@dosco Ah I see. So this handles the post processing of the data to handle the querying.
English
Seth Fenster
7.7K posts
Seth Fenster
@sethtjf
Legal engineer building @litvue
Katılım Eylül 2017
715 Takip Edilen855 Takipçiler
you can recreate this whole bespoke system with just graphjin. assume postgres has a knowledge table plus one table-returning hybrid_search(q, project) function containing FTS + pgvector + recency/RRF. the queries are just examples the agent will auto generate them you don't have to. graphjin.com
Cerebras@cerebras
English
@scottastevenson Hard agree. I went from Google -> ChatGPT/Claude -> Google over the last few months. I'm also treating google's search search box for longer queries by default
English
Hate to admit it but a lot of simple questions I used to type into ChatGPT I type into Google search now
AI for Ask Jeeves type questions will be completely commoditized, free and ad-supported.
Claude Cowork/ChatGPT Work is where the future will lie for the foundation model app experiences
English
“The ___ in ___ isn’t ___. It’s ___.” Is ai’s favorite pattern btw
Filevine@filevine
The bottleneck in legal AI isn’t the technology. It’s trust. @AIforAttorneys explores how law firm leaders can build team confidence, eliminate frictions, and drive true AI adoption.
English
@artemysone @owenbjennings Maybe but even more reason for an agnostic harness so that you’re not locked in come time to renew
Also practically speaking some types of enterprises (like law firms) can’t afford to be locked into one vendor in case of potential conflicts
English
@owenbjennings I just don't see enterprises have multiple enterprise contracts. Feels like they will pick 1 provider, right?
English
@solirvine @gordon_cassie What does that even mean? Do they just copy the search results from WL into their briefs???
English
@gordon_cassie If you dig into these cases, all of the lawyers at fault seem to claim the same defense: they thought the AI chat was just an upgraded westlaw-style search input. Either way, it's user error, and no good lawyer would make these mistakes.
English
Mounting body of court opinions calling out specific AI tools that caused hallucinations.
Harvey and Westlaw already named. Of course they will all eventual be named because they are all wrappers on the same LLMs.
Rob Freund@RobertFreundLaw
Another lawyer submits briefing with fake, AI-hallucinated cases, gets sanctioned. Lawyer blames Westlaw's "Precision" AI tool, court doesn't care. "This Court addresses a matter of profound concern regarding the integrity of the bankruptcy process, the professional responsibility of counsel, and the proper administration of justice." "The Petitioning Creditors, through Mitchell, responded with motions to quash containing fabricated quotations, non-existent citations, and invented legal authority apparently generated by artificial intelligence and submitted them to the Court without verification." "As justification, Mitchell asserts that he 'has the "Precision" version of Westlaw, which, upon information and believe [sic], does incorporate AI capabilities.'" "S.D. Tex. General Order 2025-04 expressly cautions lawyers against using generative AI services without checking accuracy, and does not differentiate between legal AI services and general AI services." "In fact, S.D. Tex. General Order 2025-04 mentions Harvey AI, a legal specific AI service, as an example of an AI tool that should be used with caution."
English
@scottew "you mean the yogi berra of our era? haha"
-- yogi berra
-- wayne gretzky
-- michael scott
English
This is fun, got a little Yogi Berra quality to it
Heshie Brody@heshie
what does this even mean
Lehi, UT 🇺🇸 English
I just don't see how openai or anthropic are releasing multiple models anymore by year end
English
@glcst yeah I agree. I say agents for simplicity but I'm often talking about an instantiation of an agent (ie a session) and I think of each session having it's own db but the agent as a stateless concept (it's configuration and tools, capabilities, etc.)
English
@sethtjf some workloads are multi-session. Some are per-session. We say "one db" for simplicity. In reality every agent can have themselves many dbs. Go big or go home.
English
What makes a good attorney? Personally, I think they should be naturally curious.
English
yes, working at @modal is as fun as it seems.
we are casting 35 more roles across new york, san francisco, and stockholm. one could be yours !
modal.jobs
English
@sethtjf That was my whole point is that it’s useful initially to introduce a new category, but ultimately it loses its meaning
English
I recently I gave the following definition:
"a new category of firms built from the ground up with the understanding that (1) AI exists and (2) because AI exists you wouldn’t build the same kind of firm today that you might have built five years ago." legallydisrupted.com/p/are-ai-first…
WillC@willchen500
Any founders of AI native firms want to chime in here?
English
should have called nuclear power giant steam power and they would have been everywhere
English
Yeah I didn't read the article yet. Wanted to add my 2c though bc this is the internet.
I'm not cynical on the concept just the naming. In the same way that I wouldn't describe any law firm as "computer-native" or "computer-first" I don't see a long-term benefit to being "ai-native" bc every firm will be using it in one way or another.
English
I addressed that in the newsletter FWIW
My take: I would argue the term “AI-first” (or its cousin native) was quite useful for introducing a new category of firms built from the ground up with the understanding that (1) AI exists and (2) because AI exists you wouldn’t build the same kind of firm today that you might have built five years ago. And I, myself, remain bullish on the opportunity to build a new kind of firm, but I get the cynicism. I also don’t quite understand why everyone of these firms needs to raise capital. If the AI is here, why not just start?
English
@solirvine @thinkymachines I find their website and blog in particular so visually appealing
English