Carl Fuda @FudaCarl

Want world class research capabilities, but don’t have the resources of a big lab? At our recent Sovereign AI event, @gabepereyra shared @harvey ’s “moneyball” approach. Here’s the playbook: 00:00 Introduction 00:37 Building a research lab on a budget 02:28 Legal Agent Bench, contracting, and the diligence dataset 03:57 Domain experts guiding synthetic data generation 05:23 Why Harvey open sourced its datasets 06:55 Working with the neo labs – and why more than one 08:20 Post-training in-house: building "Associate 1" 09:44 The model serving matrix: 60 countries, fallbacks, SLAs 11:05 Deciding what stays in production 12:29 Simple open source switches and model routing 13:55 Moneyball: "If we win on this budget, we change the game" 14:53 Q&A: Training with sensitive data 17:16 Q&A: Competing for research talent 18:46 Q&A: Designing rubrics that actually challenge frontier models 20:19 Q&A: Where the pipeline breaks — data, research, or infra 22:59 Q&A: The tension in open sourcing a benchmark 25:02 Q&A: Biggest remaining open problems 27:10 Q&A: Competing with horizontal products

When I was an undergrad at USC, I emailed @JeffDean inviting him to come play soccer with the club team after his talk on campus. To my surprise he ended up moving a meeting with the dean to come and play with us. What is truly special about Jeff is that despite being responsible for some of the most impressive engineering and AI work in history, he still finds time for everyone - even if it means taking a break from inventing tensorflow to kick a ball around with an undergrad who looked up to him. I was fortunate to see some of the great work Jeff, @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix did at Brain and DeepMind. They are truly unique - so excited to see what this team can build.

Jeff Dean @JeffDean

a week ago

Announcing Discovery Loop! I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine

keyboard_arrow_left Previous keyboard_arrow_right Next

897 2K 22K 6.5M 6K

Command Center is now available to all Harvey customers. See adoption across your organization, ask questions over your own data, and turn the answers into leadership-ready reports in minutes. Then manage new releases and permissions, all in one place.

Thanks to @ItsJulioPereyra and Harvey’s benchmarking team for improving the Legal Agent Benchmark (LAB), preparing it for retrieval-augmented generation (RAG) development, and continuing to support open-source legal AI research and development!

Julio Pereyra @ItsJulioPereyra

2 weeks ago

@Overfit_Dicta Hey we have some new docs for you! github.com/harveyai/harve… One of the main reasons we open-sourced LAB is to get public scrutiny and feedback and we’ve loved to see people dig in and find things we could improve. Thanks for helping us continue to make LAB better.

1 1 6 4K 2

Harvey is now AIUC-1 certified.

Grok has always been very strong on law

Tech Dev Notes @techdevnotes

a month ago

Grok 4.5 is #1 on Harvey's Legal Agent Benchmark

53 71 649 1.6M 103

Q2 recap for @harvey - +$100M NNARR - 53% DAU/MAU Key hires (including Q1) - Anique (CPO) - prev VP of Product at Rippling - Rachel (CMO) - prev CMO at Notion - Brooks (CISO) - prev CISO at Roblox - Keith (CSO) - prev CPO at Google Product - Agent unification - cloud agents can use all Harvey product surfaces - Command center (EA) - monitor adoption and ROI by use case - Contract intelligence (EA) - agentic contracting platform for enterprises Eng - Migration to cloud agent infrastructure - Integrating open source inference providers - Scaling document processing (54TB / week) AI - Legal Agent Bench - Open source post training - Published multiple research directions with partners We invested heavily in cloud agent infrastructure at the end of last year and in Q1. In Q2 we also unified many of our product surfaces (collapsed as @winstonweinberg says) by making them all tools accessible by our cloud agents. Prior to this, there were a lot of capabilities in Harvey that were often only discovered by power users. As cloud agents get better and our product becomes more connected we are seeing users discover more of the product by learning from their agents (see plot of product surfaces per user).

keyboard_arrow_left Previous keyboard_arrow_right Next

We're training the first in a series of legal foundation models. Harvey Labs is hiring across the post-training, agent, and data stack. We're also interested in acquiring teams and neolabs with talent in these areas. DM @gabepereyra if you're excited by frontier legal AI.

Gabe Pereyra @gabepereyra

2 months ago

Model strategy for @harvey: We are working on the first model in our legal foundation model series, inspired by @cursor_ai's Composer. Two goals: 1. Allow us to serve frontier intelligence across our product surface areas at an affordable price and a strong security posture.

66 69 935 279K 943

Anthropic is also the first model provider to feature @harvey’s Legal Agent Bench as part of their core go-live benchmarks. Glad to see legal becoming a first-class use case for AI.

Niko @nikogrupen

2 months ago

@AnthropicAI released a new model family today, Fable 5, which they’re describing as a Mythos class model with safeguards. @harvey had early access and it posted all-time high scores on both Legal Agent Bench (13.3%) and BigLaw Bench (93.4%). Due to data retention

0 2 10 13K 2

Harvey is integrating with @Intralinks, the AI-enabled dealmaking platform for high-stakes financial transactions. The integration expands Harvey's growing ecosystem of deal and enterprise data connectors.

PitchBook data, directly in Harvey. With our @PitchBook integration, bring private markets data into the same workspace where you draft, analyze, and review: from company profiles and deal history to fund, investor, and market benchmark data.

I actually think this is directionally correct. Tech companies are starting to spend the same on tokens as they do on eng headcount. Other knowledge work industries will follow with some delay. Kirkland spends billions on lawyers so it’s not crazy to assume they will spend a similar amount on tokens in the future. Most companies are starting to realize they need to own their AI. As models get better law firms will need to better articulate their value to in-house teams using codex / claude for legal. Kirkland is sitting on one of the most valuable AI datasets in the world (all their associate and partner feedback and trajectories). The cost of replicating this dataset is essentially Kirkland’s yearly labor costs. I’m very bullish they can convert this into AI differentiation above the foundation models. There’s no need to build everything from scratch like a lot of the commentary assumes. The infrastructure for building / training / deploying agents has matured significantly. It’s very possible for a firm like Kirkland to use inference providers and rl-as-a-service companies to leverage their valuable data without needing to hire OpenAI level talent. We’ve already seen promising results doing this for similar firms and post training models on lawyer feedback. We think the best future for legal is one where every law firm is able to convert their unique expertise into their own AI and use that to provide their clients with exceptional service. We are starting to invest heavily in training our own models and helping firms do the same. Excited to see a top firm like Kirkland taking a big bet like this.

Techmeme @Techmeme

3 months ago

Kirkland & Ellis, the world's highest-grossing law firm, is setting aside $500M to build its own AI platform rather than rely on tools available to its rivals (Financial Times) (Visit Techmeme dot com for the link and full context!)

70 66 1K 1.6M 467

@AnthropicAI launched Claude Opus 4.8 today. Legal Agent Benchmark is the newest knowledge work benchmark included in their public model card -- awesome to see frontier labs making legal intelligence first-class! Check out the results below:

Claude @claudeai

3 months ago

Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors. Available today at the same price.

4K 8K 67K 15.4M 8K

x.com/i/article/2051…

Now live in Harvey: Mistral. Our platform brings together leading foundation models and routes work to the right one for each task. Today we’re teaming up with @MistralAI to make their models available in Harvey, starting with EU customers.

Heading into the long weekend like…

The most powerful agent products will be the ones that effectively marry offline agents -- to build and maintain knowledge bases / indexes / datasets -- with online agents that complete work over those knowledge bases. Contract intelligence is one of our first explorations of what a production, enterprise-grade instantiation of this idea looks like.

Harvey @harvey

3 months ago

Introducing Contract Intelligence. Contract agents take first pass on inbound contracts, apply your playbooks, and generate redlines. See how clauses and negotiated positions are trending across your business. Escalate what needs legal judgment, while agents handle the rest.

8 19 229 62K 154

Over two-thirds of the AmLaw 100 and Vault 100 now use Harvey. Grateful to our customers and excited for major product announcements this week.

Five months ago, I argued against the President's $4 trillion tariffs at the Supreme Court. In 237 years, the Court had never struck down a sitting President's signature initiative. Legal scholars said it was impossible. Some of my own colleagues said it was impossible. We won. 6-3. But the real story isn't what happened in that courtroom. It's what happened in the months before. And its the subject of my TED talk, coming out tomorrow. I had the best legal team in the nation, especially Colleen Roh Sinzdak, the most outstanding legal strategist I know. Huge thanks, too, go to the Liberty Justice Center (and in particular its fearless and hyper-intelligent leader Sara Albrecht), who organized the client small businesses, as well as to the brave small businesses themselves. I also had four teachers preparing me. A mindset coach who'd worked with Andre Agassi. An improv coach who taught me that "Yes, and" works in Supreme Court arguments the same way it works everywhere else. A meditation coach who taught me stillness. And Harvey. Harvey predicted many of the questions the Justices asked — sometimes almost word for word. Brilliant. Tireless. Occasionally insufferable. Here's the catch: Harvey isn't a person. Harvey is a bespoke AI I built over the last year with a legal AI company, trained on every question every Justice has asked in oral argument for 25 years, and everything they've ever written. Tomorrow, TED releases my talk about what really happened — and what I learned standing at that podium. AI can predict. AI can analyze. What AI cannot do is the one thing that actually won the argument. Connect. Read the room. Hear not just a Justice's words, but her worry — and answer the worry. That is the irreducibly human skill. Find yours. Go deeper. In this age of AI, that's where your edge lives. The talk goes live Thursday, May 7 at 11am ET: go.ted.com/nealkumarkatyal What's the irreducibly human skill in your work — the thing AI can't touch?

keyboard_arrow_left Previous keyboard_arrow_right Next

Congrats to Anthropic on this launch. For Harvey's long-horizon agents to deliver the best work product, they need reliable agent infrastructure that enables better memory and multi-agent orchestration. We're proud to partner with Anthropic on this work.

Claude @claudeai

3 months ago

Live from Code with Claude: we're launching dreaming in Claude Managed Agents as a research preview. Outcomes, multiagent orchestration, and webhooks are now in public beta.

558 1K 15K 3.4M 8K