Zack Shapiro

@DeFi_Andree No, unfortunately. You have to work in the building.

@zackbshapiro Wait can anyone just walk in there?

Fun fact: the top two floors of that pointy building in the middle (the old SDNY courthouse) are one of the most beautiful libraries in the city. Used to hang out there all the time back when I was clerking.

@digitalages 💯

@zackbshapiro AI-native seems to imply that some unified ai operating system exists for lawyers, but it’s just the ability to use the right ai tool in your work with zero friction that can be a chat app or an embedded plugin or an agent anthropic gets you most of the way there

Is the run by lawyers who are themselves AI-native? IMO this is more about the instincts of how and when to use AI, rather than some numerical percentage of workflows.

(Also, it's important that it actually be run by *lawyers*, and not engineers)

Bitcoin Policy Institute's Quarterly Report for Q2 is live. You can read about our latest investigative research on CCP influence, our meetings with Taiwan’s central bank, Bitcoin’s role in two Armed Services committee hearings, the fight for dev protections in CLARITY, and much more. Attached below is the executive summary. You can access the full report here: btcpolicy.org/articles/q2-20…

Cool billboards, @ryanjdaniels I like the differential lightning!

FYI, this is an enormous blessing in disguise for solos/boutiques that are pushed to use the frontier directly.

“In my whole life I've never been good at something I wasn't very interested in. It just doesn't work. There's no substitute for strong interest.” — Charlie Munger

@marine_duck That’s the Heppner case. It concerns use of consumer-tier version of Claude by a non-lawyer with training by Anthropic allowed in the settings. This has nothing to do with enterprise use of AI by law firms. But thanks for your concern about my looking silly.

@zackbshapiro Saying it doesn’t make it so I just found an american case that proves I am right If I was you I would stop replying before you start to look silly aoshearman.com/en/insights/ao…

@marine_duck What makes you think it’s not privileged?

@marine_duck Because using AI doesn’t defeat privilege

@zackbshapiro Why aren’t you advising them to do it right now?

@marine_duck Why has no law firm in America tried this?

@zackbshapiro I think you are wrong and I think it’s a great case for a law firm who wants to sue and find out all the information provided by their opposing counsel to an LLM in a matter

@marine_duck I think you are confused. A law firm using technology like email, cloud storage, or (with the right privacy settings) an LLM does not vitiate privilege.

@zackbshapiro If the information is not held or owned by the law firm. The legal privilege is also a legal right held by the client, not the LLLM, so the LLM can’t claim the right as a defence to a legal action. Those are two arguments off the top of my head

If the information provided to an LLM isn’t covered by Client-Attorney privilege then the agreement can’t eliminate the risk. Having everything in-house retains the client-attorney privilege. To put it another way, if opposing counsel can just subpoena an LLM to get all information provided by a firm to an LLM on a matter in preparing their case then there is a huge risk to using legal ai

Good essay but clearly the context needed to provide to the LLM to give it a good answer: the “background on the client, the posture, the business objective, the counterparty dynamic, the legal and commercial issues etc” all these things are clearly proprietary and sensitive commercial information that no good lawyer should be inputting into a LLM Big law firms are going to have to build their own models to protect this IP or AI labs are going to have to legislate that legal work inputted into a LLM is legally privileged before lawyers can have confidence using these tools

@marine_duck Most big law firms have enterprise agreements with frontier labs that protect client data. Building their own models would be expensive and not something they are even capable of.

@SteelToeScribe @investing_law Yes, you should not rely on AI to cheat on your homework. The market will swiftly teach that lesson to anyone who defies it. But it is equally important (at least) not to mistake that lesson for any version of “don’t rely on AI for legal work.”

@zackbshapiro @investing_law AI is the best mentor Michael is (clearly) making a broader point, which you know

“I can do it all with AI” stops working when you need to negotiate on a call. 😅 Or explain something to the client on the call. Or do…anything on a call. Unless of course you make an AI version of yourself pre-trained on all laws and legal docs, and give it your Zoom credentials. 🤖

@investing_law Of course you can. You use AI to understand the document better and prepare what your responses will be to the other sides' likely requests. It's about using the AI to enhance your thinking/lawyering, not to turn your brain off and just rely on reading or memorizing AI output.

@zackbshapiro Was on a call yesterday where opposing counsel ran through questions on a 100-page document without providing any context in advance You can't prepare for that using AI

The limiting factor is not the tech, it's knowing how to use it properly. x.com/zackbshapiro/s…