Josh Stein

@twilio gave us “Ask Your Developer.” @Anthropic just made it hardware. 😆 cc: @jeffiel

Big congrats to Zoe Weil and the @sequen_ai team on the Series A, and to my partner @lisamxu on a great call here. Excited for @thresholdvc to be part of it. Sequen is building a powerful new approach to real-time behavior optimization, helping enterprises move beyond static personalization toward dynamic, in-session experiences in the AI era.

For years, @parkerconrad has been making the case that the real unlock in business software is the underlying employee data model. That’s why this launch matters: Rippling’s AI isn’t a thin wrapper on top of disconnected tools, it’s built on the system of record for the workforce. The customer posts are worth reading.

Spellbook has taken a very different path than Harvey, Legora, NewCode, CoCounsel that focus on a general application sold to Biglaw and instead created a purpose built AI product they sold direct to commercial lawyers. This could end up being the more durable path to market.

One of @levie’s many gifts has been the ability to see the future that’s coming, clearly articulate a vision for it, and push @Box toward that goal. Consumer → Enterprise. The iPad. Desktop → Mobile. And now AI and agents. It’s hard to describe just how rare and special that ability is. 🚀

Agents need files to keep track of their work, they leverage files as context about the tasks they’re doing, and use them to share back and forth with their human counterparts. @levie spoke with @CNBC about and the importance of agents having their own filesystems.

@scottastevenson So many bonus points for the old-school CRT!

Casually vibe coded a working Super Nintendo game in the background while vibe coding Spellbook. Here it is running on my original console. We live in unprecedented times.

3 months in, today is fun because we start to collaborate with builders. We have been hard at work. 🔥 Our goal by the end of year is to make the entire AI stack adaptable. Data is the foundation all AI progress has been built on. So, it is our natural starting point.

100% agree. Law is full of smart, ethical people trying to do good work. AI makes it possible to do great work, and much faster. But when revenue is tied to time, efficiency reduces billings. Classic structural misalignment and incentives drive behavior. A similar dynamic in residential real estate drove our investment in @Redfin. Agents were paid a fixed percentage of the home price regardless of effort or efficiency. There's a real opportunity with AI to drive not just productivity but reinvention of the legacy business model.

Legal AI products have diverged: those designed to sell to BigLaw innovation committees, and those designed for end users. In BigLaw, there is no incentive to reduce billable hours with AI hidden behind the curtain. AI needs to be a PR play: show clients that the firm is innovative and not falling behind. Legal AI is so visible because every firm has been clamoring to make AI press releases for 2 years. This is the “evolutionary pressure” that drove companies like Harvey and Legora to simultaneously launch “Client Portals”—a way for law firm clients to log in and see the “innovation” first hand. We launched a client portal years ago, before our AI product, due to similar pressures. Clients hated it. They wanted to engage with their lawyers where they already work: email, Slack and Teams. After that experience, we committed to focusing Spellbook on the end users, the lawyers who would use us day-to-day to get work done. For many of our customers, there is no sales meeting or presentation. There is a 25 minute test drive. We focus less on top-down sales through committees, and more on bottom-up sales through delighted users. It’s a slower path. We prove value to users one by one rather than relying on selling top-down licenses across firms. But we think this subjects our product to better evolutionary pressure. With patience, we’ve grown to 4,000 in-house legal teams and law firm customers since 2022—more than Harvey and Legora combined, and we’ll soon cross $100m ARR. 4,000 customers who use us purely for the delight and utility of our product, who are rapidly expanding usage. This also forged an incredible fit with in-house legal teams, who need AI for the pure utility. AI buyers in every vertical should consider: “Should we buy AI designed to sell to the innovation committee, or to end users?” These approaches result in very different product DNA.

I’m excited to share that @thresholdvc has invested in Inertia Fusion’s Series A, announced today. Really excited to be working with @jeffiel again! linkedin.com/posts/joshstei…

Proud to share that @thresholdvc has invested in @adaptionlabs, which is announcing a $50M seed today. I couldn't be more excited to partner with @sarahookr and @sudip_r0y as they build (at lightning speed!) fundamentally different AI systems - models that are highly efficient and adapt in real time. Congrats to the entire Adaption Labs team on this milestone. Even more exciting things coming very soon - stay tuned! 🚀

@scottastevenson @SpellbookLegal There’s always that moment where you wonder: "is this actually market?" This is so huge. Bravo.

Introducing Compare to Market, one of the biggest things we've ever launched in @SpellbookLegal. A lot of contract negotiation still happens in the dark. Opinions. Anecdotes. “This feels market.” For decades, lawyers have had to rely on experience and intuition to judge whether a term is standard. But markets move, and memory doesn’t scale. With Compare to Market, Spellbook lets you see how a contract stacks up against similar agreements in realtime. And unlike previous AI tools, you can actually inspect the stats backing the insights. You'll soon be able to filter by geography, or industry rather than relying on AI that has been trained on "everything". Lawyers are getting sick of "AI slop": AI outputs that look reasonable at first glance, but that aren't really grounded by anything. This is just one more step in our mission to build the most trustworthy, data-backed AI contract review tool.

Incredible run over those 20 years. Bravo, sir.

Congrats to @scottastevenson and the @SpellbookLegal team on Dropbox selecting Spellbook for AI contract review. A great example of AI delivering real, operational leverage inside large enterprises and also how the ecosystem will come together with great integrations like this. Do @BoxHQ next! 😀

2026 is the year of "no slop" AI Today we're launching preference learning in Spellbook Lawyers are sick of outputs that lack nuance and grounding in their practice. Spellbook now learns contract review preferences automatically—as you accept and dismiss suggestions.

I'm starting a new project. Working on what I consider to be the most important problem: building thinking machines that adapt and continuously learn. We have incredibly talent dense founding team + are hiring for engineering, ops, design. Join us: adaptionlabs.ai

Never surrender

@SpellbookLegal At Threshold, we back founders who turn breakthrough technology into indispensable tools for real work. Spellbook is already doing that for legal teams worldwide, and we’re proud to be their partner in this next chapter.

@SpellbookLegal Scott and the Spellbook team have shown exceptional clarity of purpose. By rethinking legal work from first principles, the team has built a system that continuously learns from every review, improves its recommendations, and becomes more effective with each use.

We’re thrilled to have partnered with @SpellbookLegal, a company with a bold vision of reimagining how lawyers do their work by leveraging the power of AI. businesswire.com/news/home/2025…