Why the Death of UI Could Change Everything for Businesses

Mark Smith 0:00

The intelligence age is not coming. We're already living in it. I'm Mark Smith, and each week I speak with the people building, leading, and navigating a world being reshaped by artificial intelligence. We explore AI technology, the changing nature of work, and what it takes to build a meaningful career in the intelligence age. This is the intelligence

Introduction and Personal Updates

Mark Smith 0:19

age. Ashish, welcome back to another podcast session. It's been a while since we've had a chat, mate. Well, it's well over a year, isn't it?

Ashish Bhatia 0:26

Absolutely. I think it was uh last year, sometime around spring that we spoke, I think.

Mark Smith 0:32

Yeah, yeah. And uh your world has changed massively. Um last time I saw you in person was on the Redmond campus at Microsoft, and um your world changed a lot. I hadn't written a book, I hadn't been published on Audible. All things I've done now since I last saw you, and of course you're working at Audible.

Ashish Bhatia 0:53

Yeah, I am. Yeah, and the last time we met, it was I think the MVP summit.

Mark Smith 0:57

Mm-hmm.

Ashish Bhatia 0:58

Uh around March, April timeframe. Um and then last year, around August, again this month, yeah, coming into my anniversary, I joined Audible. So which is super exciting.

Mark Smith 1:12

I am keen to hear what what's the last six months been like for you from an AI perspective? What's changed for you? What are you focusing on um now and and what impact are you having in the AI space?

Transitioning from B2B to B2C in AI

Ashish Bhatia 1:27

So I've been doing a bunch of things, right? Again, transitions are hard, but transitions also create a reset. Um, at least in my world, I was a Microsoft person for a long, long time. I spent more than a decade there. And most of that time was with the enterprises, right? Um coming into Amazon and Audible, I'm I was moving from B2B to B2C, and that was a profound change. And I always say, right, uh when you are in a B2B situation, uh, you're a lot closer to your customers, like yourself, right? We could meet at MVP Summit or Power Platform Conference or something else, and I could have a chat with you, what you're liking, what you're not liking, give me your feedback. We have all sorts of programs, uh, where we're doing previews with customers and so on and so forth. So you're very intimately close to customers, and they'll almost always tell you what's wrong, what's right with your services because they're paying customers. In a B2C space, you don't see customers that often in that intimate setting, uh, and you see your customers mostly through reviews, comments, data, right? Uh, and so on and so forth. So you look at them from a data lens. Uh we do a lot of UXRs, we talk to customers all the time, but you still look at them from a very data lens, which is different from B2B uh kind of side of the business. Audible is interesting because you will be I mean, uh you'll not be surprised. You you you you contribute to it, and thank you for contributing content to Audible. Um anytime I go to a meetup, I bump into people at uh these conferences. I run into customers all the time, right? When I tell them I'm from Audible, they would like, hey, we are customer. This is something that I like about Audible or not. Like I've m met people like my student age or very early in my kind of learning path. I was introduced by my pa parent or somebody to Audible, and I've continued that journey, so on so forth. So it's good to kind of hear some of that. Um,

The Rise of Personal AI and OpenClaw

Ashish Bhatia 3:41

but again, it's a very different lens how you interact with customers. On a personal front, right? I think we were chatting pre-game. Something changed earlier this year, right? And I feel that has created profound opportunities for everybody to kind of tinker with AI a lot more. Uh coding models became a lot more smarter. Uh harnesses came along. I've been tinkering a lot with the harnesses. In fact, right before this call, I was setting up my new Mac Mini with uh an OpenClaw instance. I've been running an OpenClaw instance for the last six months. Uh, and I was kind of transitioning from the the older substrate, which was a Windows kind of machine, to a Mac Mini to again just get a little bit more robust substrate uh that I'm running my OpenClaw on. I've been running uh the OpenClaw Boston kind of community. We meet up every month. Uh and every other month we try to do a hackathon. It's been slow in the summer, but that's something that I'm running with a few from a few more people who are interested in kind of personal AI. I feel personal AI is super interesting and important. Um it is by far one of the most intimate technology that we will ever kind of play around with. Uh again, forget about social network. Sure, we shared our location, our contacts, our images and whatnot with it. But again, imagine the kind of conversations we are having with Claude and ChatGPT and whatnot, right? We we're sharing health data, we're sharing financial data. Again, keeping sanity of the data within your own kind of bounds will become super important. That's why kind of personal agents are super important for me.

Mark Smith 5:24

I love this. Like I I feel that everybody needs to be learning about harnesses now and and and why harnesses are so, so important. I didn't realize when you were coming on the podcast that you were this deep in open claw. Um and the reason is I say this is that I lasted three days on open claw running on Windows. And uh that was it. Like I so I do everything in the you know, I did it in Azure VM, destroyed that and went to Ubuntu, and I've been on Ubuntu now for um from as you say, substrate level four. Well, ever since, so ever since February. And what blows me away is how big my system is now. And like I've got 26 agents. Um they I've built so much software. Um, you know, I've got uh one agent team, which is a DevOps team, which is made up of nine agents, each have their segregation of duties or the very specific deep function, they're skilled, each agent is skilled according to their function. Everything is gated or pipeline controlled through DevOps. Um and you know, I've built in my Ralph loop, so it's always getting better and better and better. And I am able to, you know, at the moment it does automatic bug fixing, you know, feature releases that, but it's now I'm getting to building much bigger things as I've gone on that learning journey. Everything from, you know, building my own um routing engine, um, you know, nine different uh AIs plugged in. I I give first class access to Chinese models just as much as US and European as well. Um I have built so much, like as I know I'm always trying and I've learned just an unbelievable amount. And so I think OpenClaw is absolutely phenomenal. Um I'm a bit disappointed that there's this legacy of it is insecure. Like honestly, um, my system, I have full pen tests done on it at a regular every 30 days against latest attack factors and stuff. This is stuff I've architected into my solution, and um you're right, it's phenomenal. But I don't see, you know, I saw Microsoft really scout and I went to deploy it and give it a fair shake, but straight away it required an enterprise GitHub subscription, and I've got premium GitHub, I've got business GitHub. I'm like, I'm not buying another third tenant setup because that's just now making me unproductive, m you know, moving across these uh and like how's and then the other thing is like in the business sense, I can't see a use case for many people needing scout. Like I can't see a big volume game in in that because you need to be technically inclined. You need it's not like I don't see it for your mum and pop to you know go and set up their AI for them. You've got to be deep in AI, I feel, to use tools like this. You're really wanting to move the needle in your own business or professional life and personal life to be going deep into these type of AI tools.

Ashish Bhatia 8:40

Yeah, you need to have tenacity to kind of run this because this stuff breaks a lot, right? And I feel that's the fun of it because that kind of teaches you, I feel systems thinking. I do with this talk about uh what lessons that I've learned from running my open cloud for the last six months. It's kind of something I do locally. And one of the things I talk about is when you're running a scenario uh uh on a very regular basis, and my the thing common thing that I do and kind of talk about all the time is my open claw orders lunch for me every single day using Grubba. Awesome. Right. So what happens is for me that is a production workload because if my workload fails, I don't get lunch and I'm not a happy person then, right? So it's it's mission critical for me. Uh and what it teaches you uh is having run that thing for six months, let's say, it'll show you patterns of failure.

Building Robust AI Systems

Ashish Bhatia 9:36

Right. Those think about those as failure modes, right? That can be auth failure, that can be, I don't know, running out of token failure or whatnot, right? And you start to see these patterns and you can then start to think about how do I eliminate this pattern, how do I overcome this hurdle. And over time you start to build a more robust system, right? And that is something that again, if you're a tinker, if you're interested in kind of owning your AI, owning your memory most importantly, because again, sure, you can offload your AI to uh Cloud or OpenAI or any other proprietary uh kind of system, but they'll take your memory as well, right? And your memory is what you want to secure as much as possible. Yes. Uh it's hard to get it out of these systems and kind of own it and take it elsewhere, right? So you can't take your business elsewhere because now you have built this gravity into the system, which is not easily transposable. And I think uh this that is one thing that kind of open claw allows you to do, that you can take the system anywhere you want. I, as part of this migration to Mac Mini, I tried Hermes Agent for a few days. It didn't work out for me, but again, I was able to export all of the memory from OpenClaw, was able to inject it right back into Hermes and it knew everything about me. What OpenClaw knew, right? And now I can take that again back and plug it into the new claw that I'm setting up. So

The Importance of Data Ownership

Ashish Bhatia 10:59

I own that thing and I can take it anywhere. And I think that is if you kind of extrapolate that for five years, ten years, like that is an asset you don't want to hand over to one of these companies.

Mark Smith 11:12

No, and and the thing is that I feel what open claw gives me is ability to learn practical skills like you get from nothing, like you don't get the practical level from just, you know, spinning up claw desktop or um you know, or codex or something like that. You get it because you have to solve problems and throw away stuff that you've worked, but you've gone down a path, but it's architecturally wrong, so you reinvent, but you're constantly learning, and of course, you behind the scenes you're building this big data set of of your learnings and practical applications, and it's made me, I don't know, I feel like I'm back at the start of my IT career where everything's exciting and new, and the ability to create is just unbelievable. And like for me, I don't just stay in one domain, right? I'm cross-domain. I've got an agent that just monitors my healthcare and it's taking everything from my watch every 30 um minutes. It does a sync with my watch to my servers, it takes everything from my digital scales to the supplements I'm taking to um uh uh my sleep patterns, my heart rate, all that.

Enhancing Life with Personal AI

Mark Smith 12:31

And it's ordering blood work for me to be done, and I go to the lab, have the blood work, I pay for it because it's not type of standard blood work, because I'm wanting to um proactively manage my health, not reactively, like the medical system generally waits till something's gone wrong. And I've got some phenomenal insights into my health that I've been able to, and particularly my heart, and that's the only thing that's kind of our I'm very low risk profile, but I've I would have never found that out until perhaps there was an incident in three, four, five years' time. Now I'm taking steps to never ever having that type of incidence in a scenario. I've got a I've got an agent called um Dobby, and Dobby is my house elf, and it runs everything from my cat food feeder, my cat door, my dog food feeder, it runs um my lights, home assistant is fully, I I tore it down and rebuilt it AI first, my whole home assistant system. So now it's going, hey, listen, that that access point, it's out of warranty, it's it's creating this failure rate. Here of research, the absolute best replacement for it. Here's the link. Do you want me to order it? And I'm just like, man, it's just it's making life easier.

Ashish Bhatia 13:44

It's phenomenal, right? It's helping you manage your life.

Mark Smith 13:48

Right.

Ashish Bhatia 13:48

And that is the point of personal AI. Like, if you look at the landscape, right, every company has an incentive. Um, and they will touch certain parts of the spectrum that you described, right? I mean, let's say some are coding oriented, Meta might do personal AI, Apple might do some kind of version of personal AI, Microsoft most likely will stay enterprisey. Everyone will do some narrow kind of flavor, some narrow kind of niche uh in your in your life. Um, but then by building all of this yourself, you can consolidate, you can be in control of how you want to manage all of this, right? And again, it's not for everyone. Uh at some point of time, hopefully it'll it'll become easier to plug things into. Uh in fact, if I if you look at your phone today, right, it's a clutter mess of apps, right? But again, people who have built out these agents, they can just talk to their agent and it gets everything kind of done for them. And that that's a great aggregator in my mind, right? I mean, like the app world of this mobile age could never have been. I mean, phone was the aggregator, but you're still kind of doing a lot of heavy lifting, logging into these apps and playing around messing with their UI. One of the things that I kind of, again, with that uh lunch example and talk about a lot is uh I feel we are headed towards death of UI, where UI becomes less important. Uh I have not gone to Grubhub uh for the last six months because my agent orders it. I don't see their ads, I don't see uh their recommendations that they kind of throw at me. My agent does all of that work, right? So any of the investment that, for example, a Grubhub-like company is doing in reinventing their UX, right? And their ad platform, none of that matters to me because again, the agent is bypassing all of that. Uh and if you're not investing in making your site accessible to the agents, then it's like you're not in the yellow pages anymore. And how do people find your business, right? So I'll take my business elsewhere because that is something I struggle with Grabhub a lot. It just boots me out and I have to kind of re-auth over and over again. That's one of my key failure mode. I've been trying to kind of circumvent that and figure out ways to get around it.

Mark Smith 16:14

I think it's um Google's come out with Web MCP and like every per every company's website, if you're not aware of Web MCP, you need to really get aware quickly and make sure your your site is running on it. Um just touching on Audible for a second, um, you know, I've got a library of over 300 Audible books. So I've been an Audible subscriber from day dot. And for many years I was on the um the monthly, you get three credits a month, and so I bought a lot of books over my well, it's I don't know how long that's been running for. So I've got a lot, and because I'm dyslexic, it's kind of like it gives me unbelievable um insight into information, like people's writings and stuff, because my my typical model is I always buy two copies of every book I acquire. Always get the auto audio version, I always get the the Kindle version. And that allows me now to um, I find I can say highly concentrated when I'm r reading at two times the audio speed. So I set Audible 2x. And because I'm reading, it creates a sm, I don't know, it's like a laser type focus where if I just do one, if I just read with my eyes, I'll get to the bottom of the page and I'll be like, what was that about? Because my mind's gone off like five paragraphs higher, you know, what something triggered a divergent. I'm still reading, you know, consciously, subconsciously, and then I get the bottom and I'm like, I have no idea what I just read. Same with audio. I can put audio in and you know, I wake up and I'm like, oh shoot, I've missed whatever, you know, whatever that was. But if I find, if I combine the two, I can have this massive retention and I really, really say laser focus. And so for somebody with dyslexia, I feel it's like a power tool. What are you seeing from your work in there around,

Audible's Impact on Learning

Mark Smith 18:16

you know, people like me with with with um perhaps they're not eye-based learners, or you know, not visual-based learners is a correct term for it, and more auditory-based. Are you seeing a lot of people using Audible because of that? Or is it and you know, particularly around developing skills? Like I'm not a novel um reader as such. I do one novel a year. Dan Brown releases a new book, I buy that one. Um, and so if he goes two years without a book, um I I have a gap. But otherwise, it's always either tech, as I when I say tech books, like explainer. I'm not talking about like um how does Python work or something like that, but I'm talking about technology, AI, singularity, the future, anything that's kind of mind expanding the possibilities or you know, systems thinking, things like that. Um, business, I'm into it. What are you what are you seeing from the inside?

Ashish Bhatia 19:12

Yeah, um Article was built out for commuters, right? If you remember the original story, it was built out for books and tape. And the idea was when you're driving, your eyes are busy, but your mind is free. Yes. And it was built for that scenario. So again, people I'm big on commute, and like I have an hour commute each way, going to work and coming back, and it consumes my time. Like I read I'm able to go through a lot more books. I'm a slow reader. Uh, and the scenario that you mentioned is exactly the one that works for me as well. I'm on the other side, I'm probably reading physical book and listening, and again, it's a great correlation. Uh they both kind of means of kind of intake help me kind of retain stuff much better. Um on the inside, again, we see a variety of uh kind of customers, users, uh in genre, romance is super big, sci-fi is super big. Um but uh I was at an archery tournament yesterday um uh at the state championship, and one of the archers, kind of young girls, she was talking to some of her friends, and she was like, Hey, I get super bored when I'm kind of uh reading a book, but I listen to audiobooks and that helps me kind of learn faster. So again, a lot of people use it for fast learning. Uh again, I'm one of that user because I have a ton of books. I buy more more books than I read, and uh Audible helps me kind of uh go through them so much faster. Um so yeah, that's mm pretty much it, I would say.

The Future of Physical Books and AI Training

Mark Smith 20:55

Anthropic with your court case has been in the news about vacuuming up first editions or one-off editions, etc., cutting the spines off them, scanning them at high rates, etc., to train their models, which is, you know, I understand how smart that would be. The flip side is our physical artifacts and things that are like are are scarce, not in print anymore, that type of thing. And they um, under their fair use policy was like, well, we destroyed the physical thing. We only have the digital, so that's fair use, is pretty crazy because all of a sudden I'm going, you know what? I wonder if I should buy more physical books because they might not be, you know, the nostalgia of it, etc. They might not be there in the future if they've all been gobbled up and digitized and only available by the gatekeeper of the digitization of it. Um

Cultural Perspectives on Books and Knowledge

Mark Smith 21:46

I'm sure you're across it. What what are your thoughts?

Ashish Bhatia 21:49

Yeah, several thoughts. Like I was I think just yesterday, as listening to a podcast about this incident that you mentioned, right? grew up in India. Uh we grew up with this idea of books are sacred. Like you if a book kind of you you meet Indians and the book falls off their hand on the ground, they'll take the book and they'll touch their head with it. Yeah. Because it's sacred. It's kind of wow, that's a it teaches you. Yeah, it teaches you something. It kind of has books have profound uh impact on life, right? I mean all sorts of books. Um so when when I heard about this incident yesterday as well, I was like the stark kind of contrast in my mind was the hypocrisy of um the entire distillation story. Right? That here you are scanning these books, training on them. They just had a lawsuit, they're paying over a billion dollars for um pirated books and learning from them. So again the hypocrisy of it is uh you're distilling on human intelligence when then China does the same thing to you, you cry foul.

Hypocrisy in Intellectual Property

Ashish Bhatia 23:04

So that was the kind of star contrast that at least came to my mind.

Mark Smith 23:08

A hundred percent that's not to me at all.

Ashish Bhatia 23:11

And again I'm I'm working on uh again do you know that I I write as well right and I've been working on some some of ideas and pieces. One of them is about China and how we see China and just the narrative in the Western world as they're kind of building all these kind of frontier models and stealing our IP and distilling from us which again it's not very different from what how we kind of started, right?

Mark Smith 23:39

I mean we started the whole ingestion of the internet.

Ashish Bhatia 23:44

Yeah. As a whole internet was ingested and nobody gave a permission right about it. So it's uh hypocritical in that sense.

Mark Smith 23:52

Yeah d defin definitely when I think about and like I'm like I always try to go hey if I was in that position how would I reinvent something like an Audible? How would I make the audible player way different? And for me it's interactivity. I'm going along and I just want to say hey capture that capture paragraph there because I want to explore that when I'm back at a computer interface like a a screen or something or I want to you know I'll hear a word um uh

Innovating Audiobook Experiences

Mark Smith 24:23

right now in the last 24 hours I've been building a solution that I got an acre and a half property um was a beer paddock when I started I'm building a food forest on it um yesterday I sent up my drone and took uh about 750 photos of my property in a fully grid pattern etc and I've been putting it through all the software that now creates me a full 3D model of my property allows me to go down to about centimeter accuracy on anything like the the file is five gigs that it created right so it's not tiny. And there's words and phrases in there I've never heard of before and so I'm finding I'm like what does that word mean? Explain it to me. Like I'm not interested in just seeing something and not understanding it. And so I'm wondering is there going to be and I've seen a few podcast players come out with this ability to go hey um grab that sound bite because I need to come back to it like and is there any thoughts around you know what the what an AI audible looks like um and how that could make you know bring content um more into a more usable format outside of just the uh you know their passive listening experience?

Ashish Bhatia 25:41

Yeah I mean right now a lot of the focus one of the things that we're driving towards is um every book um available as an audiobook in every language that's kind of the mission statement right um like we at the tech is there where it is possible to take every single book and if it is digitized again with the right permissions uh and author and publisher kind of buy-in uh be able to digitize it and make it available in every possible language. We do a lot of original content people absolutely love that uh because it's the multicast and it's performance art and whatnot, right? So those are super popular and that's kind of uh the thing that kind of runs the business the most but then being able to just make all books available in all regions and all possible languages is kind

The Future of Audiobooks and AI

Ashish Bhatia 26:35

of the mission. We infuse a lot of AI in terms of uh the interactivity part of what you were hinting in the sense a lot of AI is getting infused in discovery and uh personalization recommendation so on so forth but also then be able to kind of recall things like hey what was this thing right who's this character what was their motive things like that right yeah um so that's definitely where we are headed as well make that listening experience our learning experience interactive um and but the things that you were talking about like taking digital notes of things right uh is super interesting and then just like you would highlight something in a physical book or even an ebook like be able to take digital notes and then go deep into those concepts especially when you're doing technical books and stuff like that. What is this term? Who else talked about this? Like go research about this and stuff like that. That can make learning so much more interesting and uh rewarding I feel.

Mark Smith 27:39

Yeah you know for me it's like I it's a spark of so many things that I build will come from a concept I hear and for whatever reason that one concept just it light bulbs for me and I'm like I could build that into something I could I could make what was talked about tangible and you know turn it into a product or a piece of software or something. You know I've you take a book like um there was a book back in the early web design days called Don't Make Me Think um by Steven Krug and you know I now go where you know I've got multiple websites and that are all AI built and and running and now I in my interactivity with them I'll just say hey show me where you've applied Steven Krug's entire thesis on this on the website and where the gaps are boom comes back gap air gap air gap air gap okay based on August 2026 what's the best in class research on how this should perform now like it's just phenomenal the the power you get with it.

Ashish Bhatia 28:43

Knowledge is available at fingertips I feel um but again it's like w what are the topics what are the areas that you want to deep dive right and then what path you want to take to go there is again totally in your control. It's again your motivation or your drive and uh how do you want to learn I think that that is interesting. One of the things

Learning from External Experiences

Ashish Bhatia 29:07

that um you again Amazon has a ton of principles Audible has principles. One of the very interesting principles that I like and appreciate a lot is uh about having these external experiences to your work. Like go engage in philanthropy or uh community thing or be part of kind of some kind of community play a sport do whatever right and draw inspiration from that and bring it to kind of work if you can. And I feel some of the stuff that I kind of do that that is super interesting. Even like the open cloud community every time we host a meetup and people come on stage and talk about their setups or their scenarios and like I'm a sucker for scenarios. What are the scenarios you are running because that gives me ideas of how to improve my own kind of setup or new ideas that I can kind of bake into my um my setup and so on so forth. So learning from others is like super interesting. And there's just so much more to learn. I think we're getting to a place where tools can build anything you want right it is the best that you take things that you want to invest your time and go build and it for me it's becoming a more human thing in that sense like what are you building? What are you trying to automate? Like how are you kind of getting your uh kind of health data into one system what does that system look like that is super interesting because that's how we all kind of learn from each other and kind of build things for ourselves.

Mark Smith 30:33

Yeah yeah totally totally agree. What so what is the next projects personal projects that you're working on um in AI? What do you want to build out? What are you hoping to do? I you you've talked about your um your Mac mini build out

Ashish Bhatia 30:44

Personal Projects in AI and Automation

Ashish Bhatia 30:48

yeah so one of the other things that I'm uh currently running and improving on uh I mentioned I I try to learn I have an insatiable desire to learn and um just kind of grow personally um I built out a much like everybody else like everybody probably has already this thing like it's a news briefing thing right for yourself like every day hey early in the morning and then before I go to bed give me my news of the day um and then my bot was giving me all of this information in its very first version in Telegram and that was fine but it's giving me all these links I have to go to these links then start sending me emails so I'm iterating through kind of the V1s and v2s of it right so then it was email. Email is fine but again a lot of that cognitive load is for me to read and can absorb all of this mentioned slow reader I don't get kind of screen time to kind of absorb all of that information. In fact even in that version a bunch of people uh who were kind of in my circle they got super interested can you send me that news briefing I would like to have it because it also kind of reads it here's the most important reading and whatnot right so there's some kind of triaging there as well. And then like I need to be able to consume it in an audio format right again coming from Audible why not so I started converting those into uh notebook LM uh notebooks right and that uh really really changed the game for me because now as I go into my morning commute I can hit a notebook LM kind of uh button in the morning and it plays the entire thing as a podcast for me the entire news briefing yes and same uh in the evening sharing is a challenge because again Google has not made it easy to kind of share these kind of notebooks I was sharing with again people in my circle and I like everyday approve people's request. I'm like can I convert this into an Apple podcast right so again it is like that's the systems thinking part like you build the system you use it you're invested in it right um and then you're constantly thinking well how do I make this thing better right and this is how we kind of build products right all along as well. So now I'm I've built out an entire pipeline which goes from curation of uh news articles morning evening uh building out the entire notebook LM kind of pipeline getting the podcast out of notebook LM downloading it as an MP3 and then creating a feed for Apple Podcasts so now every morning when I go into my car I have an Apple podcast on just my personal podcast ready for me to play and then same in the evening. So I'm adding a little bit more things to it like have uh a research briefing like hey what is the top kind of trending let's say research paper of yesterday like tell me more about it and so on so forth. So that's the kind of project that I'm working right now. That's cool.

Mark Smith 33:51

I hadn't thought of taking it into you know something like Spotify and or um Apple podcasts. I've done a similar type thing mine all sits so in OpenClaw I get a um through telegram every morning the full audio can be up to half hour of content. It's all hey hey Mark here is the researched as of the like I'm very big on research windows that AI you can train them into uh the current context historic is irrelevant in the speed that we're going at um and in the things that we're doing and and then it gives me the text all the hyperlinks that are mentioned so I then can just click a hyperlink if I want to drill into something detailed and you know 6 a.m in the morning that's always sitting there that audio is already loaded ready to go for me but I hadn't I don't do any commuting at all but I tell you what as in when I'm out working I always have podcasts on in my you know in my earbuds and stuff and so now you've got me thinking hey I I should pipe that to a personal podcast feed and it's funny because when I was working with AI on it it wanted to put it through my BuzzSprout account and I'm like no no this is only for me personally I don't want it out in the public domain. But it's um you've got me thinking now you got me thinking.

Ashish Bhatia 35:09

It's easy easy for sharing as well now because everybody can subscribe to that RSS feed and they have uh the feed what uh service are you using for your kind of audio if you mind me asking yeah I've been so it's just doing it on my Ubuntu server directly building it you know and uh and and compressing it down. Yeah I was thinking of eleven labs as well but I what I have what I enjoy about LM notebook is it it has this tandem kind of audio which makes it a little bit more interesting. It's not monotone in that sense. And LM notebook will go get a little bit more extra information about all of that. One of the other things that I do with uh all this news curation is again you've gone through the podcast sometimes you just forget all of that stuff right after a few days. Remember Karpathi recently or actually it's been a while now did this uh LM Wiki kind of project to try to replicate that that uh when my claw kind of pulls on all of these news articles and curates the kind of audio pipeline it also takes them and drops them in Obsidian and based on that obsidian there is a kind of another agent which runs through it, uh processes all those files and kind of builds out this knowledge graph and I have probably a thousand articles living in that knowledge graph. So anytime I need to go back and say hey tell me all the kind of let's say model leaks uh or releases let's say which happened in the last uh four months right so it'll crawl through all of that and pull out uh the insights rather than having to go to on the web. I mean sure it can go on the web and pull out the all of that information but sometimes in those articles again they're interesting things that you have kind of bookmarked for yourself and whatnot so you can always go look back. It's like your knowledge base locally in your system that you can tap in.

Reflections on Microsoft and Customer Connections

Mark Smith 37:02

I like it. Ashish thank you so much for coming on the podcast I'll get you to send me through the link to your your daily brief and we'll put it in the show notes. Last kind of question I have for you is you've been out of Microsoft a year? Is there anything you miss?

Ashish Bhatia 37:20

I there are a lot of things to miss about Microsoft. Do you remember like the the MVP summits and the power platform conferences just the amount of kind of mutual admiration admiration that you have with your customers and that just bonding you don't get to see that in the B2C space at all. I mean as I said right your customers are everywhere and you only see them through a data lens so there is a veneer between you and your customers but again that personal connection that touch with your customers and just having those opportunities to sit chit chat have dinners go out for drinks all of that stuff uh was uh super interesting I again miss working with my colleagues uh I'm I hope they're doing good a lot of them have gone out and are doing wonderful things for themselves uh so the world has changed for a bunch of us uh yes but it's interesting to go back and look in times and good times that you spend there and kind of cherish them sometimes

Mark Smith 38:33

You've been listening to the intelligence age on Mike Smith the Intelligence Age is already here the question is what we choose to do with it