---
title: "Zero to Umm — Jared White: Building AI You Can Trust in a Courtroom"
url: https://stacklist.com/card/428bab9e-7283-41da-b9b8-bbfeb64eb7ef
stack: https://stacklist.com/c/podcast/stack/5bb1c307-cd3d-4b7e-a200-d2f087a72449
summary: "Jared White, founder of Matey AI, discusses building trustworthy AI for criminal defense by teaching computers to read rather than write. The interview covers his path from quantitative trading to legal AI, the economics of modern AI development, and why starting with customer needs is critical for founders."
tags: "ai-legal, criminal-defense, founder-interview, trustworthy-ai, legal-tech, startup-strategy"
key_entities: "Jared White (person), Kyle Hudson (person), Matey AI (organization), CrimD (technology), Cursor (technology), human-in-the-loop (concept), trustworthy-ai (concept), Austin (location), Zero to Umm (event)"
classification: "transcript"
content_hash: "sha256:fae0f5aa9d18c807d24123529dfb0d736d08ca2b4e4bd506445c5c998daa3c1f"
acp_version: "0.2"
token_counts_approximate: 8346
visibility: public
agent_accessible: true
status: "final"
---

# Zero to Umm — Jared White: Building AI You Can Trust in a Courtroom

# Zero to Umm — Jared White: Building AI You Can Trust in a Courtroom

**Show:** Zero to Umm (interview format)
**Guest:** Jared White — Founder & CEO, Matey AI
**Host:** Kyle Hudson
**Voice mode:** stacklist_brand
**Audience:** Business / founders & builders
**Episode length:** ~53 min

---

## Description

We taught computers to do math, then to write. Jared White thinks the real unlock is teaching them to *read* — and he's proving it in the one place that punishes a wrong answer hardest: the courtroom.

## Episode summary

Kyle sits down with Jared White, founder and CEO of Matey AI, whose path runs from building quantitative trading desks in Austin, through a stint in crypto, to an AI-native legal company built for criminal defense. Jared explains why he started in the least forgiving domain he could find, how the economics of building have flipped — he reads a Cursor bill as an asset, not an expense — and why the shift that matters isn't AI that writes, but AI that reads and reasons over data no human has time to get through. It lands on the advice he'd give a younger founder: start with the customer, because implementation is the easy part now. Builder to builder, a candid look at building something you need people to actually trust.

## Key topics

- From quantitative trading desks to legal AI — a fifteen-year path through software, markets, and crypto
- Why teaching computers to *read* — not write — is the unlock Jared bet the company on
- Why Matey started in criminal defense: the least forgiving place you can put an AI
- The new economics of building — a Cursor bill as an asset, three engineers out-shipping eight
- Pricing as a craft (version eighteen and counting) and the land-and-expand math of small, fast contracts
- Human-in-the-loop, and shrinking it toward zero as the models get better
- The advice Jared would give his younger self: start with the customer, not the technology

## Timestamps

[00:00:00] Intro — coffee mugs, merch, and a Matey logo story
[00:02:40] How Kyle and Jared connected, and founder-led sales
[00:05:35] Pricing as a moving target (version 18 and counting)
[00:07:33] Reading a Cursor bill as an asset, not an expense
[00:10:26] Jared's path: software, burnout, and building a trading desk
[00:15:13] The crypto chapter — and what FTX and SVB ended
[00:17:25] Naming Matey: the Wordle of the day and "your first mate"
[00:20:22] Why this feels like the early web all over again
[00:21:42] The real unlock: teaching computers to read
[00:24:34] When AI coding finally got good enough to trust
[00:28:17] Why start in legal — the least forgiving problem
[00:30:55] "Executive Mate": an AI that could run a company
[00:31:49] Scaling the team, hiring misfires, and the data problem
[00:37:37] Human-in-the-loop, shrinking toward zero
[00:40:38] The raise, and a go-to-market year ahead
[00:44:59] The trajectory beyond the next round
[00:50:19] Advice to a younger founder: start with the customer
[00:52:36] Wrap-up

## Highlights / key quotes

*(Lightly cleaned from the transcript for readability — confirm against audio before pulling as public quote cards.)*

- "We've been teaching computers to do math for a really long time. We've finally taught them to read." — Jared White [~21:42]
- "If we can trust it in the courtroom, we can trust it through the rest of our lives." — Jared White [~29:25]
- "We're building and selling trust at scale with AI." — Jared White [~31:00]
- "Start with the customer first. If you don't have a customer, then don't do it." — Jared White [~52:00]

## Guest bio

Jared White is the founder and CEO of Matey AI, an Austin-based company building AI-native tools for criminal defense. Before Matey, he spent roughly fifteen years across software and quantitative finance — building and running trading desks — followed by a stint applying machine learning to asset valuation in crypto. He started Matey in 2023; its flagship product, CrimD, helps defense teams turn massive volumes of case discovery into cited, trial-ready insight.

More at matey.ai

*About the show: Zero to Umm goes inside the messy middle of building a company — the real chapters, not the polished version. Hosted by Kyle Hudson.*

## CTA

New episodes of Zero to Umm feature founders in the thick of building — the real chapters, not the highlight reel. Follow so you don't miss the next one. Curious what Jared's building? Take a look at https://www.matey.ai/ 

---

## Full transcript

**Kyle** [00:32]
Hello, hello. How you doing?

**Jared** [00:37]
I'm doing great. Had a wonderful morning so far, got my coffee going. So here we are — the virtual experience.

**Kyle** [00:45]
Here we go. Cheers. You're repping the GitHub mug.

**Jared** [00:54]
I am. This one's vintage — like 2012 or something. I like collecting coffee cups from all the technology that comes into my life. I've got a rather elaborate collection from various businesses — some that still exist, some that don't.

**Kyle** [01:16]
Funny, this one's vintage too — almost two years old now. I can't give it away. But from a merch perspective, I've been thinking about our strategy. I've got stuff upstairs — Finn from Intercom, all these things from events. What I want to do isn't volume, it's quality. I've got a couple of pieces where the hoodie is so soft, so nice, that you actually want to wear it.

**Jared** [01:51]
Yes. Funny story about that. The logo for Matey is meant to be sails — we're actually shifting that to waves, so we'll have a new logo in a few months. But it's meant to carry a bit of numerology, almost like the way the Mayans did numbers. You could do stars, or the moon on one side, the sun with some stars — and build a notion around it, like "this is Q4." Embroider that on a hoodie, and that's what you give away at your event or offsite. Someone sees it and goes, "That's Q2 — that's a keeper."

**Kyle** [02:26]
I love that. Well, I'm excited to talk today — thanks for taking the time. Our connection through Darwinian [VERIFY: Darwinian Ventures?] is interesting — through Andrew and Emma [VERIFY NAMES]. I met Andrew through a mutual friend; we went through Blue Startups Accelerator together in Hawaii. Then Andrew and I got thick as thieves. We hit a point, from a sales perspective, where I'd been founder-led, white-knuckling it to a certain MRR — and then you go, I could keep pushing in this direction, but if I don't want ulcers and I want to keep my sanity... They came in, stood up our team, and now we're cranking. When did you all meet?

**Jared** [03:42]
September of last year — similar situation. I was at a CEO dinner put on by Alumni Ventures, who are on our cap table and invested in our seed round. They said, "What does everyone need? Let's help each other out." I said, we've stood up the sales team, we've got some things going — I'd love feedback on whether I'm doing the right things, if there are better ways, am I missing the mark, am I doing better or worse than I think. Turns out Andrew was sitting right next to me. I'd met him one evening earlier and didn't really know what he did. They said, "Then you need to talk to Andrew." So we talked, it snowballed, and we've been working together since September.

**Kyle** [04:48]
There are these chapters. This whole idea of "zero to." When I started Stacklist, I loved the book *Zero to One*, but it's one of those where you sit around thinking, what is "one" for me right now? A customer? The first major round close? I felt like this was a chapter where I got right to that point — I could almost see money spilling off the table because I couldn't keep up. Do I get someone more junior and train them up? It's been a good process. But I'm glad we got connected through them.

**Jared** [05:35]
Andrew's been extraordinarily helpful — developing our pricing strategy, coaching the sales and SDR team. Because at the end of the day we're essentially enterprise sales, even though some of our customers are a single-person law shop. It still feels like an enterprise sale to them, because we might be about the cost of a rent payment. That's a big deal. If you're telling someone one of their largest expenses is going to be this new thing — but you'll make more money because you'll be better and faster — that's a hard decision. So how you structure the pricing really matters. We started with "every time you want to use this, you pay a certain amount," and that's a recipe for decision lethargy — then you don't hear from them for eight months and wonder if they're still a customer. So, developing the right intuition around pricing — I think we're on version twelve or thirteen.

**Kyle** [06:55]
Pricing is such a learning thing when you start. You think you've got it solid, and it's like, okay, this is version one of eighty-three.

**Kyle** [07:33]
We're in such an interesting time with AI coming in. People talk about their Claude bill being a thousand a month, and from a SaaS perspective that seems high, because we're used to paying for Intercom and PostHog. But then you go — wait, what would a junior engineer cost? A fractional controller? Legal review? Suddenly a thousand a month is pretty good.

**Jared** [08:09]
Pretty good deal. We have a ten-thousand-dollar-a-month Cursor bill sometimes — but we have three engineers instead of eight, and we generate more software with three engineers and a ten-thousand-dollar Cursor bill, which isn't even a full-time engineer's salary. And I'm one of the three, so I have other things I do too. With that, we now generate more software on a weekly basis than we did with seven or eight engineers a year ago on a monthly basis. So I associate that ten thousand dollars of spend with a hundred and fifty thousand dollars of software engineering. Versus — I won't name names — our database provider. That bill hits and I go, man, I have to do something about this.

**Kyle** [09:07]
Totally. I'll give you a good one: Slack. I was at a large design firm, part of a huge consultancy, and we got to a point where the Slack bill for the whole company was like a million a year. Those are the things where you go, I've got to get rid of that. But now it's so different to jump to a piece of software that costs that much, because we haven't gotten to the place where we correlate it to the operational and labor cost. Anyway — I want to dive into Matey and your background. Before we get to the day you buy the domain name, tell me where you came from. Were there things in your upbringing or family that contributed to this path?

**Jared** [10:26]
Great question. I always wanted to be an entrepreneur as a kid — much more interested in working for myself than for anyone else. But I had about a fifteen-year career before I started anything of consequence of my own. The very first software I wrote was in the early '90s — I was part of a group that did competitions, and I sold them software on floppy discs. I still have six floppies from that. It was Visual Basic. Then I got a software job. About ten or eleven years later I was super burned out. A bunch of my friends had moved to Austin to work for a prop trading company, making two, three, four times what I was making, getting to work at eight-fifteen and leaving at three-thirty. No working late. I thought, this is a way better life. So I moved to Austin — if these guys can trade, I can learn to trade.

**Jared** [12:19]
A few years go by. I traded through the financial crisis and did pretty well — basic day trading. Then I told myself: I know how to write software and I know how to trade, so I should write software to trade, or at least help me trade. So around 2006, 2007, I'm building the basic underpinnings of quantitative trading and a lot of the infrastructure around it — at a time when not many people were, outside of Renaissance, Jane Street, a few others. In an office in Austin, a few of us started playing with it and ultimately built an entire trading desk. I ran that through 2015. Then I was going through some personal stuff and wanted a place to check out, so I shut it down and took a job at a quantitative firm in Austin. I looked around and thought, this place is going to go under — they make a million a year less than their expenses; where's my bonus coming from? So after sixteen or eighteen months, I was out. I wanted a remote job — I didn't want a desk in the same place every day; I wanted to travel. So I took a remote software job for a couple of years. Then I got a call from folks I'd met: "You used to run a trading desk — why don't we do this together?"

**Jared** [14:18]
That led to my second trading desk, which I ran through 2022. But my heart really wasn't in it the second time. I had much better intuition, but the market had changed so much — things we'd had to learn from scratch, NASDAQ was now advertising to you. It was no longer a chase or a thrill. So in 2022 I took a call and went to work for a startup as director of machine learning — a quantitative crypto thing. Not trading; selling an analysis of NFTs. What came out of that was interesting: we did a valuation strategy and solved the problem of valuing illiquid assets in a really novel way. I thought that was great. But then FTX happened in November '22, Silicon Valley Bank in early '23, and crypto was no longer going to be a thing.

**Kyle** [15:44]
Did you go down the NFT rabbit hole?

**Jared** [15:47]
I really didn't. To be totally honest, I couldn't believe it was happening. I thought it'd be a really cool use of the technology to do fun things.

**Kyle** [15:59]
I think it will be. I think it was the underpinning — the testing of the rails. Ethereum's throughput now is an unbelievable amount of money. I can see a future of immutable blockchain assets — your house, a contract. But I was in the waking-up-at-2 a.m. to see if NBA Top Shot was dropping LeBron packs phase.

**Jared** [16:29]
I understand why it happened — we were all stuck at home during COVID, people had time on their hands, and it became almost like gambling. What disappointed me from a technology perspective was that the hash wasn't a hash of the image. I really feel it should be a hash of whatever's underneath it. But they'll solve that — then the title company can go away, you can do that on the blockchain, automobiles, whatever. Lots I'm excited to see. Anyway, that company closes down — it was funded through an investment vehicle of Eric Schmidt from Google, and they stopped funding their crypto incubation. We were burning so much money. That got shut down around March of '23. So a couple of us — one I'd brought from my trading desk — started Matey in March of '23.

**Jared** [17:25]
The advent of the name: it was the Wordle of the day. We were looking for something other than "copilot," because everyone was using Copilot — GitHub Copilot, all that. It felt overused. I wanted something that described what we were building, which was your first mate. Your friend. Something like R2-D2 and C-3PO, but in your pocket.

**Kyle** [18:37]
Love it. From a branding perspective, it's got more than "copilot" — it's collaborator, partner. There's trust, a relationship, an understanding that comes with it.

**Jared** [19:00]
Exactly — it's a trusted person. Like you'd put your first mate in charge of the ship and go ashore. The next in command. Eventually this mate comes to you and says, "I've been observing what's going on. Legal Mate is our first mate — there are other mates we envision." When you wake up in the morning, it knows all the Slack messages that came in while you slept, all the emails, the ones that did and didn't come in — because not having an event is an event too. "These are the three things I typically see you do; I've done the first draft, discard the two you don't want. Here's your optionality for the day." You get to orchestrate your life instead of typing it all out.

**Kyle** [20:22]
I love this period we're in. You and I have gone through these before. Late '90s — my first computer was an 8086 luggable Compaq that weighed sixty pounds, with two 5.25 floppies. I bought my first domain name, zipzap.com. I sent my first email and ran downstairs: "Mom, you'll never believe it — you can send a message to someone in Japan and they read it, for free." Then websites came around, businesses realized they had to have one — that transformational period. I feel like we're back there. It was Q4 of '24 when ChatGPT blew everyone's mind and got into the zeitgeist — but you all were, what, eight months before that? As you started, what were the magic pieces when you saw not just automations and workflows, but the ability to actually partner with something like Matey?

**Jared** [21:42]
The big unlock for me: we've been teaching computers to do math for a really long time — since the Babbage machine, the mechanical way we taught math. And now we've finally taught computers to read and write. ChatGPT was this notion of writing — I could ask it to write something. But I found it much more intriguing that it could read and then write. You give it a corpus of data and say, "What about this?" To me that was the real unlock — because it's not just about everyone prompting models to generate new content that who has time to read. It's about consuming everything that we, as a society, don't have time to consume.

When we were thinking about what to build, this was when Silicon Valley Bank was going under. I had my eyeballs on the news all weekend, seeing conflicting reports, and I thought: it'd be so great if I had a ChatGPT that could read all of this, orchestrate the information, and give me some degree of confidence. That was the original user experience — I drew it on my daughter's art pad that weekend. News was one data source, so "NewsMate" was envisioned but never built. But I'd talked to a lawyer months before who told me about his experience with ChatGPT: "You've got to build me some things." So legal found us more than we found it. We could tackle the whole discovery problem — combat the information warfare that happens in these cases, increase access to justice.

**Jared** [24:34]
Then, separately — I was late to AI coding. It's not that I didn't use AI to code; I'd put a question into ChatGPT's best models, get a function, tweak it, put it in. But December of last year was the unlock. I think Opus 4.5 or 4.6 had come out, and Cursor was finally good enough that I could trust it and not have to redo the work afterward. That's when the whole reasoning thing took shape. We'd been talking about using these models for true understanding, reasoning, planning, orchestration for a while — and this was finally good enough to really use. At least as good as a software engineer I would hire.

**Kyle** [25:17]
Four-six was that moment for me too. Remember the period between ChatGPT entering everyone's consciousness and then 4.6 — prompt engineering, structuring "who are you," all of that. Then 4.6 was like, "I need help with some stuff," and it went, "Cool, I got this." Ever since, it's been faster and faster. Where did you all really come together and say, this is it, here's the domain?

**Jared** [26:06]
March or April of '23. Got the domain, the Slack, the G Suite. I knew I wasn't going to bootstrap — I wanted to raise. So we spent that summer building product and trying to fundraise. Terrible time for a pre-seed on a pre-revenue company.

**Kyle** [26:36]
And AI wasn't in people's minds the way it was the next year — even more difficult.

**Jared** [26:44]
Right. At the tail end of summer '23, I told the guys, "I don't want to do all these fundraising pitches. Let's build product until we have something better to show." So we focused. I brought in another fellow — my brother, actually — who built out the front end and the user experience. He's the reason the application is beautiful; the rest of us couldn't do that. Things started to click, we had a few customers, and we got accepted to the Neo accelerator. That's when everything took off. Through 2024 we signed a deal with the Colorado state government — used for criminal defense on their alternate defense counsel, a secondary public defender system. Then things really started to fly. All along there were three or four legacy players that mattered, and we were building AI-native from the ground up — not a vector database bolted on top of a pre-existing database engine.

**Jared** [28:38]
Fundamentally, the reason I was drawn to legal first: it has the least room for error on hallucinations. It's the most rigorous use case. In the future I might not make that choice — I'd rather build the easy problem and get quick wins. I tell people there are a hundred and fifty things you have to learn to build a billion-dollar company. I haven't built one, so caveat — there may be more, and I may think I know some that I don't. But a hundred and twenty-five of them have to be learned firsthand. You can't read it in a book.

**Kyle** [29:17]
It's not a workshop. It's not a course you download. You've got to feel the pain of it and go, "That's not it."

**Jared** [29:25]
Exactly. We started with the hardest problem: how do you build a system that can access — at the time — gigabytes of information, when context windows were four thousand tokens? How do you access a gigabyte of data without hallucinating? We built that. Now it's a petabyte of data, and it's so much richer in how it reasons through finding answers. Throughout, we're building this AI-native version of what the legacy industry made over the years. Those old tools rely on techniques like TAR [VERIFY: technology-assisted review] — and honestly, the way I describe it, those things exist because language models didn't. If we'd had language models when those were built, we'd never have developed them. So we fundamentally know they need to go away. Let's build a native system that can be used in the courtroom to address a very real problem. And if we can trust it in the courtroom, we can trust it through the rest of our lives.

**Jared** [30:55]
Ultimately, "Executive Mate" is where I'd like this to be — an AI that can run a company, starting with the legal department and the accounting department. Plenty of people are doing the sales stuff. If you can control those three things, you can run an entire company with a system like this. But you can't do that if you can't trust it. That's why we say we're building and selling trust at scale with AI.

**Kyle** [31:27]
That's amazing. What was the progression after the accelerator, when things started to click — from the small team to scaling up?

**Jared** [31:49]
We did a sizable seed round in November of '24. I'd always believed we'd need a certain amount to get to Series A, and we raised a substantial portion of that. So we're about to do another fundraise to get the rest. I wanted to stay as lean as possible — create intensity, urgency, pressure — but not so hard that it's four people in a room. After that round we started developing the team. It was funny: I was trying to build the sales team and gave offer letters to two different people to be VP of Sales. Everyone's super excited — "best thing ever, I've been selling in this industry forever, wish I'd had this twenty years ago" — and then, "by the way, I have a non-compete." You could've mentioned that. Then, "you should talk to my protégé." Same thing. Goodness gracious.

**Jared** [33:55]
Eventually I hired someone else to do sales, and that didn't work out either. At the same time, we were building the engineering team, and I made some substantial miscalculations there. Our problem space existed in one area, but really existed in another. What we had to solve first was how to ingest all this unstructured data. I thought we were solving the AI agent problem, so we built this amazing agent — but then you have to feed it context, and getting that out of body-worn camera footage, surveillance footage, recordings of phone calls becomes a petabyte-scale problem very quickly. That consumed a substantial amount of engineering effort just to make the system work at scale reliably — because it has to be lights-off. If you have to touch every case, it's not scalable. And all the while we're selling customers, selling customers, selling customers.

**Kyle** [37:21]
Did you go through periods of human-in-the-middle — eighty/twenty, then changing the percentages? What areas, and how did you navigate that?

**Jared** [37:40]
It was a natural progression as the models got better. As they got better at orchestrating, we could take more out of the loop. Initially the human does the first thirty percent and the last thirty — you're doing forty in the middle — and then that expands in both directions. On the front end, the human primes it: "This is what I'm looking for," a prompt of a certain sophistication. On the back end: "What do I want to do with this data?" As the models got better, we could take a much simpler, more ambiguous prompt and bootstrap it into something the underlying agent can resolve — maybe asking a clarifying question. And at the end it might say, "I've observed you typically want a Word document now — want me to save it that way?" So the human's now doing the first fifteen percent, the last fifteen. Eventually you want that as close to zero as possible: the data comes in, we have an idea of what you'll prompt, so we do two or three of those for you. Instead of asking "Word or PDF?" we just make both, and you throw one away. But all of that surfs the wave of what the frontier model orchestration engines can do. That's why new models get exciting.

**Kyle** [39:58]
I'm so disappointed Fable went away. I got three, four days of Fable and then — nope, can't have that anymore. Anyway — you're prepping for a fundraise now, right? Toward Q4? The lead-up is heads-down, living in a hotel room, focused on the raise. What's the fundraise focused on, and what does next year's sprint look like?

**Jared** [40:38]
It's almost a hundred percent go-to-market. I feel very comfortable with where the product is. Sure, I'd like an SRE, another engineer to really make use of the infra we've hardened over the past couple of years — but it's essentially go-to-market. Leading up to the fundraise, everyone's buying alpha, and the proxy is revenue — and the curve on it, how quickly you got there. We really just started official go-to-market. Even though we signed that first Colorado state contract in September '24, it wasn't a ten-million-a-year contract — it was "we'll use you on the cases we want to use you on." That's a win. Now we've transitioned to packages. The pricing discussions — version thirteen, probably really version eighteen, because we didn't count the first several. It's about tuning that into the civil space. We understand how to sell to criminal defense, but the civil market is where twenty and fifty times as much money exists, with legacy players everyone hates using but has to. We've got eight or ten logos in that arena. We go in and sell a tiny contract, because the land-and-expand is huge — I don't want a big contract that takes six months to sign. I want a three-case compartment they'll sign in two weeks, so we get them in the system, collecting revenue, getting feedback. Then they're at a fifty-case compartment at fifty grand a month before you know it. If we can acquire two to four logos a month over the next six months, that'd be great.

**Jared** [42:57]
By the end of the year, things start to click — you see hockey sticks on individual contracts as word spreads virally within organizations. Then, once we do the next round, I want to quadruple the go-to-market team, maybe add marketing, maybe get some paid digital back in place now that we can service a different segment. That's what at least fifty, maybe eighty percent of the use of funds will be.

**Kyle** [43:33]
And this round — you're raising an A?

**Jared** [43:36]
I wouldn't call it an A. I'd call it the remainder of our seed. When we started, I figured I might need about X. We raised about sixty-five percent of that. So — seed two? I don't know what we call these things anymore.

**Kyle** [43:50]
Rounds are so squishy these days. We did an angel, friends and family, some strategic investors — and people ask, "Are you doing a pre-seed? A seed?" You've got pre-seeds raising ten million and seeds all over the place.

**Jared** [44:16]
Right. People call five million, fifteen million on however much revenue a Series A — but I've seen C rounds at fifteen million. It really just depends.

**Kyle** [44:27]
Totally. So — historically people would ask, "Where do you see yourself and the company in five or ten years?" Now, if I said five or ten years, it'd be jets and spaceships, with how fast everything's going. But what would you see for yourself just past this next round — once you've lily-padded past it, where does that trajectory take you?

**Jared** [44:59]
It's product-focused. Right now we're collecting the data — when you're in litigation, you use Matey because you've got discovery data, or you're producing it to opposing counsel. We're the data layer of all that. The next generation is what we call First Mate — your trusted companion that knows exactly what you want beforehand, and the optionality of what you might want, and just does those things. I hate to use phrases like this, but the cursor of discovery. The fundamental architecture we're building now is in service of being able to do that in the future. You can't just scale up a Mongo database forever — it has to be done in a fundamentally different way. So as we ramp the go-to-market engine, you'll see a lot happening under the covers, because it's all in service of essentially every lawyer in America putting everything they do inside Matey.

There's a vision beyond that I don't like to publicize just yet. But let me say it this way: what I want to build essentially creates a market that doesn't currently exist — and it doesn't exist as a function of the ethical constraints of the legal industry. You do an end run around that to provide the net asset value of law to law firms. It's a really exciting trajectory. I foresee starting to go to market with that in about eighteen months — post-Series A for us. The vision: much like a ten- or twelve-thousand-dollar Cursor bill that I associate with positive value and an asset on my balance sheet, that's what Matey becomes for a law firm.

**Kyle** [48:32]
Amazing. It's funny — having a moonshot, the long-term vision. Think of Google starting out thinking, "This is going to be the index of the internet" — such an extreme future vision. What's interesting now is you can start to connect all the dots and get so much feedback so early that the vision becomes clearer — it used to be year over year, now it feels like month over month.

**Jared** [49:20]
Exactly. And it's nice that companies like Google and Meta give you the clues about what it takes to go from point A to point B — the fundamental ingredients. Of course a lot changes with AI versus social media or indexing the internet, but the fundamentals of what you're trying to capture, and what you use to lead product development, are the same. It's a fascinating time.

**Kyle** [50:03]
I love it — I'm like a kid in a candy store. It's back to my CSS3 and HTML book, writing the first website, putting it live: "Oh my gosh, this is amazing." We're back in that time. So — if you went back to the early entrepreneurial version of yourself, would you do anything differently? What advice would you give people who are in that place right now, thinking about setting out on their own?

**Jared** [50:50]
A few things. Number one, if I were talking to my fifteen-year-old self: keep going, and solve the easy problem first, not the hard problem first. And a very interesting thing I was taught through the Neo program — this was a massive unlock. Think about Google: you start with really awesome technology and ask, "What can we use it for?" You give it away for free because venture will fund it, you collect eyeballs, some subset works, you find a monetization scheme, you get money, you capture the market. You go from "here's an implementation of a thing," to "what problem is it the solution to," to "who will pay for it."

Today it goes in the opposite direction. Start with the customer first — who's paying? What problem do they have that they're willing to spend money to solve? What's a solution they'll pay for that provides excess value to them and to you? Then go implement it — and implementation is now trivial because of things like Cursor and Codex. So start in the opposite direction. Start with the customer. If you don't have a customer, don't do it. Things will be a lot easier.

**Kyle** [52:36]
Amazing. This has been so much fun — I appreciate you coming on. My favorite thing is hearing the story: the spark of inspiration, the team coming together, everyone jumping into Slack going, "What do we do next?" Especially where you're at, finishing up the seed, thinking about the A. It's so exciting, and I appreciate you sharing all of it.

**Jared** [53:10]
Thank you so much for having me. It's been a pleasure — wonderful meeting you. We'd love to chat again sometime.

**Kyle** [53:16]
Amazing. Let's hang out and catch up for a bit.