Abe

580 posts

Abe banner
Abe

Abe

@abraham_jmb

1L / SWE / amLaw 100 SA

New York / Dallas Katılım Nisan 2022
1K Takip Edilen557 Takipçiler
Abe
Abe@abraham_jmb·
honestly hard to compare because i only really thought about legal AI abstractly before i started consulting and being on the inside of a firm. and mostly I thought legal AI couldn’t be all that diff from software engineering. i don’t think lawyers are about to be automated away end to end as much as I use to. i think firms will face moderate pressure from in house use of AI and to a degree AI native law. i think half of the use of juniors could be automated with agentic computer use (organizing files) and an outlook MCP integration to Claude, but who the hell is going to let an agent have anything but read access to imanage. bottlenecks everywhere. ai is great for improving work product, but harder for getting efficiency gains (kind of the opposite in software engineering)
English
1
1
4
451
Abe
Abe@abraham_jmb·
@moosegrav lmaoo when i started seeing opposing counsel responding in minutes to 2am emails on a deal that wasn’t closing for at least a week and a half, i realized…
English
0
0
1
628
moses
moses@moosegrav·
@abraham_jmb I remember apologizing for emailing someone at 8:45pm my first week on the job since it was so late. I have since learned.
English
1
0
3
760
Abe
Abe@abraham_jmb·
the best part abt working around biglaw attorneys is almost everyone has edge. i remember asking earlier in the summer if I could get into the building after an event that ended at like 10p, everyone laughed. this is walmart, we're open 24/7. def don't have to worry abt the optics of emailing back work product at 2a. you might even get a response.
English
5
1
91
23.5K
Michael?
Michael?@Suethehumanrace·
@abraham_jmb is edge the same thing as availability? not rhetorical but that’s just what it sounds like it is
English
2
0
10
2.7K
Abe
Abe@abraham_jmb·
@teddyvroberts lol 🎯, def been known to suffer from SS, im predisposed
English
0
0
1
1.2K
Ted Roberts
Ted Roberts@teddyvroberts·
@abraham_jmb You may be suffering from Stockholm syndrome. I’d get that checked out if I were you.
English
1
0
9
2K
42 North Counsel
42 North Counsel@42NorthCounsel·
@abraham_jmb How has your view of the promise of legal AI developed since being on the inside of a firm?
English
1
0
0
155
Abe
Abe@abraham_jmb·
i dont want to see the ant api bill from my harvey usage. i ride opus 4.8 in harvey like nobody’s business as i’m getting assigned increasingly more legitimate junior tasks despite being a summer, since some of my work product is actually usable because of my opus riding.
English
3
0
20
3.3K
Abe
Abe@abraham_jmb·
@willchen500 maybe, fable could also j not avail at my firm.
English
0
0
0
30
WillC
WillC@willchen500·
@abraham_jmb So Harvey lied about fable being on there. That’s disappointing. I think soon you will get nerfed GLM instead of frontier models so tokenmaxx while you can.
English
2
0
1
108
Abe
Abe@abraham_jmb·
@willchen500 lmaooo. but yeah, ofc i would use subsidized fable, but its not on model picker. beside the fact that harv know i could tokenmaxx 100k a month on fable, ant put rogue data retention policies on fable. also i bet all frontier models on harv are set to low thinking mode.
English
1
0
0
76
WillC
WillC@willchen500·
@abraham_jmb Can you pick fable? If I were you I would just use it on every query since it’s free and unlimited ok Harvey I presume?
English
1
0
0
90
WillC
WillC@willchen500·
@abraham_jmb Keep burning those tokens 🔥🔥 it’s fixed pricing for your firm don’t worry.
English
1
0
6
382
Abe
Abe@abraham_jmb·
@KristenForti3 100%. i think if the trend of equal intelligence tokens dropping 90% / year continues, these companies will be largely ok. but they still might have to move to new usage pricing model bcuz of a demand explosion.
English
0
0
2
197
Kristen Forti
Kristen Forti@KristenForti3·
@abraham_jmb Token usage in legal models will be an interesting problem in the coming months/years!
English
1
0
2
210
Omar Farooq
Omar Farooq@potencytoact·
@abraham_jmb living in the jargon-space of codex and claude are the new cognitive moats
English
1
0
1
133
Omar Farooq
Omar Farooq@potencytoact·
People say Opus, Fable; etc are verbose. It’s actually that we can’t keep up with them. Every outputted statement bears load in one way or another.
English
1
0
1
187
Abe
Abe@abraham_jmb·
also made the repo public so anyone can run it locally and keep their docs on their own machine: github.com/opensourcelega…
English
0
0
1
387
Abe
Abe@abraham_jmb·
if anyone wants a free docx compare software, just vibe coded one. redline-compare.vercel.app core features: 1) two word docs in, litera-style redline out (red = deleted, blue = inserted, green = moved) 2) downloads a real tracked-changes word file 3) change-by-change navigation 4) nothing stored feel free to add bugs or feature requests to the thread.
English
2
1
23
2.9K
Ben Sovocool
Ben Sovocool@BenSovocool·
@abraham_jmb Nice! If possible it’s handy to have a toggle to turn on margin change lines.
English
1
0
0
137
Abe
Abe@abraham_jmb·
legal is hurry up and wait. unlike the bun rewrite where the scope of what can be done continuously is quite expansive (months of human work compressed into one agentic loop) legal is naturally bottlenecked down to discrete tasks that are dependent on opposing counsel's commentary. so you might turn around a document in 4 hours that use to take 4 days because youre using AI, but you still have to wait for opposing counsel comments. and you may not be able to move onto the next task until aforementioned task is finished. that doesn't mean alpha can't be generated with AI. AI opens up firm's capacity. just gotta watch out for when all the extra clients you take on have opposing counsel dump documents on you at the same time.
Nat McAleese@__nmca__

Q: what is the law firm equivalent of the bun rewrite?

English
1
0
17
3.4K