It's a Renaissance Woman's World  ## Document Notes ## Highlights - **Database Designed For AI Agents** - Agentic databases simplify AI agent data by combining vectors, relational data, embeddings, and history into one engine. - Amal highlights TigerData's hybrid search and zero-copy forks as elegant simplifications for agent workflows. Transcript: Adam Stacoviak This is the year we almost break the database. Let me explain. Where do agents actually store their stuff? They've got vectors, relational data, conversational history, embeddings, and they're hammering the database at speeds that humans just never have done before. And most teams are duct taping together a Postgres instance, a vector database, maybe Elasticsearch for search. It's a mess. Our friends at TaggerData looked at this and said, what if the database just understood agents? That's agentic Postgres. It's Postgres built specifically for AI agents, and it combines three things that usually require three separate systems. Native model context, protocol servers, MCP hybrid search, and zero copy forks. The MCP integration is the clever bit. Your agents can actually talk directly to the database. They can query data, introspect schemas, execute SQL without you writing fragile glue code. The database essentially becomes a tool your agent can wield safely. Then there's hybrid search. Tiger Data merges vector similarity search with good old keyword search into a SQL query. No separate vector database, no elastic search cluster, semantic and keyword search in one transaction, one engine. Okay, my favorite feature, the forks. Agents can spawn sub-second, zero-copy database clones for isolated testing. This is not a database they can destroy. It's a fork. It's a copy off of your main production database, if you so choose. We're talking a one-terabyte database, fort, in under one second. Your agent can run destructive experiments in a sandbox without touching production, and you only pay for the data that actually changes. That's how CopyOnWrite works. All your agent data, vectors, relational tables, time series metrics, conversational history, lives in one queryable engine. It's the elegant simplification that makes you wonder why we've been doing it the hard way for so long. So if you're building with AI agents and you're tired of managing a zoo of data systems, check out our friends at TigerData at TigerData.com. ([Time 0:00:42](https://share.snipd.com/snip/76cb4b80-f5d7-4fc9-94be-4ee2940e9ff3)) - **Building Infrastructure For Physical Engineering** - Amal describes Istari as an infrastructure platform for engineers building physical things like planes and rockets. - She explains the platform connects disparate engineering tools, enables AI pipelines, and preserves compliance and IP control. Transcript: Amal Hussein Yeah, let's start with what I'm building. What is it you do here? Yeah, I know. I mean, we were actually sort of like semi in stealth last time I was on the show, so I couldn't say as much as I can now. And even now I'm like, but we're essentially an infrastructure company for people that build in the real world, right? So think of like your mechanical you building? Engineers, your aerospace engineers, etc. They use all these different kind of digital engineering tools, super expensive licenses and super, you know, old code bases. You know, some of these software, like they were, you know, some of the, some of the software that's still used today to build like your airplane was like first invented with like the Apollo launch at NASA, like, you know, like no, like no kidding. So with all these kind of disparate tools, there's all, you know, everyone has their own favorite provider and this and that. And so what our kind of infrastructure platform allows you to do is kind of connect all of your data together and kind of give you situational awareness and kind of you can then pull data Out and extract into standard formats. And then you can use AI to do all kinds of things. You can build custom workflows. You can, you know, make sure that like as you're building things, things are still in compliance, right? Because you're able to kind of pull data from like all these different models. And so, yeah, so essentially it's a platform for enabling that kind of a digitally threaded workflow for people that like haven't really been able to do this type of stuff before. Like People are emailing each other files with lots of really sensitive IP, as you can imagine, people building airplanes and other things, rockets, all kinds of really sensitive IP. And there's no easy way for people to collaborate with each other across the same team, let alone being able to actually collaborate with vendors, right? And being able to share like tire specs with, you know, the person making the tire for the, for the plane, for example. So all this, this software, this infrastructure software that we build at Astari is installed on your network. And so that's the other cool thing. So we're like a GitLab in that, like, installed, self-hosted software, you know, works on your cloud, like all the clouds, including all the gov clouds. Right. And one of the things I'm responsible for is kind of our integrations platform. Right. So how things get in and out. And we have agents and we've got kind of a, know, a whole SDK for people to like write their own integrations. We have a number of ones that we maintain and write. And so in addition to kind of owning kind of like an ecosystem platform, I also own kind of an underlying data platform in which I can't get into too much. ([Time 0:11:49](https://share.snipd.com/snip/ee86a535-602e-4e2d-b50c-4b0fcde2eab7)) - **Moon Dust Vacuum Case Study** - Amal recounts Blue Origin using Istari and NTOP to iterate on a moon-dust vacuum part with AI pipelines. - She reports dramatic improvements: faster iterations and higher quality for hardware design workflows. Transcript: Amal Hussein You know, they were designing a part for a, well, really they're building a moon vacuum. So basically Blue Origin is trying to like survive the lunar night. No one has been able to survive the lunar night because of the huge temperature fluctuation. And it's two weeks long. No one's been able to stay on the moon for that long or nothing has been able to survive. And so they're going to be sweeping up regolith, which is like moon dust. And they're using that to like power up a battery. And so, you know, to sweep up this regolith, like it's like a very harsh substance. It's not from Earth, obviously. There's a lot of like, you know, unique composition to it. And so they need to design the correct kind of cylindrical shape for this moon vacuum that's going to sweep up this moon dust. Blue Origin leveraging Astari's platform, along with, you know, NTOP, another partner of ours, but they're an actual digital engineering tool. Like they're an actual tool provider. Like they work on like CAD software, basically, parameterized models. You know, they were able to leverage setting up these AI pipelines to do all this rapid iteration, right? Where you can pull data from their models, give it a bunch of inputs, right? Because our platform is code first, so you can string together these complex workflows with AI. And what you can do with Astari is you can set up boundaries for AI. And that's like one of our huge value propositions is like we eliminate AI hallucinations, right? Where you can like sort of vibe code or rock it by, you know, because you're able to set the boundaries pretty tightly around like specifications and make sure that like things are still In compliance. And so, you know, Bluardrum was able to kind of leverage AI pipelines to do a bunch of iterations with our software. Our software like helps you, you know, with all of that, you know, connectedness with your data, setting up boundaries for AI. You know, we have an obviously MCP server that like lets you do all the things. Right. And so they were able to kind of do all these iterations with the, with the part developing the part, which kind of like, was like, you know, 75% faster, 40%, like better quality, et cetera. ([Time 0:16:21](https://share.snipd.com/snip/5c31fa5b-0ed6-4c05-b5db-c151320dfa1a)) - **Balance Process With Practicality** - Do keep processes light where possible and add structure where it matters; prioritize productivity and clear communication. - Amal uses Jira, roadmaps, and balances past tech debt, present delivery, and future planning as a leader. Transcript: Amal Hussein For sure i mean you know i mean i i'd say like you know this is where like making the sausage is kind of like boring in the sense that like there's nothing i you know i feel like it's you know Everyone's doing the same thing across companies just in different ways. So I think for us, we're somewhat light on process where it doesn't matter and we're not so light on process where it does matter, as you can imagine. But I think for me, we're not really dictatorial around agile or ceremonies or whatever. It's like, do what's productive and communicate. And so yes, Jira, unfortunately, I would love to be on Bivotal. I think it's something we are considering in the future. But Jira, lots of Jira, lots of spreadsheets, lots of time road mapping. I've talked about this before on other podcasts, but as a lead, you're splitting your time between the future, the present, and the past. The past because you're cleaning up tech debt and there's old bills that you need to pay down on your code base and processes and docs or whatever else, other gaps. And then the present because that's the day-to And then the future. And as a lead, I I spend a lot of my time in the future cause I'm kind of like, you know, I'm like in the expedition ahead of the team, kind of exploring new grounds, finding like, okay, like, Yeah, this, this, this looks good. We can start making camp here, you know, like that's like, that's me day to day, but yeah, I'm doing, you for me, it's a lot of planning, a lot of design work. I'm obviously, we're kind of still a small enough company where even in a leadership role, you're very technical and hands-on. And I really hope that I never have to change that about myself. Because I think it's really important to be able to do the work of the people that you are leading, because if not, you are completely out of touch. All kinds of bad things happen. You also have no empathy for their problems and all of that. So just a lot of architecture, a lot of design, code reviews, unblocking folks. ([Time 0:40:13](https://share.snipd.com/snip/eb1ac37c-6e01-4a8e-9b05-8d016bef26ad)) - **Accountability Scales With Role** - Leadership shifts from answering technical details to owning outcomes and accountability across teams. - Amal contrasts being responsible as a principal engineer versus being accountable as a director. Transcript: Amal Hussein I think what I'm really most excited about these days, Jared, is just teams. Like, you know, you know, the power of teams is like really becoming apparent to me because like I figured out like, wow, like, you know, how do I just enable people to do their best work? And I feel like I've like kind of found the secret sauce of at least that's working for me right now and my company. And, and it's just really incredible to see people just kind of go and like the, the kind of results that you see when people are just like, kind of unshackled and unburdened and where People feel like they have agency and when people feel like they can, you know, they're engaged and they care, you know? And so in that sense, like I'm having a lot of fun with teams right now. And, you know, and the muscle that I'm, you know, always going to be trying to grow is, you know, how to be, you know, how to continually improve my leadership with teams. Right. Because I think for me, it's, you know, I'm now at this new mastery level where I'm accountable and responsible for the work of multiple teams. I've done that before as a principal engineer, but I was not accountable. I was responsible and now I'm responsible and accountable. Adam Stacoviak And so it's like a whole. How is that different? Exactly. Spill it out for me as if I'm like 10, not five, but give me 10. Okay. Yeah. Amal Hussein I mean, so I think it's one thing to like, I'm the person that has to like go to executive leadership meetings and present like our roadmap or present like, here's what we were delivered Or here's what we weren't able to deliver and why. Right. And so at the end of the day, it's on me to answer to like why this is either good bad didn't ship shipped late you know what whatever right whatever it is and so I think that's the difference Um you know versus kind of a principal engineer I would never for me anyway as a leader I would never ever put any of that on a on an IC like any any anything that goes well is for the team to Own and celebrate right anything that doesn't go well is like bubbles up and that's on leadership like that's always failure always goes up literally the wrong direction kind of thing Yeah exactly exactly and so i think um so i think in that sense like it's it's it's very humbling and it's a responsibility. I take very, very seriously. I'm like enjoying it and, and yeah, learning a ton, you know? ([Time 0:42:28](https://share.snipd.com/snip/3914c63e-4a2f-486e-b1e9-56ea2a487191)) - **Schedule Deep Mastery Blocks** - Try to pick one technical topic per quarter to deeply master despite AI-assisted upskilling. - Amal warns AI can atrophy deep skills and recommends intentional focused practice to regain mastery. Transcript: Amal Hussein Like upskilling quickly on something new but i feel like i don't have the time especially as a busy manager now like to like do that deep mastery of something. Right. And like, and you know, you master something through like, you know, suffering a little, burning yourself a little, things like that, you know? Um, and so I, I feel like mastery is kind of like a little bit of a dying art, you know, uh, in general. So like, um, so I want to figure out, like, you know, I think I want to maybe pick one thing, a quarter or one thing every six months to try to really deeply master. ([Time 0:45:18](https://share.snipd.com/snip/71b5d24b-3b5f-42e8-b964-5e8d7455f811)) - **AI Changes Workshape, Not Human Need** - AI increases productivity but doesn't eliminate the need for human roles in product and safety-critical domains. - Amal emphasizes humans remain essential for product judgment, safety, and cross-discipline coordination. Transcript: Amal Hussein Weird times. I mean, you guys are in it. You tell me, Jared, like what's happening? You know, what's the sentiment these days? Like what's the temp? Still not great. I know the Zerp days are over, but still not great. Jerod Santo Have gotten way better over the last year, even if the models haven't gotten way better. I think that's the current trend is the tooling and the ability to give tools to the models and the, just really the good old fashioned software engineering that has been built up around Models that are like nominally better than they were a couple of years ago, I have just really unlocked a lot of productivity. And recently, so much so that, I mean, a while loop and a tool and a large language model can do amazing things right now. I think we're all seeing that, but it's unlocking more productivity from smaller teams, from indie developers. But what Adam said hits the nail right on the head. I mean, you still have to be able to create, maintain, and build. Hopefully not in that order. I did it out of order, but you know what I mean? Software products, not just software. Yeah. Amal Hussein Yeah. And humans need humans, guys. Can I just put that out there? Humans need humans. Stop trying to take humans out of the equation in the sense that this notion that there's going to be one person running a billion dollar company like just barf at that okay um and also Adam Stacoviak Just yeah like it's more possible than it has ever been though right you have to admit that yeah i think it's not necessarily impossible that somebody gets there is it healthy no i mean Is it possible yeah it really is possible but that person's gonna be pretty talented person they're not just stupid like at that point like you could scale right you would do it for bragging Amal Hussein Or ruthless maybe all those things yeah it's a billion dollar company that runs in one time zone got it But no, I mean, it's all good. I think I'm just a little bit bitter about all of the, oh my God, there's going to be no software engineering jobs in six months. ([Time 0:58:32](https://share.snipd.com/snip/e2845380-a600-4f9f-8050-92da5976ae1a)) - **Efficiency Often Creates More Demand** - Jevons paradox suggests gains in efficiency from AI will drive more consumption, not less. - Jerod argues increased productivity will create more bespoke software and thus more demand for engineers. Transcript: Jerod Santo Think the shape changes yeah the org shape the team shape those things change necessarily yeah but there's just more smaller i think there's more smaller things and less big things Adam Stacoviak We were saying before in our call today jerek we mentioned the the piece of team size there's the idea that you know back in the the getting real book from 37 signals i believe they prescribed A four-person team maybe a five-person team like four engineers one designer or some sort of leader i believe was their prescription and then jeff bezos you know kind of coined the idea Of a two pizza a two pizza team i believe it was is what it was which the idea was like four people or less uh could you consume two pieces with four people or less probably right i think it's Jerod Santo More than four people i think it's closer to six it's two pizzas well you haven't seen the way we eat here in the midwest yeah i don't i mean yeah i mean i'm speaking from experience here Adam Stacoviak Okay i'm gonna to have a pizza guy, I'm one person. Multiply me by the amount. Actually, Adam and I together are a two-pizza team, it turns out, you know? Amal Hussein Yeah, that's so funny. Adam Stacoviak So I think we'll see the pizza team shrink, maybe to two people. Amal Hussein I think there's certainly a trend and there will be a trend of, you know, smaller teams, right? Because I think, you know, but I think there will be lots of small teams, right? And to your point, like definitely the craft is changing, you know, as it should, right? It's, you know, and I urge everyone to like, you know, change along with it, right? Right. You know, so it's definitely, that's very important and it's uncomfortable, but you'll be better for it, you know, if not, you know, there's a lot of, there's a big need for blue collar Jobs. If you're in North America, you know, like retrain as an electrician or all kinds of, you know, all kinds of need for other high paying, which could be more lucrative in certain scenarios Jerod Santo As well. And perhaps even a more enjoyable way to make a living. Amal Hussein Absolutely. Jerod Santo Certainly there will be job displacement. Things will change. But if you, if you look at Jevon's paradox and you look at these things, like historically, when we've been able to do more with less, which is what these tools let us do, we can do more with Less. Humans have never chosen less like, Oh, we can do more with less. Let do more and it's just always been more more more and that's the the ambition and the drive and the ingenuity of the human race is more so we're not going to do less we're not going to choose That we're just going to do more more more which require more people yeah ([Time 1:02:25](https://share.snipd.com/snip/afc4a51f-096c-4271-9a82-df942d1c3dae)) - **Automate Realistic Verification** - Do automate verification by writing programs that act like real users to validate deployed features. - Adam recommends external test harnesses and simulation of human behavior as table stakes for safety. Transcript: Adam Stacoviak What if you wrote a program to verify the program? So rather than just writing tests in your integration, for example, what if you actually said, okay, let me assume my tool is deployed. Let me actually write a program that acts as I would. I'm going to program it with my intellect and how I would think about it. And it's going to automate future Adam or future Adams to test this thing. Like you got to reframe the way you're thinking about it. You really do. A hundred percent. Amal Hussein I actually, that's kind of the way I prompt now. I always ask AI to write the software for the thing that I like, like if I'm trying to do analysis for something or whatever, I always actually have it write software and tests so that I Like no one's changed. So just like % agreed. Adam Stacoviak Yeah, test passing is table stakes. Having a test harness external from the actual application is not tables. It's now table stakes. Now it's table stakes, yeah. Yeah, now it's table stakes. But like, you know, that's how we have to think about it. ([Time 1:23:19](https://share.snipd.com/snip/f78e24fc-f380-4c10-996e-1faaf1ffab25)) - **Generated Code Needs Testable Design** - AI can generate tests and code quickly, but quality requires isolation, dependency injection, and proper testing hygiene. - Amal notes many codebases make testing hard because they lack good isolation and design for testing. Transcript: Adam Stacoviak Yeah, test passing is table stakes. Having a test harness external from the actual application is not tables. It's now table stakes. Now it's table stakes, yeah. Yeah, now it's table stakes. But like, you know, that's how we have to think about it. Jerod Santo And it's darn near free. That's the best part about it. I mean. Adam Stacoviak It is free, Jared. Jerod Santo I mean, it's so low cost. It's basically free. That's why everything is kind of cool. It's like, yeah, might as well. Remember how hard it used to be to write tests? Amal Hussein Yes. Because most of the time, you know, I hate to say it, but your code's a little shitty and not well isolated, so it's really hard to write good unit tests, and then you're like, oh, you know? ([Time 1:24:05](https://share.snipd.com/snip/4402c71c-5a7f-4c2f-81ad-5d065fc1bc9b)) - **Istari's Multi-Language Distributed Stack** - Amal lists Istari's core stack: Go, Rust, Python, TypeScript, and Kubernetes for installable distributed software. - She explains the product runs a control plane, data plane, and agents across clouds and customer networks. Transcript: Amal Hussein But, but yeah, core platform is in those kind of more kind of traditional languages. And yeah, I think in terms of like, what's next, I think, you know, like every scale up, you know, we're kind of at, you know, we're just kind of growing our platform. What's really great is you know we have this like straw man that was built and then we kind of like iterated on that and so now it's like it's interesting to see like I'm at the like maybe Fourth big juncture of the of the platform just you know and to kind of be part of like a growth cycle for like living, breathing software, um, going through so many big kind of big growth Spurts has been very fun, right? Because like, I've never joined a company this early, you know, usually we're just kind of refining like much more narrower verticals. And so to kind of be like growing horizontally, to be getting taller and bigger is fun, you know? And so that's kind of what, that's, that's what I'm, that's what we're all kind of going through, building more, building more services, building more features, making our platform, Like just, you know, integrate with more things. And just in general, like, yeah, we're doing a lot of great stuff with, you know, supercharging our kind of AI workflows. And yeah, I don't know. ([Time 1:27:31](https://share.snipd.com/snip/c6dc571c-7921-44fc-b9b5-6cd1a3e5932c)) - **Own Your Writing And Syndicate It** - Do revive personal blogs and use POSSE (publish on your own site, syndicate elsewhere) to regain community control. - Jerod and Adam recommend Hugo/Hugo-like static sites and syndicating posts to social platforms for discovery. Transcript: Jerod Santo But will you like refresh your blog? Will you like? Oh yeah. I'm refreshing my blog already. So you refresh yours. I'll refresh mine. Adam has, and we'll just write on our blogs and we'll, we'll talk to each other there. And then we'll follow our, you can join the posse party. You just, you just post to your blog and then your blog syndicates it out to all the networks so people who don't want to come to your blog they see that there maybe let's stop over have a Amal Hussein Read i love that idea you know i think that's like that's i think that's very the indie web is cool again indie web is yes what would be a good platform for that? Adam Stacoviak Mine's already in Jekyll. Jekyll works. No, is it in Jekyll? I just converted from Jekyll to Hugo. Jerod Santo Yeah, I was going to use Hugo. Hugo is super simple. You can have Claude just say, just take my, you know, I have a Jekyll blog. This is like literally what I did. I have a Jekyll blog. Here it is. You're in its directory. This is your prompt? You're going to inject this? Something like this. Yeah. Yeah. Copy paste this. Okay. And I'll put some special characters in there. Convert this to Hugo, dude. Like quick. Dude. I want to get blogging. That's basically what I said. And then I said, thank you. So awesome at the end. Dude, where's my blog? Dude, where's my blog? That's a really good title ([Time 1:33:10](https://share.snipd.com/snip/d0284861-9a96-4827-bd96-8409183607e8)) - **Aerospace Conferences Feel Academic** - Amal describes attending AIAA where aerospace conferences are formal, paper-driven, and deeply technical compared to tech meetups. - She contrasts that seriousness with software conferences and feels humbled by aerospace engineering rigor. Transcript: Amal Hussein And I think I feel the weight of that importance. Like, you know, when you are like changing, when you know you're changing something in a big way that has like a big ripple impact like that's how i feel you know because it's like we're Doing something so revolutionary for this industry and we're doing it in a way that's so you know it's decentralized and it's very open and like you know we have like you know you plug Into our platform we don't care about like this tool or that tool. We're kind of agnostic about so many things. It feels like we're doing something good. And I think the weight of that is kind of what's translating into the seriousness. But I feel like... No wonder you won't let people vibe code. I mean, come on. Yeah. It's okay. It's okay. I'm fine. You know what it is also like people building physical engineering. Like I went to like my first big physical engineering conference, which is AIA this past January. And people are wearing suits and pants and you know how, and they're so serious. And every submission, like there's like thousands of talks. No, no, no. Literally thousands. And like, everything comes with a technical paper and like, here's the bibliography. And here's basically like this art, you know, peer reviewable, you know, like artifact that you can, you could have. And, and I'm like, man, tech conferences are just like giant parties disguised as conferences. And like, here I was at this very big conference that was absolutely not a party. And it was like a very academic conference. And so, you know, you know, people do take themselves a lot more seriously and rightfully so. Right. Like, you know, like building things that like impact people's lives, you know. So so that that that that checks out. ([Time 1:38:53](https://share.snipd.com/snip/dddc5fc3-81a2-44fe-95f0-4bbb11abd624))