Since 2013, Casetext has empowered lawyers to provide higher-quality and more affordable representation to more people, through the power of AI

Casetext’s co-founders shared a vision to expand access to the law through the application of advanced AI. Ten years later, they continue their mission to challenge the status quo, with an eye toward the future. 

From BigLaw to a Sunnyvale living room

As a student working in the Community Law Clinic at Stanford Law, Jake Heller was struck by how inaccessible legal information was compared to information in other fields. He understood for there to be equal access to justice, there needed to be equal access to the text of the law itself.

By early 2013, Heller had been a litigation associate at Ropes & Gray for more than a year, following a First Circuit Court of Appeals clerkship and fellowship with the Massachusetts Governor’s Office of Legal Counsel. A Silicon Valley native who’d been coding since he was nine, Heller was perplexed by the disparity between outdated legal tech and powerful everyday tech. “I’d spend four nights in a row until 4 AM looking for one specific court case for a client—but I could find a specific style of Thai restaurant within a mile of my house in minutes.”

So Heller teamed up with Harvard Law grad and practicing attorney Joanna Huey, who shared his frustration with the state of legal tech and his vision for disruption. Their idea to make the law publicly available through crowdsourcing—a Wikipedia for the law—was accepted by startup incubator Y Combinator, and in October 2013 a startup was born, backed by $1.8 million in seed funding.

A month later, Heller left Ropes & Gray to work full-time on building a crowdsourced legal database, while Huey left their startup to pursue a career in legal education, eventually becoming a law professor.

Soon after Huey’s departure, Heller brought Pablo Arredondo and Laura Safdie on as co-founders.

Safdie, a high school friend and former roommate, was one of Heller’s earliest advocates. A Yale Law grad, Safdie clerked in the Southern District of New York and worked as a Senate Judiciary Committee staffer before joining Simpson Thacher’s litigation practice. She was similarly frustrated with the outdated legal research tools available and supported Heller in his decision to leave practice to focus on building a modern alternative to existing legal research technology. She also strongly urged more brainstorming for the company name, in lieu of Heller’s Law Penguin.

Heller was connected by a former professor with Pablo Arredondo, a fellow Stanford Law graduate practicing IP litigation at Kirkland & Ellis. Arredondo, a self-taught coder and fellow at the Stanford Center for Legal Informatics, had previously co-founded Occam, Inc, a Sequoia-backed startup focused on building an alternative legal research engine. Arredondo, who was in the process of building another search engine, shared a similar vision for expanding access to the law and improving legal research.

The trio of co-founders worked out of Heller’s living room in Sunnyvale, where they continued developing a free legal research platform offering millions of cases, statutes, and legal texts annotated by a community of practicing attorneys. 

By February 2015, Casetext was attracting more than 250,000 users each month and had raised $7 million in Series A funding. But they were facing serious barriers to continuing to offer their free platform. Chief among them: though lawyers absolutely saw the value in such a resource, they simply didn’t have enough time to contribute to it. So the trio made the strategic decision to focus on the technology Arredondo had been building.

Finding success with “Spotify for legal research”

While working as a litigator, Arredondo had been developing AI-powered technology that would later become CARA AI. Initially known only as CARA, the tool first launched in 2016, when Casetext had just 17 full-time employees. CARA proceeded to wow the market with its search functionality. A kind of Spotify for legal research, CARA suggested case law instead of artists or songs—it reviewed a document, such as a judicial opinion, pulled out key information, and identified relevant cases to read next. 

Two years later, in 2018, CARA AI debuted as Casetext’s first paid product and became the solution that put the company on the map. Customers loved the technology because it not only identified their own gaps in briefs, but would catch the opposing counsel’s intentional omissions. 

In January 2020, while more than 5,000 law firms were using CARA AI, Casetext launched a powerful new search functionality, Parallel Search, part of a drafting tool on the Casetext platform that was built on breakthrough AI called transformer-based neural nets or large language models (LLMs). The first legal research tool to use this paradigm-shifting technology, 

Parallel Search proved so popular with attorneys that in June 2020, Casetext made its capabilities available as an upgrade to the Casetext research platform. 

OpenAI taps Casetext to bring GPT-4 to the law

As Casetext continued updating its platform to take advantage of the latest AI, the team tracked the progress of transformer-based neural net technology and LLMs, particularly the gains being made by OpenAI in generative transformer-based technology. The nonprofit AI research lab had launched in 2015 and was making notable progress with its Generative Pre-trained Transformer, or GPT, technology. 

Since Casetext’s inception, Laura, Pablo, and Jake have remained focused on their vision that AI would transform more than just legal research, advancing the efficiency and quality of work across every facet of a lawyer’s practice. They envisioned integrated technology that would empower attorneys to spend more of their limited time on complex work, ultimately elevating their practice. And they were convinced LLMs were the key to this technology.

When the team was given a look at GPT-3 in 2020, they knew the technology was advancing toward what they needed to power Casetext’s next phase of legal AI offerings, but wasn’t there yet. They never imagined it would be only months later, and that Casetext’s next phase was less than a year away.