Roko's Basilisk
Posts
Loyalty Before The Law Degree

Loyalty Before The Law Degree

Plus: Anthropic's clash helps sales, Genesis reveals Eno, NVIDIA doubles down on infrastructure.

Roko's Basilisk
June 18, 2026

Here’s what’s on our plate today:

🧪 Why is Spellbook just betting $1M on people who don't have law degrees yet?
📰 Anthropic's feud may be paying off, Genesis unveils its humanoid, and NVIDIA's infrastructure push.
🛠️ Three tools worth trying: Spellbook, Harvey, and Hugging Face Legal Datasets.
🗳️ Poll: Is Spellbook's student strategy smart or a quiet problem?

Let’s dive in. No floaties needed…

In partnership with

AI agents now read your docs almost as much as humans do

5% of traffic to your docs is now AI agents, not humans. If your documentation isn't structured for machine readability, your product is invisible to Claude, Cursor, and every other coding agent your buyers use daily. Mintlify is built for both audiences.

See how it works

_{*This is sponsored content}

The Laboratory

TL;DR

Fellowship as a foothold: On June 9, 2026, Spellbook launched a $1M Legal Fellowship Fund, offering law students $25k in project funding, five years of free access, and mentorship from co-founder Daniel Di Maria.
Old playbook, new stakes: LexisNexis and Westlaw built loyalty this way for decades, but AI drafting tools shape how lawyers work, not just where they search.
Capability gap makes familiarity the product: Claude Fable 5 hit 13.3% on Harvey's Legal Agent Benchmark, up from 10.4%, meaning most complex legal tasks still fail, so habit fills the gap merit can't.
Deep pockets, deeper reach: Harvey ($11B) and Legora ($5.6B) can outspend Spellbook's fund many times over if law schools become the new battleground.
Stakes: Whoever wins the muscle memory of this cohort may win procurement decisions nobody remembers making.

Why Spellbook just bet $1M on people who don't have law degrees yet

A key function of marketing is to communicate with consumers in ways that ensure a brand comes to mind whenever they think about a particular product category. Known as brand recall, this principle has become central to how organizations market both products and services. And, one of the most effective ways to build that recall is to familiarize future consumers with a product long before they are responsible for purchasing decisions.

Spellbook is investing in law students today in the hope that the AI tools they learn in school become the ones they carry into legal practice tomorrow. Photo Credit: Scholarship OWL.

There is a familiar pattern across industries that sell to professionals. Companies encourage students to adopt their products while they are still in classrooms, universities, or training programs, knowing that habits formed early often persist. Their logic is that when a student learns a particular workflow in a particular piece of software, they frequently carry that workflow into their first job, their first independent decisions, and their recommendations to colleagues, simply because it is the version of the task they already know. Over time, familiarity turns into preference, and preference can evolve into long-term loyalty, making student adoption one of the most durable forms of brand building.

Spellbook, a Toronto-based AI contract-drafting company, just made a version of that bet, at a scale that's small in dollar terms but interesting for what it signals. On June 9, 2026, the company launched a $1M Legal Fellowship Fund, offering law students $25k in project funding, five years of free platform access, and one-on-one mentorship from co-founder Daniel Di Maria.

According to Law.com, the program is open to law students of all levels and tasks accepted fellows with building their own legal tech tool over an academic semester, with Spellbook's teams providing testing and working sessions along the way.

Along with the fellowship, Spellbook also expanded its push into legal education through a broader Academic Partner Program. The fellowship is part of the company's wider effort to familiarize future lawyers with its tools early in their careers. Spellbook says the program provided more than 1k students across more than 50 law schools with free access to its platform during the previous year, while its academic page currently advertises free licenses through the end of 2026. The company has also sought to extend that strategy beyond the classroom.

Separately, the law firm Kennedys partnered with Spellbook on what LawSites described as one of the first large-scale initiatives in the legal sector designed to address AI's impact on early-career legal training.

That strategy would be noteworthy in any market. In legal AI, however, it is unfolding at a moment when the underlying technology itself is advancing quickly enough to make the next generation of lawyers a particularly valuable audience.

What a fellowship actually buys

The thing to note about law schools is that they are one of the few professional pipelines left where the training is still genuinely formative, where the tools a student learns on aren't incidental; they become muscle memory. A first-year associate doesn't relearn how to draft a contract from scratch; they draft it the way they were taught, using whatever software was sitting open on the laptop during their clinic hours. If that software was Spellbook, the associate's first instinct, several years later, when their firm asks whether to adopt an AI drafting tool, will be shaped by the one they already know how to use.

Stack the fellowship, the academic licenses, and the Kennedys training partnership together, and a pattern emerges that has less to do with charity and more to do with sequencing. Free access while you're a student. A fellowship if you're ambitious. A training partnership if your first firm already uses the tool. By the time anyone in this cohort is asked to weigh in on a procurement decision, an earlier version of them will already have answered the question, without realizing it was being asked.

Spellbook’s new fellowship and academic partnership programs are designed to give law students funding, mentorship, and years of hands-on access to its AI legal tools. Photo Credit: Spellbook.

What makes this strategy particularly interesting is that the legal industry has seen similar approaches before. For decades, legal research companies such as LexisNexis and Westlaw invested heavily in law school access, recognizing that students who spent three years learning one research platform were often more likely to carry that preference into practice. Spellbook's fellowship appears to be applying a similar logic to the AI era. The difference is that legal research tools primarily helped lawyers find information, whereas AI drafting platforms are increasingly becoming part of the work itself. If that shift continues, the habits formed in law school could influence not just which products lawyers prefer, but how they approach legal work in the first place.

Why this matters more right now than it would have a year ago

Normally, this kind of move would be background noise, the legal tech equivalent of a free trial. But it's landing at a specific moment in legal AI's development, and that moment is what makes it worth a closer look.

The same week Spellbook announced its fellowship, Anthropic released Claude Fable 5. According to Harvey, the model achieved a new high score of 13.3% on its Legal Agent Benchmark, which assesses whether AI can complete complex legal tasks end-to-end. Harvey described the improvement from the previous model's 10.4% score as "a meaningful step up." Even so, a 13.3% score means the system still fails to fully and correctly complete most of the difficult legal tasks it is tested on.

What this tells us is that the underlying models, the actual intelligence doing the legal reasoning, remain a considerable distance from being reliable enough to win this market on capability alone, and that distance could persist for a long while yet. In a market where the core product can't differentiate itself on its merits, companies don't stop competing; they shift what they compete for. Familiarity becomes the product in the absence of a clearly superior one.

That helps explain the strategic logic behind programs like Spellbook's fellowship. It also raises a more complicated question: what exactly are these initiatives buying?

The part that's easy to miss

There's a tension worth naming here, and it isn't cynical, just structural. Spellbook's fellowship genuinely does fund real projects, real mentorship, and real access. The students who go through it will, by most measures, come out ahead, with money, connections, and a tool they know inside out.

But the thing about loyalty built this early is that it doesn't require consent the way a sales pitch does. Nobody sits a third-year law student down and frames the exchange explicitly, funding now in return for a default recommendation later. That's not how it works, and it doesn't need to work that way for the effect to take hold. The loyalty forms quietly, as a side effect of familiarity, the same way a person doesn't consciously choose to prefer the layout of a kitchen they grew up cooking in, their hands just move that way.

The legal market is arriving at this realization later than other industries, partly because law has historically been a slower-moving, more credential-bound profession, in which the firm where someone trained mattered more than the tools they used. That's beginning to shift, as the tools themselves start to carry as much weight as the pedigree. Spellbook is not the only company that has come to this realization, which raises the question: What happens if several companies pursue the same finite pool of students?

What this looks like from the other side of the table

Harvey, valued at $11B after a March 2026 raise, and Legora, valued at roughly $5.6B, both have far deeper pockets than Spellbook's million-dollar fund. If familiarity is becoming a meaningful part of the competition, the companies with the most capital to spend on educational access carry an obvious structural advantage, the same way a company with a larger marketing budget eventually outlasts a scrappier competitor on sheer reach, except here the reach is into law school curricula rather than billboards.

There's also a quieter question underneath all of this, about what happens to the students themselves, not as future customers, but as people learning to think. A law student whose formative experience of contract review happens entirely inside one company's interface absorbs that company's assumptions about what good drafting looks like, often without ever seeing the roadmap those assumptions came from. This isn't necessarily harmful; it's simply happening, largely unnoticed, while attention stays fixed on benchmark scores and billion-dollar valuations.

Back to the muscle memory

Think again about that pattern professionals fall into early in their careers, the workflow learned once, on one set of tools, that quietly becomes the default for everything that follows. Nobody remembers exactly when the habit formed, years later. What persists is just the habit itself, the felt sense that this is how it's done, this is what the work feels like.

Spellbook's fellowship is, on paper, a $1M line item announced during a busy news week, easy enough to file under minor. But minor announcements are sometimes just early ones. The benchmark wars happening in parallel, the 13.3% scores, the billion-dollar valuations, and the model releases are all loud, visible, and contested. The fellowship is quiet, small, and barely contested at all, because right now, almost nobody in the room is a customer yet.

Which raises the obvious question: by the time they are, will anyone remember there was a choice to make?

Thursday Poll

🗳️ Spellbook is locking in loyalty from law students before they're customers. Smart strategy or quiet problem?

Launch fast. Design beautifully. Build your site with Framer Agents.

Framer is a pro website builder trusted by companies like Miro and Perplexity, helping creators, teams, and businesses ship production-ready sites faster than ever. With AI agents built directly into the canvas, teams can design pages, manage CMS content, write copy, add SEO, and audit for issues — all without leaving the tool where the real site lives.