Adam Griffith

@DennisonLaw I’m worried this man does not have enough hats

Another reason why the AI explosion gives special advantages to mid-career professionals: your email inbox is a treasure trove of data on your style, skills, and approach. Google Takeout is your friend: takeout.google.com/?pli=1

@DietCoke_Esq Similar: “Here’s our standard order form that quietly incorporates like 30 pages of one-sided T&Cs via URL.” No read, only sign!

I hate when the opposing side is like "this is our standard agreement." Yeah, thanks for sending it over as a PDF—I'm definitely just going to agree to these unilateral terms and this insane indemnity clause

@SellersCounsel @ZachAbramowitz “Hey associate, you’re not able to take care of yourself while also billing 2300 hours/hour to enrich me? Damn idk, maybe you’re stupid/lazy? Have you tried wanting it harder?”

I know several small firm owners that never go to their kids sporting events or birthday parties. Meanwhile, the most successful BigLaw partner I know doesn’t miss any of his daughter’s volleyball games or life events for children of his associates. I talked to a BigLaw associate yesterday who billed 2300 hours last year but ate dinner with his family and put his kids to be every night. Those who want to make it work will find a way to make it work.

@RobertFreundLaw … so FAR

We're reaching new lows with lawyer AI misuse. Court expresses concern that deposition "testimony" submitted by counsel via declaration "was either generated by generative artificial intelligence or otherwise fabricated."

We all had (were?) that friend who tabled for Westlaw or TR in the common area at lunch. Prepare to see that same thing for Legora and Harvey. Not sure if lawyers will become siloed into one ecosystem in the same way, but I won’t bemoan part time jobs for law students

Claude workflow finding: when working with documents, I’m finding that a Word comments in the doc are an ideal way to call Claude’s focus to specific text blocks. Added benefit of letting me ask it to execute many changes at once without needing to list out in prompt.

What are the limits of using AI to help run a legal practice? @zackbshapiro draws on his own experience with Claude AI to demonstrate a transformative effect. He joins @lawbytespod to discuss how he uses it and implications for the future of legal services x.com/mgeist/status/…

I’ve never been so tempted to frame and hang a court sketch in my office as I am right now.

@alex_prompter lol @ increasingly slopped lower effort ripoffs of the @heynavtoor post from this week Next up 🚨BREAKING: Claude do law, pls save munny. Why use many word when few do trick?

🚨 BREAKING: Claude can now replace a $1,000/hour corporate lawyer from Skadden Arps. For free. Here are 10 prompts that protect you from getting screwed on your next contract, lease, or business deal (Save this before it disappears)

@lawheroezV2 My experience: - best results, use a form you already like - good results, give Claude a handful of agreements that are similar but not on point, let it decide what to use - scratch drafting works well with a lot of detail in the prompt. Ask it to critique its own work afterwards

Been seeing some lawyers saying that drafting contracts from scratch on Claude/Cowork using Claude’s training data rather than based on a trusted template/precedent is pretty good. Anyone tried and can give specific examples and their experience? I’m very curious especially for long complex contracts like SPA or APA. Not an NDA.

I am once again reminding you that the below is *not* what good legal prompts to Claude look like. You need to be extremely detailed and specific about the output that you want, not ask it roleplay as a Skadden associate . . .

I wrote about the errors I consistently see when using AI tools to draft contracts. Curious what patterns the other lawyers out there have noticed and how you plan for them/correct them

@teddyvroberts It is totally possible to check for, or avoid, hallucinations within Claude itself. This is adding Word plug in training wheels rather than learning to ride the bicycle, and there are a hundred companies out there trying to sell you training wheels

@StagBadd I get that the models are getting better, but why is it the wrong approach?

This sort of tool is the wrong approach. Hallucinations have dropped off so much outside legal citations in the last 2 years, it's impossible to imagine frontier models won't figure out legal cites. They're already avoidable if you practice a little rigor in your AI drafting

@SinaiLawFirm This is exactly the curriculum in a lot of liberal arts undergrad programs. We can either say that law school is about learning to practice a trade (justifying the hurdles to practice), or it's a mindset lesson which can be taught in 2 years w/o the Bar, but not both.

Hot take - I don't think law schools should teach you how to practice law. It would be a colossal waste of time. Law school should teach you how to research, write well, and develop an analytical mindset to become the best lawyer possible.

One more IP Sunday Scary, without an aggressive reverse engineering clause that contemplates AI tools, licensing agreements for products that expose the underlying function are basically just giving away the product for free: malus.sh

The punchline is: keep it secret, keep it safe, if you want any kind of IP most at all

What you get is an ability to enforce your rights against someone who obtained, copied, and used your IP without your permission.

Some thoughts on IP protections for Ai-skills based businesses. The picture for IP rights to AI generated code is pretty clear for now, but there’s going to be a lot of businesses that rest on sophisticated Claude / Codex skills files. What about them?