Dec 5, 2025 · 44m · no-priors
No Priors Ep. 142 | With Harvey Co-Founder and President Gabe Pereyra
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of No Priors, Sarah Guo and Elad Gil speak with Harvey Co-Founder and President Gabe Pereyra about scaling legal AI infrastructure, applying reinforcement learning to complex legal reasoning, and transitioning from individual copilots to enterprise-wide organizational intelligence.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 28.1% of the talking time here. How this is scored →
The hosts as informed peer 5.5 Guest teaching 4.3 Guest disagreement 0.5 The hosts pushing back 0.6
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 39:21 Gabe refutes the air mattress rumor
Gabe humorously but firmly rejects Sarah's premise that he sleeps on an air mattress, detailing the logistical saga of his missing bed frame.
Hardest push from the hosts ▶ 36:00 Elad challenges Sarah's premise on founder ambitionElad counters Sarah's claim that early coding AI founders were less ambitious, arguing they were aiming for giant models while Harvey succeeded by nailing product.
Biggest teaching moment ▶ 18:20 Gabe redefines verifiability in production software versus legal outcomesGabe educates the hosts on how real-world distributed systems engineering mirrors complex legal transactions because true verification only occurs years after deployment.
The host holds their own ▶ 22:28 Elad outlines historical enterprise software implementation playbooksElad demonstrates seasoned enterprise knowledge by linking Harvey's forward-deployed engineering strategy to the playbooks of Oracle, Dell, and IBM.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Expanding from Law Firms to In-House Corporate Teams | 5 | 4 | 0 | 0 | Sarah and Elad prompt Gabe to explain legal workflows beyond consumer perceptions. Gabe gives an in-depth breakdown of private equity fund formations and side letters, comparing legal contracts to complex codebases. | |
| Agentic Reasoning and Reinforcement Learning in Legal Matters | 6 | 5 | 0 | 0 | Elad outlines agentic logic trees in software engineering and asks how legal translates. Gabe draws upon his DeepMind RL background to reframe legal matters and associate workflows as RL environments. | |
| Transforming Law Firm Structure and Training Next-Gen Partners | 7 | 4 | 1 | 1 | Elad shares insights from his Series B diligence calls regarding partner leverage ratios and firm restructuring. Gabe explains how law firms can leverage partner feedback data to accelerate training for associates. | |
| Partner Expertise as Expert Reasoning Traces | 6 | 5 | 0 | 0 | Sarah connects senior partner expertise to distinguished systems engineers. Gabe details how transactional partner Gordon Moody's tacit architectural knowledge represents expert reasoning traces missing from public datasets. | |
| Verifiability and Reward Functions in Legal Reinforcement Learning | 5 | 6 | 1 | 0 | Sarah questions how RL scaling applies to law given the lack of objective verifiability. Gabe explains legal reward functions and notes that real-world software engineering also lacks simple unit test verifiability over multi-year horizons. | |
| Forward Deployed Engineering and Enterprise Implementation Playbook | 7 | 4 | 1 | 2 | Sarah questions why a software platform is establishing a Forward Deployed Engineering unit. Elad steps in to contextualize FDE as the classic enterprise software deployment playbook seen across Oracle, Dell, and IBM. | |
| The Unexpected Speed of Enterprise Legal Adoption | 5 | 6 | 1 | 1 | Sarah asks why Harvey does not build its own law firm, citing community interest. Gabe explains lessons learned from Atrium and articulates why tech execution and legal practice are incompatible under one roof due to conflicts of interest. | |
| The Multi-Trillion-Dollar Professional Services Platform Opportunity | 6 | 4 | 0 | 0 | Sarah illustrates cross-border counsel complexities using global enterprise M&A examples. Gabe expands the market opportunity beyond legal into the multi-trillion-dollar professional services collaboration space. | |
| Transitioning from AI Researcher to Hypergrowth Founder | 5 | 3 | 0 | 0 | Elad asks about Gabe's transition from IC researcher to founder. Gabe shares the founding story with Winston and how early exposure to GPT-4 gave them high conviction before generative AI became mainstream. | |
| Comparing AI Form Factors in Legal Tech versus Coding Tools | 6 | 5 | 1 | 2 | Sarah and Elad debate why coding tools emerged later than Harvey. Elad pushes back on Sarah's suggestion that coding founders lacked ambition, and Gabe explains the distinction in initial product form factors. | |
| Scaling Technical Talent and Engineering Recruitment | 2 | 1 | 1 | 1 | The conversation shifts to lighthearted banter covering pull-ups, TikTok algorithms, and Gabe clarifying that he sleeps on a regular mattress without a bed frame rather than an air mattress. | |
| Non-Consensus Views: Organizational AI and Collaborative Systems | 6 | 5 | 0 | 0 | Sarah asks for non-consensus predictions, and Gabe argues future value lies in organizational-level AI productivity rather than individual copilots. Elad enriches the point with analogies to Figma's collaborative shift. |
Statements from this episode (20)
Pereyra: Harvey reaches nearly 1,000 customers and 500 employees
“We're almost at a thousand customers, 500 employees. Started about just over three and a half years ago, and so been kind of scaling quickly since then”
Pereyra: Enterprise AI bottlenecks are orchestration and governance, not model intelligence
“When you get to that scale, a lot of the problems we're solving are not just model intelligence problems. They are these orchestration, governance, and kind of all of the enterprise product problems that you run into at scale.”
Pereyra: Harvey signed Walmart and works with AT&T and Fortune 500 clients
“So we recently announced we signed Walmart. We're working with AT&T, a bunch of these fortune, 500 large private equity firms, global 2000, kind of the largest consumers of legal services.”
Pereyra: Legal and Coding See AI Traction Because Both Workflows Are Unstructured
“Legal is so difficult is the workflows aren't structured. So the same way with programming, it's really hard until these models to build tools for programmers. You basically just had an ID and then programmers did stuff in all the different languages, but you …”
Pereyra: In Legal AI, the RL Environment Is a Client Matter
“And in legal, that RL environment is a client matter. So you have all of the context of a fund formation, an acquisition, a litigation, and the models are starting to learn. Let me go in the document management system and see if I can find this, go in the data…”
Pereyra: AI models won't change senior partner or senior engineer roles anytime soon
“When we think of the best partners we've worked with, I don't think the models are doing what they do anytime soon. And I think what's interesting is I think the role of law firm partners actually doesn't change that much in the same way. I don't think the rol…”
Pereyra: Senior systems engineering expertise won't enter AI models soon
“None of this is public. This will, won't go into the models for a long time.”
Pereyra: Legal AI requires expert reasoning traces, not just public filings
“Like all you get from these public mergers is like an SEC filing. And so you do see the final result. But most of the value or what you need, I think, to eventually improve these models is the decision making process the same way you need these reasoning trace…”
Pereyra: Complex legal drafting lacks binary verifiability for AI reward functions
“For something like generate this merger agreement, it's really hard to just give some binary like this is good or this is bad. And I think this has been like a big research problem, like with all the labs we work with, and also internally, there is just this o…”
Pereyra: Real software engineering lacks objective unit-test verifiability at scale
“I think you actually have the same problem in programming, where I think in the short term programming is verifiable, where you can look at unit tests, but once you get into real software engineering, like the unit, there is no unit test. It's like I deployed …”
Pereyra: Fortune 500 legal departments lack standard systems compared to law firms
“When we start working with the Walmarts, the very large banks, the Fortune 500, they're much less standardized than these law firms, and so there is just this massive amount of work where we go to a large bank and they say, we don't have any document managemen…”
Pereyra: Law firms are acting as Harvey implementation partners for clients
“Law firms are starting to do this for their in-house clients. So they're starting to go and take Harvey and go to their clients and say, Hey, buy Harvey and we'll help you build all the workflows and implement it because we have the scale and the expertise to …”
Pereyra: Building a tech-enabled law firm requires running two incompatible companies
“The big challenge that they ran into was, you're essentially just building two different companies, right? You're building a law firm, And you're building a tech company, and it's already really hard to, like, build product engineering, do AI, scale sales, and…”
Gabe Pereyra: Client conflicts limit AI law firms compared to software providers
“Solving that equation at scale is a much bigger opportunity than if you build a single law firm Because you get conflicted out. You can't scale this.”
Pereyra: Legal is a $1T market while professional services is $3T-$5T
“And I think to your point, the scope of this, like legal is a trillion professional services is something like three to five trillion.”
Pereyra: Narrow AI point solutions fail to leverage generalist model power
“If you had built something where it's like all this does is like check that your Python code doesn't have bugs, which you could have done better with 3.5, like you wouldn't have built something like cursor. And the intuition was just these models can help you …”
Pereyra: Legal AI Succeeded Early by Focusing on Document Upload and Citations
“I think it was finding the right form factor. And I think in legal, it was maybe a bit more obvious where the initial form factor was essentially like the initial feature we built that none of the products had at the time was upload a document and do something…”
Pereyra: Coding AI Lagged Legal AI Due to Model Capabilities and IDE Integration
“I think with coding, The initial models were also not quite as good, that you needed maybe a bit more capabilities of the base models, and then you needed, I think, figuring out the right way to, like, integrate this into the IDE.”