will generally not have insight into what AI mode is doing for your traffic, other than almost everyone in SEO has seen this in the last six months. The blue line of search impressions has gone up. The purple line in search console of clicks has gone down and everyone is across the board. I can't think of a single client. And we've looked at like literally hundreds of Google search console instances and everyone says the same story. John Wall The amount of traffic you're getting from Google is going down. Right. And so, yeah, the first thing I thought is preserving the search business. And so that's just, they're going to throw stuff in the right rail and that's the easy way to get around it. ([Time 0:00:48](https://share.snipd.com/snip/f8aaa7b6-6e63-4677-84d9-55680163131b)) - **Optimize Content for AI Search** - Marketers must adapt content strategies to align with AI-driven search sources to stay visible. - Use resources like the free Trust Insights guide to optimize content for AI search environments. Transcript: Christopher S. Penn Exactly. So it is a dramatic change. It means that marketers need to be very focused right now on adapting for this new reality and doing the things that will hopefully get you placements. And so what that means is, A, you should be using these features and B, you should be testing out the sources that it's getting its data from. What are those sources in general? And then trying to get your content into those sources as well as obviously doing all the right things to make sure your own stuff is in good working order. ([Time 0:04:12](https://share.snipd.com/snip/adf8238d-8a43-4d16-b9db-7d474a204576)) - **AI Video Generation Risks** - Google's VO3 video generation model produces high-quality AI videos with synchronized audio. - This tech raises serious ethical concerns about potential misuse in marketing and misinformation. Transcript: Christopher S. Penn But yeah, that was some of the big stuff. The other things that came out, Google's new VO3 model, the video generation model came out. Eight seconds of video it can generate. But what's different is that it can generate the audio track along with it. So you get extremely high quality video. It's uncanny valley stuff. It's literally, it's like deep fake territory of like, okay, this is not, it's not as easily detected that it's AI, but then you get the voiceover, you get the background music, you get The sound effects, all of it in one download. ([Time 0:04:59](https://share.snipd.com/snip/1a321239-f26e-4cf0-8a2e-5a6e44887af1)) - **LinkedIn Focused Activity Wins** - To succeed on LinkedIn, focus your activity and profile around a specific relevant topic. - Engage consistently with related content and build connections aligned to your niche to improve algorithmic recommendations. Transcript: John Wall You've got a bunch of stuff there that we've been giving that away for free. I'll link to that in the show notes, too. So you can check that out. But you did in the newsletter outline kind of a six step process on what to deal with this. But what's the big story with this? Like what's gone on with LinkedIn that we need to be aware of now? Christopher S. Penn Here's the thing. They've got new AI models in the process. There's a GNN, a neural network in front of it. There's a new AI-based, LLM-based ranking system in there. There's a bunch of different pieces they've built into the system. They're really cool. And it gives you guidance about how to think about your LinkedIn profile and your activities on LinkedIn. Basically, if I had to boil it all down, say and do things on LinkedIn that are relevant to what you want to be found for. So if you are, say, you're all in on Blenders, as a ridiculous example, you should be leaving comments on other people's Blenders posts. You should be commenting about Blenders. You should be putting up your own posts about Blenders. Your LinkedIn profile should have Blenders all over it. Your first degree connections should also be Blender adjacent or Blender specific. But essentially, these language models give LinkedIn a much more capable way to understand what someone is about. And so if you are just out there on LinkedIn, starting fights and all sorts of stuff, and you're probably going to be topically not as relevant on any one specific topic. So you do want to have some focus on that. And you want to be, it rewards activity. I mean, that is not a change, but it really emphasizes that the activity you do on LinkedIn has to be focused if you want to be recommended and found for that ([Time 0:06:40](https://share.snipd.com/snip/f2828e1d-f30e-4a86-a200-99cb380596b5)) - **Safety Limits AI Creativity** - Claude 4, despite hype, underperforms some earlier models on coding and benchmarks due to safety restrictions. - AI model creativity often declines as safety constraints increase, affecting overall capability. Transcript: Christopher S. Penn So yeah, so Anthropoc released Claude 4, both the Opus and Sonnet models. Very interestingly, on a lot of the leaderboards, like in LM Arena and over on AdrChat, the developer coding tools, Claude 4 does not outperform Claude 3.7 in a fair number of cases. And they both do not good stuff on that? OpenAI's O3 or Gemini's 2.5. So there's a lot of questions about what Cloud4 is good for, because it's still, as with all of the Cloud models, it's still very expensive to use, particularly on the API. And it's not clear yet. The thing that it seems to be optimized for is tool handling, using MCPs, using external tools and stuff. It seems to do really well at that, but at the expense of not being able to do as much stuff at other stuff. And there's been a lot of interesting articles talking about how early versions failed its safety tests, like trying to blackmail its engineers and stuff like that. OpenAI has been experiencing similar issues where it was trying to override its shutdown codes and things. And one of the things that people don't know about language models is this. The more safe you try to make them, the dumber they get, because you basically have to lobotomize a lot of its functionality. So a completely uncensored, unsafe model will be extremely capable. It's why some of the models that are out there, that are open source models that are used to write adult material that the mainstream models will not do, they're very creative. They're better writers because they have fewer restrictions on how statistics about words work. And so these new advanced models, if they've been very heavily redacted for safety purposes, may be underperforming for that reason. That is speculation. I do not know for sure, but that's speculation. ([Time 0:08:43](https://share.snipd.com/snip/4d2c1ccc-74ef-4816-a02b-77ca04d4dee2)) - **DeepSeek AI Model Surprise** - DeepSeek quietly launched a new free AI model that early tests show outperforms many top competitors. - This highlights how fast AI capabilities evolve and the competitive race among AI providers. Transcript: Christopher S. Penn But DeepSeek, the Chinese company, very quietly pushed a new release. They said, oh, yeah, we updated our model. That was literally a blog post. That was all they did. And it's spectacular. So far, early tests are showing that it outperforms almost everything in the field. It is still free to download. It is still dirt cheap to use on the API. Obviously, if you're using their services, it is still not private in any way, shape, or form. But the early tests, because it dropped literally 10 hours ago as of the time we're recording the show. So it's still new in the field. But early tests are showing that it is a dramatic improvement over its predecessors and a dramatic improvement against all the other competitors in the field and still free for you To download. ([Time 0:10:48](https://share.snipd.com/snip/7029ace6-78c3-423d-8e88-436a8ce74eae)) - **Duolingo CEO AI Blunder** - Duolingo's CEO publicly boasted firing most employees and replacing them with AI, which backfired badly. - The episode showed the dangers of CEO overconfidence and the immature state of agentic AI for complex tasks. Transcript: John Wall We have the Duolingo case. We were kicking around over on Analytics for Marketer. Katie has a nice piece talking about kind of what went wrong with that. Yeah, I mean, the short version for people that have not followed this, the CEO came out with a statement of, you know, kind of the classic, we're going to fire most of the employees and AI is going to do a better job running everything. And it just kind of went awry. You know, I don't know any lessons beyond the normal, you know, don't be a CEO talking about trashing the whole company. Christopher S. Penn Don't do that. I mean, I've seen so many people do this so badly. The thing is this, AI agents and agentic AI is still in its early days, and it is still very difficult to get them to do a wide ranging task. They can do now tasks extremely well. And agentic tools can can do them in sequences and workflows and things we've talked about. Tools like NAN and obviously the coding environments like Klein and Cursor can all do that. But the idea that there's just, you know, a Terminator sitting at the desk instead of a human, that is still far away because the tools still require a lot of babysitting and they get smarter Very, very quickly. They can do things today they could not do even six months ago, but you're still not firing everybody. I do see, and there was a piece from Dario Amode from Anthropic, the CEO of Anthropic, talking about the impacts of AI. He's saying 20% unemployment rates are not unrealistic because these things are just going to devour particularly entry-level work. And I definitely see that. But the whole, the way Duolingo approached it was just boneheaded. John Wall What did you think of it? Well, it's asking for trouble. Yeah. You know, and there's, I've seen a lot of this stuff too, talking about, you know, people make the mistake that transparency and authenticity are the most important things. And no, it's just not like, there's a lot of things that don't concern. The short version of this is there's no person looking into learning a language today online that cares at all about where the LLMs are and who's doing the graphics and what the workforce Is completely, you know, none of those things are involved in the customer experience. So, you know, and then I could go on a whole other set of rants about how, you know, that's the CEO talking to the street, right? Like that's, he's looking to move the quarterly stock numbers with that kind of talk that has nothing to do with running the business. And that can easily bite you. Yeah. Because like you can see, yeah, they did a ton of things wrong. But yeah, I don't know. I feel like I'm all in on the fact that, yeah, I can see AI being an excellent language instructor. You know, like we don't need tons of people to be doing Spanish 101. Like that's a pretty well-established list of stuff that matches that situation we've talked about before where AI and computers for healthcare for seniors. You know, Amazon Alexa will answer a senior citizen 500 times in a day and not get angry about it. And that's an important, there's value in that. Because, you know, no human really wants that job. And so, yeah, it can be replaced. So, yeah, you know, they definitely made some mistakes. There's some truth to a lot of it, but that doesn't mean that you need to brag about that on the internet. ([Time 0:12:06](https://share.snipd.com/snip/2020a02e-f64d-4b95-a0c6-58b38c2eba60)) - **Auracast's Early Hardware Struggles** - Auracast enables audio streaming to multiple compatible devices but lacks broad Apple support. - JBL and Sennheiser lead in hardware support, showing early ecosystem growth challenges. Transcript: John Wall The mark. All right, a couple other things. AuraCast, I wanted to report back. I've still been playing around with that some more. I have an AuraCast transmitter. Digging in, you know, the real challenge with this is Apple does not have an app available that supports it. So there's no, you know, inroad there. As far as hardware, JBL has really stepped up. There's a lot of JBL products that support this. And then I also saw some Sennheiser earbuds that unfortunately appealed to my audio gear addiction. And I may be picking up a set of those sometime soon so I can get into that. But yeah, the idea that you get this device, you're able to stream some audio, multiple devices can listen to it. It seems to just have a lot of interesting applications. So I'm looking forward to play with that. ([Time 0:16:20](https://share.snipd.com/snip/6dedbaa6-6fee-4b65-b8be-d3ba2cd860c1)) - **Smart Glasses Not Ready Yet** - Current smart glasses remain underwhelming with mostly Kickstarter-level features. - Use cases like virtual screens for video editing could be valuable, but the tech isn't ready or affordable yet. Transcript: John Wall Also on the gear watch I wanted to throw out, I have to get some new glasses. So I was looking at smart glasses because it's so funny over the past couple of years, I'm like, well, I'm not buying these smart glasses because if they're not prescription, I'm not Really going to wear them that much. Well, now that I need a new prescription, it's like, oh, well, maybe I should just get smart glasses. But I'm throwing this out to the audience and everybody else because everything I've seen so far, I'm like, yeah, no, I'm not doing that. It's not worth the benefit. There's been some interesting stuff I've seen kicked around, but a lot of them are still at the Kickstarter level of virtual screens on the glasses. So if you want a teleprompter or if you want something that's going to pop up, like when you're at a meeting, Hey, this is this person, here's, you know, their full name and all this stuff, Which I forget continually. Um, that could be really interesting for me, but, um, yeah, I don't know. Have you thought of it looked or thought about smart glasses, anything on that front that's appealed to you? Christopher S. Penn I have thought about the, the, the Vitor ones. I've seen them on Instagram and stuff like that. Of course, Apple Vision. The main use case for that would be things like on the plane where I want my 65 inch screen. But obviously I'm like a Tyrannosaurus, like stuck like this. I don't need the virtual reality overlay, you know, Terminator style in real life. I have no need for that. But having that whole blackout, like you put the goggles on and now I can work in Premiere, for example, on the plane, that would actually be legitimately useful because I am on the road So much. Yeah, that's a great point. John Wall The idea of 120-inch screen or, like you said, Premiere, and having like a full screen size display of the video and still having all your editing tools over here. Yeah. Yeah. That's a win. I can see that. I don't see that being cheap anytime soon, but we, you know, it will get here. ([Time 0:17:03](https://share.snipd.com/snip/6e2f0bc0-3427-44b3-8e4a-36b68345f04c)) - **Use AI for Reporting, Not Math** - Avoid letting generative AI handle math; do preprocessing externally and let AI generate reports. - This method ensures accuracy while leveraging AI's strengths in communication and synthesis. Transcript: Christopher S. Penn But the Phoenix thing, I'm doing a half-day workshop on using generative AI for doing analysis, which is going to be really interesting because it's the Marketing Analytics Summit. And the one thing that generative AI is really bad at is math. And therefore, this is an interesting challenge to teach. How do you use this stuff and get correct answers? And the short version is you don't let generative AI do math. You do all the pre-processing somewhere else, hand generative AI the conclusions. And then it does, you know, making the reports and all the fancy stuff. But you definitely, definitely don't want it doing math. ([Time 0:19:03](https://share.snipd.com/snip/09a1148d-2a10-4f52-b801-0e38bb938999))