←
Daniel Levine450 posts
Daniel Levine avatar

Daniel Levine

@daniellevine

CEO @MinuteBox | AI for corporate and transactional legal teams #legaltech

◷ Joined April 2008▧ 23 Photos and videos
Daniel Levine avatar
Daniel Levine@daniellevine·18h

There is a complete disconnect between the R&D and GTM teams at @OpenAI and it is impeding the great work I’m sure @jboehmig is trying to do. I’ve spoken to half a dozen legaltech startups at @tltfsummit this week. They all told me the same thing. None are building on @OpenAI and probably not for the reasons you think. The models across the frontier are all good enough. I’m sure Astra for Law is great and would be worthwhile building on top of, however, their commercial team requires large minimum spend contracts in order to get ZDR. This makes it impossible for young companies to choose them when they start building. It was incredibly easy to obtain ZDR from @AnthropicAI and @GoogleAI. In less than 24hrs we received ZDR certificates or confirmation from both. Then we started building on top of them. OpenAI? Had to take a call with an extremely brash account executive. We had to commit $25k minimum spend to get ZDR. This was before we wrote a single line of code for our new harness. Very shortsighted and many I’ve been speaking to this week had similar experiences. Platforms are supposed to make it easy to work with them and reduce friction. In the last decade, AWS and GCP would throw millions in cloud credits to get you to build on their platforms and lock you in. I’m not sure why OpenAI went in this direction, but they should consider making it easier for startups the build on top of their great models, especially if you are spending so much time and resources building for legal, which must have ZDR. We will spend over 7-figures with Anthropic and Google this year. 8 figures next year. $0 on OpenAI. All because the commercial team made it difficult when we started. #tltf2026

OpenAI is the only frontier lab focused on model capability for the legal domain. Today, we released a huge leap forward with our latest configuration of Astra for Law. openai.com/index/astra-law-6… Here's why it's the best choice for anyone building in legal:

Steven Pulver avatar
⟳ Daniel Levine retweeted
Steven Pulver@StevePulver·Nov 16

Officially launching today! So proud of the team for what they’ve built on top of our market-leading, proven system of record infrastructure as we double down our focus toward an intelligent system of work built for paralegals and legal professionals. Much more to come soon!

Announcing MinuteBox AI. An AI agent that helps you complete corporate legal and governance work in the records you already keep in MinuteBox, available today in the US, the UK and Canada. minutebox.com/blog/introduci…

0:46
Daniel Levine avatar
Daniel Levine@daniellevine·Nov 14

Our AI governance review would’ve caught this. Yikes.

$NVDA failed to buy D&O liability insurance. This is just amateur hour for a company of this size. They are going to be buried under litigation. This just keeps getting worse.

Post attachmentPost attachmentPost attachment
Daniel Levine avatar
Daniel Levine@daniellevine·Nov 14

This is actually one of the things law firms struggle with the most in the “business of law.” “Did I get a good return on my legal spend?” It is so tough and almost impossible to answer because, following most legal expenditures, a return on legal spend is quantified precisely by having no follow on legal expenditure to the one you paid for. I.e. no dispute or no litigation. It is very tough for clients to see “no event happening” as a return on legal spend and, further, attribution to the legal spend itself is nearly impossible. since there are many other reasons for “no event happening.” This is very similar to insurance, where it’s possible to see a quantifiable “return” on the basis of an insurance settlement, but for most people, a much better return is no loss at all in the first place. Paradoxically, it’s much easier to quantify the return if your house burns down and you get a great settlement than if you pay premiums forever and never have to make a claim. Most people would probably much prefer never having their house burn down in the fist place. But this will lead them to ask if they got a “good return on their insurance spend?” The problem of value in law (at least in solicitor work) is attribution. Litigators don’t have exactly this problem. To be sure, they have many others.

Our firm has done >100 deals over the last 10 years including add-ons spending ~$100M on legal. Total disputes are <$5M. Did I get a good return on my legal spend? Why shouldn't I ask the question if I could find a way to spend $50M even if I lose on a dispute or two.

Daniel Levine avatar
Daniel Levine@daniellevine·Nov 14

It is remarkable how little people seem to understand about how lawyers work, where their fees come from and where their value lies. The reason a lawyer uses his or her 79 page financing precedent over the one that is AI generated or that you bring them from LegalZoom or even from your last financing isn’t because they want to bill you drafting their own docs. They aren’t doing that. The reason actually saves you money. It’s that they know their own documents inside and out. They know every definition, every reference, and they know every typical point of contention already. So this idea that you’re going to generate your complex transactional documents in AI and just have a lawyer “look it over” is nonsensical and will cost you more, because the lawyer now has to scrutinize every word, every definition, every clause and every reference for the first time. This is harder and takes more time. To be sure, the practice of law and the consumption of legal services is going to change. But it won’t be in this direction of AI does 90% of the work and you’ll just get it checked over to save on fees. You can already do that today without AI and every lawyer will tell you it’s cheaper if they just do it themselves. This is the same reason so many lawyers don’t like giving ILA. The idea of putting up with “a little lower quality” makes no sense either, especially in the context of mid and large cap transactional documents where a missing comma can be a $5 million mistake (see Oakhurst Dairy). Legal is one of very few functions where what you are paying for is precisely the opposite of “a little lower quality.” You are paying for extreme scrutiny and attention to detail so that when a dispute arises 5 years later, you are in as strong of a position as possible. Lawyers earn their money in that hardest and last 10% of the work. It has very little to with the present-day transaction. Most of that is dealt with in the term sheet or LOI. So accepting “a little lower quality” is more like saying “I’ll accept all the risk if there’s a dispute over this transaction sometime in the future.” You can have the same outcome with no lawyer at all, save all your legal fees and just pay for it later. You’ll have to account for that somewhere with present-day dollars.

Our firm spends over $10M a year on legal fees. Almost all of it on M&A transactions that are all pretty similar. The funny thing is we have very few disputes to point to where you say this is the value we got for that legal spend. My understanding is that AI is not there yet. It can't really do a good job marking up a legal document while understanding the context for the transaction. But I am pretty sure in the next 5 years it will be. We could still have a lawyer look things over but have AI do 90% of the work. And we would be willing to put up with a little lower quality if we could significantly cut our spend. Change in the legal world is coming....

Daniel Levine avatar
Daniel Levine@daniellevine·Nov 14

I will never sleep on a bed that doesn’t have an MCP server.

Hey @daniellevine - happy to go through these 1. Orion is up to a $1000 cheaper. We put a big focus on our design and manufacturing, for simplicity. 2. Completely agreed that you shouldn't have to subscribe to your bed, which is why we made the subscription optional. 3. We're adding support for as many alternative forms of control as possible: Siri, MCP, API, Alexa. What else would you like to see? 4. We take user privacy very seriously and understand that consumer health data must be treated carefully. 5. Would love to get your thoughts on the hose that we've built. 6. We have a built-in quiet mode, which allows the device to run at lower speeds when you want to tradeoff sound for performance. 7. We would love to get your take on our app, which is currently going through a redesign. 8. We're working on home assistant. 9. We have most of our account management and secondary devices built into the app. It's just where it belongs. 10. We completely agree on having a more full offline mode. Easier said than done, but in the works. 11. I can't comment on Eight Sleep's upgrade pricing.

Daniel Levine avatar
Daniel Levine@daniellevine·Nov 14

I have wondered for a while why there seemed to be no real competitor to Eight Sleep, so this is great to see. I have an @eightsleep Pod 4, and it is one of my favourite things. I don't know how @harrygestetner1's @orionsleep compares, but, in a sea of positives, here's my list of negatives on Eight Sleep, 1. Stupidly expensive. Totally worth it. 2. Subscribing to your mattress topper is total insanity and shouldn't be a thing. 3. Having to slap the side of your bed to change temperature is a horrible UX. It works about 50% of the time and is incredibly annoying, especially in the middle of the night. The haptic vibrations in response to your bed-slaps are also loud and will wake whomever is on the other side of your bed. 4. @m_franceschetti has tweeted about aggregate sleep stats from monitoring customers that use Eight Sleep, which is creepy and weird. If you collect customer data, you should treat it with respect. 5. The hose is thicker and more cumbersome than you might expect. 6. The Pod can get quite loud, especially if you really crank up or down the temp quickly. Usually doesn't last too long. 7. The latest redesign of the app is trying to let the company tell some sort of AI health story, but is much less usable and is a complete step back. 8. No official Home Assistant support (would solve number 3 and 7). 9. They ask you to fill out a Google Sheet with your address when replacing your water filter, which just seems like an odd anachronism for such an otherwise modern company. 10. It should work without Wifi. 11. Upgrade path to newer pods is unreasonably expensive. Unclear if worth it. How does Orion compare on these points? PS, no matter what you do and no matter what anyone tells you, I hope we can all agree, do not ever, ever get a BedJet

Anthony Pompliano @APompliano is bashing my product because he is an Eight Sleep investor. Pomp, if you are going to publicly claim Eight Sleep is better and call us a knockoff, disclose your direct financial interest first. Perhaps if you were better rested, you wouldn’t need to dunk on founders. For anyone who does not make money off our demise, I want to give you an early Christmas gift. DM me your bed size and I will send you a free unit. Sending out 5 today.

Post attachment
Daniel Levine avatar
Daniel Levine@daniellevine·Nov 13

This is easily provable. If you lose your childhood home because of a lawyer's mistake then surely there will be a bar association complaint and a cause of action with the insurer, naming the lawyer who made the mistake. Since this is a totally true founding story, let's see the receipts.

The world doesn’t need another Harvey. It needs a Mike. Our AI legal agent remembers your business, takes the first pass on contracts and helps you find vetted lawyers. Built by @IndexLawHQ. Backed by @ycombinator. Give it to Mike: usemike.com

1:25
Daniel Levine avatar
Daniel Levine@daniellevine·Nov 13

Ironically, this launch highlights exactly why the world needs @harvey and lawyers, rather than vibe-lawyering. A real lawyer would have searched and warned the company before launch about a likely confusingly similar common law trademark owned by @MikeOSSTeam and @willchen500

The world doesn’t need another Harvey. It needs a Mike. Our AI legal agent remembers your business, takes the first pass on contracts and helps you find vetted lawyers. Built by @IndexLawHQ. Backed by @ycombinator. Give it to Mike: usemike.com

1:25
Daniel Levine avatar
Daniel Levine@daniellevine·Nov 13

This is a problem of incentives, not cluelessness. Investors are signalling with their capital that they want to invest in “AI-native” startups like “AI-native law firms”, which creates an incentive for a founder to say their startup is AI-native. This doesn’t change that nobody knows what that means.

Investors in the legal space, when you hear the phrase “AI Native Law Firm” know the person across from you either hasn’t a clue or thinks you’re an idiot.

Daniel Levine avatar
Daniel Levine@daniellevine·Nov 11

For the same reason law schools don’t ban Westlaw/Lexis and tell you to go search the stacks

Can you explain why prohibiting AI would knee cap potential earnings, rather than the opposite? Wouldn't allowing AI lead to de-skilling and therefore worse potential earnings?

Daniel Levine avatar
Daniel Levine@daniellevine·Nov 11

Opinion and taste is far more important to creating great software than anything else. It matters the most when going from 0 to 1. Once you release your product and it is out in the world, it no longer belongs to you. You had plans to do this and plans to do that, but you started getting customers and they asked you to do this other thing or that other thing, and all of a sudden you are weighing your direction with the direction of the market. The market is never wrong, but creating software by committee seldom produces great products. That’s why product vision is so important. The best products have a strong sense of opinion and taste, but are directionally guided by the direct feedback of paying customers. Finding that balance is how you create world-class products.

The idea of "talk to your users and ask what they want" as a silver bullet for PMF discovery is majorly oversold Like obviously you should talk to your users and potential users But if you don't have some unique insight and opinions about what you are doing....why are you the one to be doing it. I see a ton of entrepreneurs trying to take this advice at face value - starting discussion threads to solicit feedback for trying to schedule interviews to learn about the space. The funny thing is that, often these entrepreneurs are steering the direction they want to go anyways so the entire thing is theatre. Like the founder of the social network for cats: "What are the major frustration you face with your cat?" ".... umm I don't have any." "Do you sometimes wish that your cat had a pal?" "...umm i guess that would be cool." "Ahh I see so you probably want some sort of online social network with a mobile app that allows pet owners to connect and arrange meaningful experiences for their feline pals." "Ahh sure... can we end the call now." At the VC pitch "we spoke to hundreds of pet owners and heard unanimously that they want a social network for cats"

Daniel Levine avatar
Daniel Levine@daniellevine·Nov 11

The “AI-native law firm” discourse reminds me of all the D2C brands of the last decade that promised lower prices at the same quality as traditional brands because they were cutting out the middle-man and had no brick and mortar, only to see all of them open up retail stores once they needed real distribution outside the podcast universe. Every "AI-native law firm" pitch boils down to one of two things: 1. Cheaper legal work, priced on outcomes instead of hours. 2. Marked-up inference with a law license and an insurance policy. Clients will figure out which one they're buying fast. If they do succeed in disrupting traditional law, I think their pricing, size and overhead will eventually climb to match or exceed traditional law once they reach scale, much like all the D2C brands. x.com/SMB_Attorney/status/21…

Another “AI-native” law firm (wtf does that even mean) launched a few weeks ago with millions in VC money and a big splash on social. Their pitch: “We review contracts faster and cheaper than regular firms.” A few people tagged me: “You’re cooked, bro!” I stayed quiet because I didn’t want to be a jerk. Anyway, that phase is over. You can tell how much experience they actually have in our profession by what they think clients are paying for. Clients don’t need us to identify 47 potential issues in a contract and turn AI slop in a half hour. If your entire pitch is “we do routine legal tasks faster and cheaper,” you’re not competing with us. You’re competing with Claude and ChatGPT. And you’re going to lose.