The Analyst That Never Sleeps: Inside Farsight’s AI Revolution

In this episode of RRE POV, Raju and Will sit down with Samir Dutta and Noah Faro, co-founders of Farsight, an AI platform reshaping how financial institutions operate. Raju first connected with the Farsight team through a fun coincidence they all happened to be part of the same MIT fraternity, Phi Beta Epsilon. Though they were there years apart, that shared bond made teaming up on this company a natural fit. The guys dive into how Farsight helps analysts move faster by generating slide decks and models in minutes, not weeks, and how it’s already driving real revenue for clients. With proactive AI agents on the horizon, Farsight is quickly becoming a must-have tool for modern financial workflows.

What is RRE POV?

Demystifying the conversations we're already here at RRE and with our portfolio companies. In each episode, your hosts, Will Porteous, Raju Rishi, and Jason Black will dive deeply into topics that are shaping the future, from satellite technology to digital health, to venture investing, and much more.

I go into my room and I'm like, okay. They, I escaped. I don't know what the hell happened. And I go into my drawers and in place of where my like socks were, were was a drawer full of goldfish. And they, like every single drawer had like goldfish in it. And I'm like, I don't, I don't know how to get this out.

Like I don't even know what to do. Um, oh my god. The snack goldfish or the animal goldfish? No, no real goldfish living, you know, like little tiny. Yeah, yeah, yeah.

I am Will Portus. And I'm Ra Rishi. Welcome to R-R-E-P-O-V, the show in which we record the conversations we're already having amongst ourselves, our entrepreneurs and industry leaders for you to listen in on.

Hey there, listeners, this is Ra Rishi, along with my partner Will Portus. Today we'll be having a discussion with two of the three founders of Far Site, Samir Datta and Noah Farro. And just for Context, far Site is an AI platform that automates workflows, insights, and decision making for the financial services industry.

It just raised a $16 million series a round that was led by Signal Fire and Co-led by ROE. So hey guys, nice to have you on the show. Yeah. Thanks for having us. Good to be here. Yeah. I think this podcast may very well have the highest average iq. Of any session we've done, I, I'm just thrilled to be here with all you MIT guys.

I know, I'm, I know I'm gonna learn a lot from the three of you. Yeah. I mean, Samir and Noah went to my alma mater, MIT and, and Will went to this Noname, west Coast college. You know, Stanford? Yeah. Who knows, like, it might become something. Hopefully that sentiment persists at the end of, uh, at the end of the session.

I know it. Like I said, I know I'm gonna learn a lot from you guys. Okay. So, you know, you can't do any podcast anymore without talking about ai. Now it's just the nature of the beast and we actually have, I think, a fantastic AI company here. But like Samir. Um, we're gonna talk about sort of the inception of the company, if that's okay.

I'll just ask you a question. So, you started your career at one of the, you know, top investment banks, Evercore, and then joined General Atlantic as a growth equity investor. So what problems did you see there that kind of led to the founding of Farside? I would recap that, you know, four or five years of my life with, with two main buckets of problems.

I think one more obvious than the other. The obvious one being there's just so much grunt work and repetitive work that goes into, into both of these jobs, especially at the junior and mid-level, but, but even at the senior levels to an extent. And that's what causes sort of the infamous late nights banker and investor health issues.

You know, just a whole host of problems there. But, but again, I do think that's the, the more obvious sort of problem. The second problem, which I think actually presents a much larger opportunity. Is the fact that these institutions sit on so much data and so much prior work that is just not leveraged nearly to the extent that it could be.

So, you know, like when I was at Evercore, every deal that you work on has a similar flavor. It is something that's been done in the past. By that group or, or by that partner. And, and you know, so much of it just lives up here in, in senior partner's minds and as junior or mid-level folks that are newer, um, to the industry and newer to the job, it's really hard to find the good reference material and find the right insights, you know, for which the work has already been done.

But surfacing that at the right time and being able to leverage that at the right time, I think is a massive, massive problem. And, and you know, obviously at FireSIGHT we're, we're looking to solve both of those. Yeah. Can, can you give us an example or two? Um, just for context, like, not all of our listeners are gonna be from the financial services industry, so, um, you know, as a reference path, I mean, I think investment banking is definitely straight down the fairway as well as other, you know, financial services institutions.

But, you know, give us an example of one or two use cases. Um, of things that ForSight does, uh, to sort of, uh, you know, improve workflow and productivity at, you know, an investment bank or a similar financial services institution. Yeah, happy to. And, and so for, for broader context, you know, everything in investment banking and, and private equity and these adjacent industries revolves around deal making, right?

And so you think about deal making, there is a deal life cycle that goes along with that. And so I'm an investment banker. I have my set of relationships and the industry that I cover. I'm constantly looking to stay in front of those relationships and be top of mind and, and serve as a thought leader with the goal being that when.

Some of my client relationships are ready to transact. Either sell the company, buy another company, sell part of their business, raise some capital. I want them to come to me as the partner and and engage my services. And when that happens, just as an example, a tangible example of one thing Farside does anytime there's a transaction, there's all of this collateral and deal material that that needs to be put together.

Literally a hundred page slide decks with these massive behemoth Excel models. Tons of emails exchanged far site based on, you know, the similar deals that the bank has done in the past will actually create the first raft of all of that material. Then give you the tools to very quickly get to the second, third, fourth draft.

So we're taking a quantum of work, uh, hundreds of pages, all these Excel models that take literally a month or two months to do manually today. And we're now getting that first draft within a span of 10 minutes, often with better insights than, than, you know, human teams would've come up with, uh, over that one month.

That's unbelievable. Uh, I need you to solve a bunch of other issues in life when you get done with this one. Like marriages, you know, other simple things, uh, also require a lot of memory and can be, uh, accelerated. Uh, okay, so, so that, that's really, really fascinating. Like when you talk about the prior art.

You know, if I implement far site at web at my bank, how quickly can I get value? Is this gonna take like, you know, six months to kind of learn all the things that I've done in the past? Or do I get value in a, in a shorter time span than that? Yeah, I, I mean, great question. I, I would say like. In any financial vertical that, that you're talking about.

There's 10 or 20 workflows that are just so commonplace that that, and, and our systems have become so good at that set that we're ready to go within a week or two on, on that set of workflows. Now there's a second set of workflows that are maybe more proprietary to each. Bank or each fund, what have you.

Um, and so that, that's something that might take a little bit longer. But you know, I, I think with all software you have this adage of crawl, walk, run. You know, we can be crawling within a week. Walking within a month and running within sort of three months. So, uh, usually FireSIGHT is not the gating item in, in enterprise deployments.

So, Samir, I I have to ask, I, I'm thinking about some of the, the environments that you've sold into and a lot of those senior partners. Who've been, you know, maybe they're m and a bankers, maybe they're capital markets advisors, but ultimately they've been client gatekeepers for a long, long time, and they guard those relationships jealously.

They take those relationships with them throughout their career and they don't. Just let anybody into the conversation. What's it been like engaging with those people? Like how, how, how, how do they perceive a capability like ForSight? Yeah. I, I, I, I think that's a great observation and, and I think what we've learned is there's sort of two different user personas that we cater to.

So one is sort of the efficiency, um, workhorse persona where, you know. They don't really care about that stuff. They care about work being done quickly, accurately, uh, and in a way that makes them look good. What you're referring to is the second user persona, which is, um, I'll call it the sort of strategic user, right?

So they have all these ideas in their mind. They're much more close to the revenue generation piece. And the way to play with those users is to actually give them, uh, the sense of configurability. So they will get their own, you know, their user portal for FireSIGHT will look different than their partner in the same group.

Ah, interesting. So it becomes really personalized to the special little world they've built for themselves. And they, they're less beholden than never actually to the generalist pool of analysts who they used to have to pull in to work on things and, and try to impart all those ideas too. Exactly.

Fascinating. Wow. Yeah, I mean that, that, that's, that's super interesting. So, you know, look, AI is out there, co-development happens rather quickly in this ecosystem, and we know there's a, there's a few other players that are sort of trying to tackle similar types of things. I mean, what makes FAR Site, you know, sort of some of the capabilities that we have, um, unique and difficult to replicate?

Yeah, great question. I, I think the single biggest thing is that, uh, there's a lot of people running at this problem. 99.9% of them have built chat first platforms. So think of that like a chat GBT or a perplexity built for finance, which is, which is very useful. But you know, our view is that that may be a little bit limiting in terms of the actual breadth of work that you can take on.

And having been sort of a banker and, and and growth equity investor, myself, uh, we've taken the approach of being workflow first. The first thing you get out of far site is a entire slide deck or entire set of Excel models within seconds. And that tends to be sort of the eureka moment for a user where I can type in three things and visibly see something that I know takes me a week or two weeks to do, generate in front of my face exactly the way I, I would do it today.

Um, and so it's really that end-to-end nature of these quantums of work. You know, you can get from a chat platform and generate a slide here or there, or generate an Excel export. But really far side's competitive advantage is in the end-to-end nature of what we do. Because that's the storytelling piece.

That Exactly, no, I'll ask you a question in the middle of this. Um, what, what, because I love it. I mean, generating PowerPoint. It's hard, but why is it hard? Like, why? Like you would think that, okay, so I take a slide deck, it's a template and I try to populate the pages, but like I know like mentally that's kind of exhausting just me for me to personally do.

But what, what makes that challenging and how have we cracked that? Yeah. I think the biggest challenge for actual generation of materials is that, let's say you, you take an LLM, you ask it a question, it gives you a textual response. When you think about how much information actually goes into a single slide for a PowerPoint that may be tens or hundreds of queries into a chat bot.

Or if you look at a hundred page slide deck, like what smear referenced earlier. We're talking about thousands or tens of thousands of queries. So when it comes to actual generation of content that is not just textual, but is spatially related, or includes charts or financial analyses or other complicated metrics and components, layering all that into one cohesive end-to-end workflow generation is extremely difficult to manage.

LLMs traditionally do a lot better on more discreet problems. And if you're able to take a PowerPoint or an Excel or something like that and make it a more discreet problem and break it down, then you can make it a holistic end-to-end generation. But if you just take a hundred page slide deck and throw it into GPT and it's about meta, and then you say, Hey, replicate this same slide deck for Google, it's not gonna come anywhere close.

And a large reason for that is the complexity of the task. That's why it takes. Tens of analysts to do this over the course of months rather than one person churning it away on their computer. Yeah, I love that. Um, and no, I'll just another follow on question there, like, if I work with investment bank A and then I've got a client that's investment bank B, how do I ensure that there's, like, how do, how does Farside ensure there's no data bleeding across, you know, the ecosystem?

Um, and that people sort of keep the sanctity of their work. 'cause I, I would imagine these investment banks feel like what they do is relatively unique and proprietary versus, or at least some of it versus a different one. Yeah, 100%. From a deployment standpoint, everything that we do is in, at the very least, their own private clouds.

Within Far site, we segment off our system specifically for that customer. Nothing that's ever interacts with models or our own data systems or anything like that crosses between their deployment and other deployments between our other customers. So they never have to worry about that from that standpoint.

But we can also deploy things on premise to if they want to deploy in their own clouds. So when it comes to data security, we're as secure as it gets. Uh, we don't have any bleeding across customers. Um, and they can ensure that when it comes to a security perspective and their ability to maintain the integrity of their private client files, for example, we're not gonna do anything with that, that compromises that.

Okay. I love it. So, Samir, I'm just going back to you. Uh, no, I'll ask you a bunch of technical questions later. I'd just like to get into the weeds on AI 'cause my favorite topic. So, uh, it just is, um. So I get PowerPoint's hard and I, I see a few competitors out there that are, you know, basically query based architectures and, you know, I, I've played around with 'em and it's like, okay, you know, so I got a little piece of data.

I kind of, I gotta cut and paste it and put it in, generating that end product in one fell swoop. Is awesome and completely agree with her. Anything else that's sort of nuanced in what we do here at ForSight that you want to tell our listeners about? 'cause I'm assuring you that some of your clients or prospective clients are gonna be listening to this.

I'd love to give you an opportunity to just kind of share some other, you know, sort of value propositions that we got. Yeah, I, I think it largely revolves around what we'll broadly call sort of strategic benefits or idea generation, right? So there, there are. Two types of work in, in this industry. One is efficiency based.

So, hey, I need to pull these logos, or I need to develop a first draft of this model. But generally I know the bounds of what to do and it can be very rule based, right? So AI is maybe helping me do it more error free, and certainly helping me do it faster. By far the area where I think we as a management team and, and overall company spend a lot of time is on those strategic benefits.

So how can I, uh, surface better ideas to my clients? How can I be the. You know, thought leader in this industry from the investment banking perspective, and, and be sharper than all of my competitors. How can I come up with better lists of buyers for a potential transaction? Right? So, you know, we, we've had multiple instances now of far site showing up and, and coming up with lists of buyers and ideas that the Cuban team wasn't thinking about before.

Those transactions actually get consummated. So now, now, like ForSight is directly attributable to, to millions of dollars of fees, uh, for our client base, which is, you know, at the end of the day that that's why we're gonna be successful. I think the efficiency piece is a great wedge and what keeps users coming back.

But, but those stories and those wins at the high level are gonna be. How you win this market. So that's, that's really powerful. I mean, the word of mouth effect of that within the, the advisory communities got a starting to be created a, a snowball effect for you. What, what, how is that manifest in terms of your pipeline and like where, where are you guys in, in your engagement with the market?

Right now, are people just coming to you out of the blue saying, Hey, we heard about this, or we just lost a big deal, because we know those guys were working with FireSIGHT. I would love to get to that point. I think we're close. I, I don't think we're quite there just from a, from a branding and marketing perspective, but, but I, I hope in the near future, I think where we are at is quite a bit of, of just generic buzz.

I, I would say across the street. Um, and I think the big challenge is, is. You know, cultivating that into getting the decision makers much, much quicker, getting the buyers much, much quicker. Because if you think about everybody that works at a bank and is a banker, maybe 0.2% of them actually have the authority to, to onboard ForSight.

So obviously that brand market, these inbounds are great. Um, and it's, it, it's been awesome to see sort of the feedback on demos and trials and, and from customers, but. So, so one of the reasons we love selling to financial services historically is because the take up on innovation is often faster than, than in other vertical markets.

And if you can, if you can onboard a capability today and make more money tomorrow. People will move fast to do it, and that's often been true on the trading side. It's been true actually in the back office side of things, but now we're talking about something right in the front office where there are millions of dollars in fees at stake.

What's the kind of proliferation like once a customer starts to onboard with far side, like how does it, how does it pollinate? Within, within the organization? Yeah. I, I think a, a large part of it is, is, is organic, right? So a lot of these success stories. Of like, Hey, you know, I, I got staffed on something at 10:00 PM last night, but I had far sight and it actually didn't ruin my entire night.

Like, that's the wildfire, uh, that's great. At the junior levels. And then, you know, at the senior levels it's like, Hey, you know, the, this far sight thing is pretty good. And look, we're we're pretty realistic, uh, in terms of where the product is at today, which, uh, at the senior levels, which is like look like.

Maybe half of the ideas are, are not gonna be great, uh, that, that come from the system, but the other half, you know, provided we're, we're getting you two or three good ideas a month that, that you're actually acting on, that's hugely valuable. Right? Oh yeah. Um, and so it, it, it becomes something, uh, I think will, to your point earlier.

Partners like to ge keep their relationships. Um, but you know, as, as a whole, within the group, partners will collaborate. 'cause you know, the overall group gets a bonus at the end of the year. And if Partner A got a few good ideas using the system, they'll they'll tell partner B about it. Yeah. The cross bank pollination is much heavier at the junior levels, at the senior levels.

It's, it's nearly nonexistent, but yeah. Yeah, that's probably, they don't even, they don't like to even look at each other, you know? Um, what's the, I mean, what's the usage pin? Like, let, let's say you, you know, your earlier clients, you kind of deployed the product and I don't know what usage metrics you, you know, kind of focus in on per client, but how have you seen that ramp across the organization?

Yeah. Uh, so, so. You know, and obviously our products has evolved since our initial customers as well. And so, uh, one thing we used to, and still do track very heavily, the first key metric was workflows run per week, right? So how many workflows is the organization running on sort of a per user basis? And that's consistently been.

10 or north of 10 workflows a week, and each workflow is typically an hour or two. So substantially time saving. And I think what we've noticed over the last call it six months is not just replacing the workflows or doing the first draft of workflows that, you know, we're being assigned top down, but actually now more work and more ideas and more meetings being prepared for that, uh, than was ever being done before.

So if you look at the overall quantum of work that's done across, you know, a hundred person investment bank today using FireSIGHT, it's probably double, if not triple, what they were doing before far the same. Damn. Wow. That's, that's fascinating. So, so that's overall like utilization of the talent that they have is effectively multiplied.

With far side. Yeah. I, I mean, I, I'll, I'll actually, I'll, I'll put this in your world, like Roger will, how, how many, how many meetings do you guys have coming up this week? Oh, I mean, it's July, July 4th, maybe it's a little slower, but, but the average week, yeah. No, I, I mean, I, this week I probably have 30 meetings.

Same. Yeah. So, so you've got 30 meetings and, and you, you'll have your team, you know, in the back of your mind you have like, all right, these are the five most important meetings I have this week. You'll have your team do a bit of prep work for that. So that's a prioritization that happens at the senior most level.

Now if your junior team or even yourself is using FireSIGHT and the barrier to getting that work done is near zero, you're gonna start prepping for all 30 meetings. Yeah. Damn. I gotta just talk to our analysts about, you know, cracking the whip here. I, I don't get that, that many briefs I get maybe like five or six out of those 30 as briefs.

I need, I need all 30. I, this is fantastic. I, I'm, I think our analysts are gonna hate you. Um, but this is awesome. Okay, so just one more piece on just like the functionality like. And maybe I, you know, like is there anything you could talk about that you're kind of in the works of working on that you can tease out?

Or is that sort of like, just, you know, like we gotta wait until it kind of shows up at our desk? Well, I, I can tease it and, and subject to editing later here, but, uh, I, I think one thing that, that the market has been, um, really demanding for us and, and, and we can't scale quickly enough, is the idea of the proactive agent.

So, so, you know, we're already sitting on top of. All of these customer relationships, all of this prior art, we, we actually uniquely have the ability to be proactive and suggest new ideas, proactively conduct workflows, uh, and almost operate as an autonomous analyst or associate or VP to the extent where you have this infinitely scalable.

Farsighted system that is servicing all the partners and all the MDs across your bank or across your fund with basically no input. Right? So right now, up till now, the overall AI paradigm has been, and this goes across markets, right, has been I, as the user, need to think of something and then go into the system and prompted, and that's how Foresight works today.

Uh, I think, I think very, very soon. Uh, and, and this is already under. Uh, uh, beta with a, a number of our sort of best customers, uh, is, is the idea that the AI can proactively do this work. So Samir, just to kind of bring that home in a practical sense, are you saying that if, let's say I'm a coverage banker in a sector like aerospace and defense, that that I can count on FireSIGHT to signal to me, Hey, you know.

This stock is up meaningfully and we're seeing really good buying activity among certain names. You should be leaning, leaning in. You should be calling these guys to, you know, other, other players in the sector have announced transactions, have announced m and a in the last 30 days. You know, these are the kind of targets that they're showing interest in.

Like that kind of proactive signaling. Yeah, and, and actually, you know, uh, probably the most exciting part about that. You're exactly right. I think that's just scraping the tip of the iceberg, uh, in terms of the art of the possible here. And as we were talking about before, like we're, we're kind of allowing senior bankers to configure this themselves.

Um, and some of the things that they've come up with, like, I'm not gonna share on, on this call and, and this audience, but, um, it, it's really interesting. It, it's unlocking a new way of thinking and a new way of viewing the world, which just frankly wasn't scalable or, or even feasible before. Yeah. No, I, I love this and.

What about other markets? Like, I'm just thinking about this and, you know, I can see this broad applicability, not just in investment banking, but a whole lot of other sectors. What, what uh, what are your sort of like sequencing of other, you know, sort of spaces that you kind of want to do the same, you know, take far sight and, and, and, uh, grow into.

Generally speaking, I think, I think all the financial services is, is very much in the, in the strike zone here. Um, to your question of sequencing, like private equity is nearly identical in the data sources. Instead of, instead of building the sim, you're now digesting the sim. So, so you know that that's a very natural, uh, act too, and, and we're already doing some work in that regard.

Then, you know, you have these hedge funds is another obvious one. Then you have these like broader, uh, mass market financial services. Products like wealth management, commercial banking. In large part, we're working with the banks that also own, you know, the leading wealth manager and the leading commercial banking arm.

And, and, you know, expansion into those I think is, is a natural next step. But broadly speaking, you know, the, the way our sort of system works is you can throw in any prior art that can be presentations, excels, what have you. And using your own data sources and, and institutional knowledge, you can recreate that for any new situation instantly.

Yeah, I love it. I, I really do love it. Okay, well I appreciate that. And Noah, I'm gonna just, uh, shift to you a bit now, uh, and just talk about sort of the tech stack. Uh, just for our audience's benefit, you know, Samir's, A CEO, and Noah is a CTO. Uh, so this is a little bit of organization around the questions.

So you've been working AI software dev for a number of years at MIT Media Lab, AWS Spritz. AI is moving really quickly, you know, just. Like internal gut check, like how do you manage software releases, you know, with this pace of change? Because I imagine, you know, we leverage a bunch of AI inside of Far Site and, and, um, LLMs inside of Far Site, and the pace of change is so dramatic.

Like how do you think about like, okay, got a lockdown on this, but might have to revisit it, you know, six months from now or three months from now. How are you juggling that? We keep a running list of all of the different AI initiatives and pushes and new state of the arts that come out. We actually have some members on the team take designated time to read Twitter and other resources that publish AI contents that we can stay up to date.

Because every single day there are thousands of new resources or papers or projects, what have you. When you think about like. The past six months alone, there's been three completely different AI agent frameworks that have been released by anthropic, Google IBM, that have completely changed the landscape of how people are actually executing things with ai.

And the rate of change with that is so fast that if you are not up to date on it and keeping track of it and making sure that you're revisiting and actually revisiting those updates, you're gonna be get left behind in in a month. It's unbelievable actually. I, I, I really, I don't even know, it's been a long time since I was a coder or, you know, part of a product development team.

I can't even imagine, like, we to think, you know, back in the day, Hey, you know, this, this code will be good for like three years or five years, you know, or 10 years. You know, and now you're looking at it and you're like, okay, you know, the, it's the efficiency gains that you can get. Can happen overnight sometimes.

And I was talking to one AI uh, company and they basically said, we've replaced our two thirds of our code base three times. Already in the span of like two years, they're just jettisoned old cold base and code base and put new things in there. So I'm glad you know, hey, gotta get the brightest minds in the world to work on stuff like this.

And so I'm glad you guys are here now. Now in terms of ai, inside of Far Sight. Have we built our own stack or are we leveraging open source? Are we leveraging a lot of the platform companies or just a mixture of sort of all of that? It's a mixture. Uh, when you think about the, what is state of the art across the AI landscape, things are very good for very specific purposes.

Um, so we use a variety of different types of models, um, that are both vision and text space. Uh, for us to be able to do everything from AI agent workflows to like just straight question answering or document digestion. There are a ton of different things that we do and we use different models for different parts of that stack.

No one model is. Like going to solve all of our problems. I think a lot of people like to coin the term like GPT wrapper. Uh, I think any company that is actually advanced in the AI landscape is going to be using a multitude of open source and closed source models for very fine tuned specific processes.

And as a result, you end up becoming agnostic to all of the different things that are happening and changing on a day to day. Yeah. And so if you say like, Hey, tomorrow somebody, you know, sort of. Distributes a new model and you are like, okay, that's better, you know, for this task. How quickly can you swap that out in terms of our software?

Sam Altman actually released a quote about this, I think like six months ago or a year ago, that was startups that are building without being model agnostic and very able to switch with the different state-of-the-art improvements that are happening every single day are the ones that are going to die.

Uh, and so at at Far site, we make sure that everything that we do is. Interchangeable with new processes that come out. You referenced like deprecating some things. We've deprecated a couple features and replaced it with completely new ones over the course of the past few years, and that's something that we're continuing to make sure that as we build new features, we're able to augment them rather than replace them.

If you're having to continue to recycle your code base and your different features and processes with every new initiative or innovation that comes out, you're spending way too much time. Taking care of tech debt when you should be setting up your system to accommodate the changes and new processes that come out.

Um, so we, we spend a lot of time on that system design. I love it. I love it. And like, what's your thoughts around generalist models versus specialist models? I mean, how even, you know, sort of open, you know, source tools or getting more granular every day, like they're sort of bifurcating their code base to focus on specific problems.

Is there certain use cases where you're saying like, Hey, a generalist, you know, LLM is fine, versus using something that's much more specialized or fine tuned. Yeah, I think that there are a lot of pros and cons when it comes to whether or not you want to use something very discreet versus something more generalized.

Uh, one of the biggest cons of using something extremely discreet is the lack of ability to pivot on your feet with new innovations like what I was referencing. So let's say for example, you spend a ton of money in resources, fine tuning an open source model for some specific piece, and then now an open or a closed source model that was just released completely blew that.

Initiative out of the water and performance. Now you spent a lot of sunk costs, a lot of sunk time in that specific initiative when you could have figured out a system or developed a system around a more pivotable model. And so that's one of the biggest cons of doing something extremely discreet, but.

There are a ton of different use cases where it's more applicable and less applicable to use generalized AI versus not. And being able to have the technical time and bandwidth to be able to investigate those routes and pick and choose what things are good for, what, uh, allows you to build a more robust system going forward.

Okay. That's awesome. Um, and let's just, you know, touch on usability. I, I, I think part of the challenge in this ecosystem is that the codes are being developed so quickly. You know, you kind of bastardize everything and you know, it becomes an amalgamation of stuff and you can't have this authentic sort of con, you know, sort of simplified ui ux, you know, how have you thought about ui ux in your architecture so that, you know, a, a new customer, a new client coming on board doesn't have to sort of get into all the little tiny nuances on how we do things, but can effectively get working faster.

Yeah, there's a quote in design that I like to reference, which is, when you think about ui, ux and like massive technological change, if you introduce a completely new ui, ux, no one's going to adopt 'cause they won't know how to use it. For us, we are taking a different lens to the user experience where we're generating workflow zero to one rather than chat.

And that may be a new user experience, but it's actually much simpler than way the ways that people traditionally interface with ai. And we've done that intentionally. When you present a chat bot or something like that to a user, they have to figure out what question to ask, what information they need to provide that chat bot for it to answer the question.

They then need to get the response and then actually do the work. That's a lot of steps that they need to figure out and spend a lot of like. Thinking time on versus if we take care of that for them, it simplifies the actual interaction and makes it a lot easier to A, become accustomed to and integrate into your day-to-day workflows.

And B, it also becomes a lot stickier because now this is how they do it. Instead of them having to figure out all of the fine grained little steps, they're now just thinking about iteration and refinement after far side generates materials. Yeah, I loved it. I loved it. We saw, I used a tool, you know, before we invent investing and I, I kind of found, you know, just the fact that you can get a PowerPoint presentation within 10 minutes.

I mean, I've never seen a tool be able to populate anything like that before. I mean. Honestly, like it's usually page by page, right? Ev different pieces of requests, different sort of, you know, nuances on each one. And you have to go back and forth, back and forth. But to see something be produced in 10 minutes is kind of shocking.

And so, you know, I kind of applaud you on, that's one of the reasons we invested. Actually, I'll spend some time on this 'cause you know, we've been thinking about AI a lot here at re and just for our listeners, I mean, I think I, you've heard me say some of this. The AI world is very, very similar to the internet world in the sense that it's really like a, a query and a response kind of model.

Like you ask the internet something, you get a bunch of feedback, and you ask AI something and you get something, some feedback, and some of it's generative. Some of it's just search oriented, and the reality is. It's kind of easy to, in, in the way AI is going, you know, every single company has, is trying to outdo one another, and the query-based architectures are super easy to replace.

You can effectively say, Hey, I'm using, you know, query one and I can just swap that out and use query two tomorrow. So when we invest in AI companies, we're looking for sort of four pieces of uniqueness, right? One is workflow integration. Um, 'cause as you know, and Samir said, workflow integration is very difficult to replace if you, if, if Far site becomes the tool.

By which you actually execute these transaction or generate PowerPoint. It's not like you're sitting there saying, ask question one, get an answer. I can replace something else to do that overnight if it's better and faster and more reliable. But generating like end product is tightly integrated and woven into workflow.

The second thing that we look at is really the systems level thinking. The fact that you are able to like. Architecture solutions such that an LLM that is better today. You know, if it becomes inferior tomorrow, you can swap that out immediately without any hiccup to the end users. The third is really sort of this whole notion of very, very, um, rapid time to value.

And what I mean by that is, you know, so you, you don't need to wait, as you said, Samir, like six months to get value outta this product. I mean, in a couple of weeks you're up and running and getting value. And the last piece that I look at is that the end user, the customers are the ones that are fine tuning.

How you, you get more value and better value over time, and the analysts, you know, are sitting there saying, well, you know, hey, you generated this PowerPoint for me, 85%, 95% correct, whatever it is. Every time they make a tweak, you know, the system learns from that and gets better. And so. That's how you build a company that has longevity and you know, you guys absolutely nailed it from a characteristic viewpoint.

And so, you know, we were all over it. Couldn't, couldn't not, I mean just had to do this deal. But it also happens that we are tethered in some weird way. So I'm gonna just pause for a second. Is there anything that else that you kind of want to talk about around the product or the company before I get into this?

You know, sort of Gatling gun section, which is my favorite part of the podcast where I kind of throw people off guard. So anything you wanna talk about Samir or Noah, that you know, kind of just would be valuable to our listeners. And it could be on any topic, it can be on, you know, sort of like how you've, how you're thinking about leveraging ai.

It can be part of the architecture, it could be why this is such an amazing place to work. It could be, which I know it is, and oh, or it could be something about the product or, you know, some features or nuances that you wanna talk about. Gimme a second. Uh, I'll, I'll have Noah think as well, but nothing jumps to mind.

Okay, well, so we can come back to that question. If you guys think about it. I'm gonna go ahead and move to the Gatling on section, which is where I ask you guys, you know, sort of random questions and you gave me fan stuff, so. This one, you know, you probably have a story around, it's not often that college buddies get together and fo found a company.

You know, it's usually coworkers. That's what we find, you know, happening in our industry. What, what made it come together with you, Kal, you know, Noah and Samir kind of deciding to do this? I can, I can start on that. So, uh, yeah, totally agree with you. And, and, and I've seen that anecdotally as well with the, the sort of company or colleagues starting the company.

And it's interesting, right, because I, I think in some ways we, we actually have an advantage, right? So, so we have sort of that shared experience from like almost a decade ago now at college, at, at, at MIT and then afterwards, like we, we all went and sort of did our own things for. Four years, uh, and, and, and learned like very different sort of parts of the world and, uh, specifically different parts of the world that we work with and sell into and partner with today.

Um, so in a way, I, I actually think it's, it's, it's hugely different. But you know, from, from our perspective, the, uh, the origin story was actually pretty, pretty simple where, um, I was living and breathing the problem. These guys were living and breathing, uh, tech that I. Thought might, uh, have a solution to some of the problems that we were facing.

And, uh, it just, it just made such a, I think, a seamless transition into starting the company. The other piece I'll say is like, I forget if it's Y Combinator or someone else, but you know, half of like early stage startup failures are due to co-founder split ups or co-founder, like disagreements. We have disagreements, but I, I, you know, there's just so much like trust and like shared experience there, though.

I, I think we're very well equipped to deal with those kind of things. Yeah, that's awesome. That's awesome. I, by the way, also started a company with some, uh, friends from college. Um, so here's the little known fact. Uh, Noah, Samir and Kal were all into single fraternity at MIT called five eight Epsilon. And I was also there and many, many years before them.

And I won't say how many years 'cause I try to tell people I'm young. Um, and maybe five years before you guys were there. No, no, no. That was a lot of years. A lot of years. And it, it is, it is a unique fraternity. I, I gotta be honest, like it's a, it's a national, it's no national chapter. It's a local thing.

And, you know, we have, you know, just like heritage and, you know, reunions where everybody gets together and, um, you know, William wants to jump in on this. Well, I just want our listeners at, at home to know that at this point, secret handshakes were exchanged, certain important robes and emblems were presented, and rituals were conducted, none of which was captured on audio.

Yeah, I can't talk about that stuff. Thank you. Thank you Will. Yeah, there was a lot of that. Um, but. It is a unique place and I think, you know, all of us are tethered to the university, uh, you know, I mean to the, to the fraternity in many ways, and it, it is a good bonding moment. So I'm gonna ask you, and if you need to think about it a little bit, I can start.

What is your funniest or best story for when you were living in the house? That is shareable to a public audience. I'm gonna start, I, there was two, I have two stories 'cause I had a chance to think about this. So one of 'em is, you know, so we're, we're tech dudes and five Beaton's got this beautiful roof deck and it's like on Memorial Drive on for four, 400 Memorial Drive and it overlooks, you know, the Charles River and this fraternity on the other side.

Of that river actually. And one point, you know, one of my fraternity brothers, I can't remember who brought home like a ton of surgical tubing and it was in the winter and we actually built this, like I, I, I think it was like, you know, ballistic sling and we were shooting snowballs, like off of our root deck.

Over the Charles, like, you know, just yanking this thing back and flinging 'em. It was unbelievable to see it. And like at one point we wanted to see how, how much strength was in this, you know, sort of surgical tubing and we shot it against, you know, the roof deck. I think it was like a chimney or something like that.

And it took out a brick. It was like unbelievable. So that's my first story. I got two. If you guys have had a chance to think of it, I'll let you go. I'm not admitting that I was part of this, nor that I know anyone that was part of this. But there's a concept generally at at, at MIT called Hacking. Yeah.

Which is these pranks that, you know, undergrads kind of play around campus in order to relieve stress and show off creativity. Uh, but much needed at a, at a school like that. So, um, one of the hacks, uh, that, you know, I was, I was privy to, was. There are these tunnels that run underneath MIT and at one point there would just a huge set of like unused industrial printers in there, right?

Like we started school in the, in the mid 2010s. Like nobody was really using these printers anymore. Uh, and so they'd move them to the tunnels below campus and these things are probably like a hundred, 150 pounds. That's crazy. And. So, so there, there was this hack where before graduation, one year, a number of those printers, those industrial printers wound up on, on, uh, lined up, uh, perfectly symmetrically on the dome, uh, that, you know, that famous dome on campus that the everyone takes photos in front of.

Yeah, yeah, yeah. And, you know, these are large printers, but from so far away at the top of that dome, it, it's kind of hard to see, but there's, there's an entire class of MIT graduates. Where the pictures behind them, you see these little specs on the dome and, and I'll, I'll tell you one other thing is 150 pounds, you know, those were taken, it's a workout.

2020 stories up, you know, going through some precarious. So I, I'll leave it at that. I'm, I'm not quite sure how that, so, yeah, I, I mean, I don't know what you're talking about, this hacking thing, but like I heard my year that there was a couple of people, I'm not gonna include myself or exclude myself in this.

That put a lit Christmas tree on that small dome, you know, and strung a lot of wiring and actually had to get a couple of repeaters and batteries and, and convert AC to DC in order to be able to do this. There's a lot of strings of light and a lot a late night doing that. So anyway, I appreciate that one.

I appreciate that. Noah, you got anything there? Yeah, so one of the biggest things that was happening in the house when I was a senior was poker. Hmm. So it became, it started out as this like fun little recreational thing. And then there was like leaderboards and like people trying to figure out who's the best one in poker.

And then it started bleeding out of the house. And so you got people from other schools coming into PBE to figure out who was the best poker player. Trying to play like MIT's best. We had kids that I've never seen before, like coming into play poker and it just became this like. Sensation of people trying to start games in the house with people that no one has ever met.

So it was like a constant like fist fight over who would have the team, who would've the table who can invite who over? Like, oh, I'm bringing over these guys from Harvard. Like no one show up, like I'm taking over this table. And since then, everyone in the house that played, we still play poker. Oh, that's so cute.

That's so awesome. So it's like a whole, it's a whole tradition and, and thing that, that, that started out since. I love that. So we became Poker Central. We, we became Poker central. Um, and I think it also was a very strong pipeline into everyone in the house by the time I left going into quant and hedge funds.

There's a very strong delineation at MIT of people playing poker and the pipeline of that to quant finance. Oh my God. That's hilarious. That's awesome. That's awesome. I'll tell you one more, uh, just 'cause I've, I've written it down. Um, so, so I don't, you guys did pledge retreats, right? You were in the Oh, no.

No. Okay. So you did our year when I was going to school. It was a long time ago. You moved into the fraternity like two weeks after you came for freshman year. You lived in the house for four all four years, you kind of made a decision very quickly. And so you had a pledge class that we had like 15 of us at the time, and you were also, you know, when you had to do your pledge retreat, you kind of went away.

It was like freshman year and um, you did something sort of like diabolic to the house. It was like a, like a unwritten rule. And so we stole all of the toilet paper. And the toilet seats. We took all the seats with us. There was no seats available. And you know, we were like, okay, this is, you know, we're definitely gonna be like, known as like, you know, the, the, the class that kind of outdid everything.

And we come back and obviously there's all new seats on there and we go back to our rooms and of course, the. Older classmen have hacked the crap out of us. And, uh, I'm like finding nothing. People are like nervous, like eggs are gonna fall from their, you know, near their desk. When they open a drawer, you know, there's gonna be some kind of like gas.

One dude did have like a, he opened his door and there was an entire doorway full of like full Pepsi cans. One layer deep in front of his door. And so what they had to do is they had to layer them all out and then escape through the window and repel down the house. Um, and they just left at there. I go into my room and I'm like, okay.

They, I, I escaped. I don't know what the hell happened. And I go into my drawers and in place of where my like socks were was a drawer full of goldfish. They like every single drawer had like goldfish in it. And I'm like, I don't, I don't know how to get this out. I don't even know what to do. Um, oh my god.

The snack goldfish or the animal goldfish? No, no real goldfish living, you know, like little tiny. Yeah, yeah, yeah. Like there was, there was water in your drawer and like living goldfish? Yes. Every one of my drawers. Yeah. Living goldfish. Wow. Yeah. That's really impressive. Okay. Yeah. But I have to ask. What happened to the goldfish?

Okay. The goldfish got donated to a, uh, at a, to a store, a pet shop. It took me a long damn time to get the goldfish out to, like, they'd layered it. They didn't wanna damage the drawer 'cause that was like house property. So it was pretty fancy. Okay. Wow. Alright. Uh, will you got something, you got something from your like, uh, fraternity days.

Can't leave you out of this man, dude, all, all those nights when like, you know, you weren't there and you're, you were, you were not associated with whatever happened. I, I had my share of those two, but I went to bed early that night. Exactly. So, uh, tell me the story then. Yeah. Yeah. Well. Not for, not for this, uh, setting.

You guys clearly had an awesome time and, and, uh, yeah, yeah, that's fine. That's fine. Okay. It, it's a great tradition. And by the way, I, I mean, I think one great takeaway for all everyone in our audience about that is that you, you meet these people in college, in, at a place like in MIT or Stanford or wherever you go to school, and some of them are gonna come together.

And change the world and you can be a part of those things that it happens. A, a lot of companies are founded on shared experiences from a really young age, and those are some of the most powerful bonds we ever see among founders, particularly 'cause of the values and the loyalty to each other and to a shared vision.

And so I would, I, I think you guys are a great instance of that, and thanks for sharing some of your history before you were company founders. Yeah. Yeah. I got one last question. Go ahead. One last question and then, you know, I'm gonna let you wrap it up, will, but, okay. Favorite movie referencing? MIT Uh, I I, I got one for this.

So I went to go see, I think it was Infinity War or maybe it was Endgame when Tony Stark was talking about Eigen values and he like cracked time travel from Eigen values. Yes, it was end game and. I had friends that were mathematicians. I had like, like all in this row, like computer science kids that just happened to take that class.

Like we all looked at each other and we were like, this is the most like facetious thing they could have said. It was the, we were like peeled over, like laughing at the fact that they were saying that like, Egen value's cracked time travel. And everyone in the, like the movie theater like turned around and looked at us and like, obviously no one had any idea.

We were laughing at a math joke, but that is Tuesday. Like something that I remember like as being like, so hilariously ridiculous. Yeah. That's H Samir. You got one? Yeah. Uh, first of all, no, I'm surprised given the. The poker story you just told you didn't go with 21, but Fair enough. And the, the math problem that they do in there, and it's supposed to be some grad level class is actually taught in 6 0 4 2, which is like the second computer science class you take.

So I guess I'm taking two off the table here. So 21, um, surprised no one didn't say that. And personally, I'll just go with the classic, the Goodwill hunting, I think is a Yeah, yeah, yeah. I'll give you guys one more. I'll give you one more. A movie called Real Genius with Val Kilmer in it. Uh, you probably have never heard of it 'cause like that was back in the day and Val Kilmer was like this, you know, hot, good looking guy and everything like that.

But all based in MIT. So, um, I get the, I get the benefit you guys watch severance. Yeah. Yeah. So I, I worked in that building for like five years. The one in New In New Jersey? Yeah. At Bell Labs. Yeah. No, I was like, I'm, I'm like full of movie trivia, you know. So one, one of our employees just bought the severance soundtrack and said, once we get a recorder, uh, record player in the office, she was gonna bring it for her.

Any. Nice. That's too funny. I was gonna ask you about your favorite hallucination that you've seen, but I'm gonna pass on it. We've, we've gone almost 50 minutes just plus here, so, um, anyway, I'll let you wrap this up. Will, if that's okay. Well, Noah and Samir, thank you so much. We're thrilled to be investors in ForSight and I think you've shared some incredible insights with our audience today.

People are gonna be. Waiting for the day that their organization brings far sight in to help them get more done. Uh, we at re are are thrilled with what you're doing. So listeners, thank you for being with us for another episode of re POV and we look forward to coming at you again soon. Take care. Thank you for listening to R-E-P-O-V.

You can keep up with the latest on the podcast at rre on X or rre.com and on Apple Podcasts, Spotify, Google Podcasts, or wherever fine podcasts are distributed. We'll see you next time.