heading back to NY from SF - quick observations
1) Basically all the AI companies are starting to converge on the same spaces within their category. Vibe is way more cutthroat than collaborative now.
2) Was trying to explain to people that most of my friends in NY don't use LLMs at all, except maybe chatGPT to help write emails. Many didn't believe me. The tech vibe has never been more different between the two ciites.
3) Everyone seems much more hyper aware about money everyone else is making now. It does remind me of finance in New York where personal net worth is now very explicitly a driving force in people's decisions (vs. before where it was more veiled).
4) It is insane how expensive everything - talking to some people about housing market/childcare there made me gasp outloud. 1.5x-2x of even NY costs for something similar/
5) Agent swarm stuff seems to have increased doomerism amongst people who previously were pretty optimistic about AI in the last year
6) Like everyone seems to want to start a data labeling company to make a quick bag lol
7) very anecdotal based on a small number of convos, but I was surprised that some people are actively pushing out life milestones they wanted to do earlier (e.g. find a partner, have a kid, etc) because they think the AI window is short right now and they want to make sure they prioritize it
(personally I don't think those things are mutually exclusive but hard to convince people of that)
There is an ENDLESS supply of intelligent people on the internet explaining why every opportunity is dead, every successful person got lucky, and everyone trying something new is too late, wrong or dumb.
Don't listen.
You deserve to share your ideas with the world.
For years, the advice was to move to San Francisco if you wanted a career in tech. That advice is now wrong. New York has become the most exciting place in the world to build.
A few reasons why:
1. The AI labs are here. OpenAI, Anthropic, Google, and hundreds of others all run meaningful operations in NYC. You can now build a serious frontier-tech career here without giving up anything.
2. Every major industry lives here, and so does its AI counterpart. New York doesn't only have "tech," but finance, media, advertising, fashion, healthcare, real estate, law, and commerce too, with their applied-AI layer counterparts being built on top of each one.
3. The talent pool is unlike anywhere else. SF gives you engineers. New York gives you engineers, plus world-class designers, salespeople, bankers, marketers, operators, creatives, and media people, all at the top of their fields. Companies will need all of them.
4. People actually want to live here. New York is unlike any other city with its density of restaurants, art, nightlife, fashion, and media - and the people working across these sectors.
If you're building something and you're more than just tech, I'd argue there's only one city for you right now.
Introducing Tenet, our first model post-trained for legal.
Tenet is a Kimi K3 base that we post-trained with @FireworksAI_HQ on a corpus of publicly available legal data, synthetic data, and human expert data simulating long-horizon legal work.
Training increases Tenet's all-pass rate by 82% on LAB and 22% on LAB Contracts relative to the Kimi K3 base model. It achieves state-of-the-art performance on LAB Contracts and places second on LAB.
These gains generalize to other leading agentic benchmarks including @mercor's Apex Agents - Corporate Law, @crosbylegal's Redline Bench, and @scale_AI's Professional Reasoning Bench.
Tenet is also optimized for token efficiency, operating at less than a fourth the cost of leading foundation models.
We additionally post-trained three specialist models for Tenet to use as subagents:
1) M&A Diligence: post-trained with @baseten on our LAB Diligence environment in an RLM harness, this model is optimized for high-scale, long-horizon tasks.
2) Review Tables: trained with @appliedcompute on our Review Table environment, this model is state-of-the-art and cost-effective at high-volume document review and structured data extraction.
3) Firm Knowledge: trained with @EngramLab on our synthetic law firm environment, this model is optimized to learn and search over a firm's knowledge via memory and structured notes.
More details on model training, environment design, benchmarking, results, and more in the article by @gabepereyra below.
What's next for Harvey’s research?
- Scaling LAB to more jurisdictions, practice areas and workflows
- Scaling compute to bring new generalist models and capabilities to Harvey
More to come soon.
Introducing Tenet, our first model post-trained for legal.
Tenet is a Kimi K3 base that we post-trained with @FireworksAI_HQ on a corpus of publicly available legal data, synthetic data, and human expert data simulating long-horizon legal work.
Training increases Tenet's all-pass rate by 82% on LAB and 22% on LAB Contracts relative to the Kimi K3 base model. It achieves state-of-the-art performance on LAB Contracts and places second on LAB.
These gains generalize to other leading agentic benchmarks including @mercor's Apex Agents - Corporate Law, @crosbylegal's Redline Bench, and @scale_AI's Professional Reasoning Bench.
Tenet is also optimized for token efficiency, operating at less than a fourth the cost of leading foundation models.
We additionally post-trained three specialist models for Tenet to use as subagents:
1) M&A Diligence: post-trained with @baseten on our LAB Diligence environment in an RLM harness, this model is optimized for high-scale, long-horizon tasks.
2) Review Tables: trained with @appliedcompute on our Review Table environment, this model is state-of-the-art and cost-effective at high-volume document review and structured data extraction.
3) Firm Knowledge: trained with @EngramLab on our synthetic law firm environment, this model is optimized to learn and search over a firm's knowledge via memory and structured notes.
More details on model training, environment design, benchmarking, results, and more in the article by @gabepereyra below.
What's next for Harvey’s research?
- Scaling LAB to more jurisdictions, practice areas and workflows
- Scaling compute to bring new generalist models and capabilities to Harvey
More to come soon.