Brendan (@Ruff_Brendan) header
Brendan (@Ruff_Brendan) profile picture

Commercial Litigation Attorney || Pacers Fan || Amateur Musician

Indiana, USA
Joined August 2015
Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)Photo by Brendan (@Ruff_Brendan)
We are picking up a corgi puppy today this is thrilling
2
1
1
27
Means the world to be back in the fight with my brothers. See y’all Friday, go ‘Cers🙏🏽
437
2,945
119
48,581
560,830
As practicing litigators know, the real gold standard is having to stand up in court and defend what you wrote. Opposing counsel isn’t going to grade your answer - they are going to try to take it apart. They’ll find the bad fact you glossed over, the case you stretched, and the concession you didn’t realize you made. That experience changes how you evaluate the next piece of work. My preferences are well and good and a worthwhile data point, but the more useful signal comes from what happens when the work is actually used.
The gold standard for evaluating complex legal work is partner review. Partners can cost over $3,000 an hour and associates $1,000, making expert review expensive at scale. Evaluating a model on LAB through expert review alone would cost millions of dollars. In practice, we combine rubric-based LLM scoring with sampled human preference. But rubrics miss errors beyond predefined criteria, and human reviewers get fatigued and make mistakes at scale. We built a generative reward model to bridge this gap by training agents to approximate partner review. We give these agents the original outputs, web search, and other tools to check our systems’ work. We find that their judgments correlate strongly with expert lawyer review. This approach will help us scale human review across model training and production products, and will be central to training Tenet 1.5.
8
121
26,617
Brendan retweeted
Retweet if you hate this baldheaded clown
19
432
12
832
15,255
Brendan retweeted
Reasons to use an em dash: 1) reclaim it from the machines 2) you don't know what a semicolon is for 3) it has incredible vibes
122
5,421
273
35,502
512,801
Brendan retweeted
Going up against pro se litigants who use AI
64
268
51
3,780
474,674
There is no one more excited about the general population feeling emboldened to DIY legal than lawyers
If you replace your transactional attorney with AI, call me in six (6) months when you need a litigator
13
66
8
733
61,570
Too many em dashes? What’s next? too much love? too much joy?
32
3,913
90
17,572
617,615
Beer should be free if you’re sad
73
2,260
112
22,656
828,567
I’m coming home, coming clean. Our brand new single “Pressure” is out worldwide June 27, 2025. Pre-save now at the link in our bio. #newmusic #independentartist
2
3
271
Ahem.... One space after periods. It's official.
151
175
98
1,500
177,107
2-1.
3
40
1
1,236
13,056
RT IF YOU GOT THE PACERS WINNING THEIR FIRST TITLE 🏆
172
5,714
146
12,718
787,145
RT @Pacers: it's our time, Indiana.
67
4
overrate that.
394
4,506
264
55,087
2,077,033
your Eastern Conference Champions 🏆
48
560
36
2,823
82,532
NBA Finals: Oklahoma City Thunder-Indiana Pacers. Game 1: Thursday June 5, 8:30 pm ET on ABC.
595
1,999
502
19,836
1,241,128
ITS A PARADE INSIDE MY CITY YEAAA
23
79
3
1,040
21,968
We Move! #YesCers
425
1,494
194
15,394
716,562
WHAT THE HALI⁉️ The Pacers take down the Bucks to advance to the conference semis 🔥
215
1,430
180
15,345
780,545
Not too bad of a way to send the Bucks home for the NBA's supposed most overrated player. Let them know Tyrese Haliburton!
1
19
2
312
12,606