cto at @WeAreLegora and hiring big time
🇸🇪
Joined March 2022
- Tweets75
- Following373
- Followers763
- Likes87
this is so good, esp when we get type info on .sql files 🥳
what @gustaf said! We're looking for more ambitious engineers, designers, and product people in NYC, London, and Stockholm!
I just stepped out of Legora’s NYC office. Hard to describe how special it is to see hundreds of people working on something that Max, Sigge, and Jake once only had each other to believe in. They will win this and everyone who doesn't know yet will know soon.
And they're hiring for all roles in NYC but especially engineers and product. legora.com/careers
If you’re going through a door 10m in front of me just leave it. Please don’t make me awkward semi-run while you hold it
When hiring, I always look for ambition. Doesn’t matter what, but they should be - or want to be - excellent at something.
This is not what I meant @bryan_johnson
Breaking news… my boners are now longer than 4 hrs
Longer than
> Titanic
> Lord of the Rings (extended edition)
Stats:
> 4 hr 2 min
> 5 episodes
> 93% strength
Out of database of 55,000 measurements:
> 99th percentile among all males
> 93rd percentile among men 20-25 years of age
> 100% percentile among men my age
> only 230 sessions exceed 4 hrs
Nighttime erections (NTE) are a tier 1 longevity biomarker as critical to systemic health as VO2 max, resting heart rate, HRV, and blood pressure.
Women have them nightly too.
NTEs decline with age. Mine have been getting stronger over the past 18 months. This is a personal best.
Low night time erection scores mean the following:
> 2x risk of heart attack and stroke in next 4 years
> early warning for cardiovascular problems
> poor nervous system regulation
Protocol for strong NTEs:
> high sleep quality
> cardiovascular fitness
> tadalafil 5 mg - I take this longevity medication daily
> calm nervous system
Jacob Lauritzen retweeted
if I was Dumbledore I would've hid the philosopher's stone in the aws console
evals evals evals
The Legora BAR is back with a new update. Since launching the Benchmark for Agentic Reasoning, we’ve continued expanding it and testing the latest frontier models against the complex, end-to-end legal work our customers do in Legora. The goal remains the same: understand what actually improves legal work as models and our harness evolve.
On short tasks the field is essentially flat, with nearly every model within 3% of the average. The separation happens on long cases, where the spread runs from 0.78x to 1.15x. Opus 5 and Fable 5.1 lead the long tier, with Muse Spark 1.3 and Gemini 3.8 Flash close behind at around 1.09x to 1.10x.
As the landscape gets more competitive between different models, the trade-offs are becoming just as important as raw quality. We build the BAR to make those differences visible, both for us and for our customers.
Thanks @etnshow for having me on! And thanks anthropic for putting me in a hotel room with a nice view
Law firms buying GPUs to train their own AI models are solving the wrong problem according to @WeAreLegora CTO @jacsebl:
"[Law firms] want to control their IP. They want to build a durable asset. They want to build a system of knowledge that's self-improving over time and become sort of their unique IP."
"Why do you go to Firm X? They have this combined knowledge of 10,000 lawyers over many years."
""If you have to get data engineers and try to fine-tune the data in there, it's really difficult to update."
"[Instead] You can build a system in artifacts. You can build it in skills, and you can build a knowledge graph, and you can build memory systems and personalisation."
"If it's artifact, you have knowledge managers and people from the firm that can look through the ontology and the knowledge graph or the skills and say, "Yeah, this is right. We like this. We don't like this." And they can edit it. They can fine tweak it super easily."
"If you have an agent that produces something, you can say, "This was wrong. Please don't do that again." And it will just update it, and then it's better next time."
Jacob Lauritzen retweeted
Law firms buying GPUs to train their own AI models are solving the wrong problem according to @WeAreLegora CTO @jacsebl:
"[Law firms] want to control their IP. They want to build a durable asset. They want to build a system of knowledge that's self-improving over time and become sort of their unique IP."
"Why do you go to Firm X? They have this combined knowledge of 10,000 lawyers over many years."
""If you have to get data engineers and try to fine-tune the data in there, it's really difficult to update."
"[Instead] You can build a system in artifacts. You can build it in skills, and you can build a knowledge graph, and you can build memory systems and personalisation."
"If it's artifact, you have knowledge managers and people from the firm that can look through the ontology and the knowledge graph or the skills and say, "Yeah, this is right. We like this. We don't like this." And they can edit it. They can fine tweak it super easily."
"If you have an agent that produces something, you can say, "This was wrong. Please don't do that again." And it will just update it, and then it's better next time."
The Jev model direction makes so much sense that it almost seems obvious in retrospect. A lot of llm usage today is purely decision making/picking from options, not writing. Forcing the same kind of model to do both types of work seems suboptimal
