/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Harvey, Abridge, Ramp, Rogo, and other AI startups are embracing open-weight models or training their own models to reduce expensive reliance on frontier labs

The $15.6 billion legal startup Harvey built its business around training AI models like OpenAI's GPT-4 to do specialized work for lawyers.

Bloomberg

Context & Ripple Effects

Harvey began as a legal-AI company built around generative models, raising a $21 million Series A backed by Sequoia and the OpenAI Startup Fund in 2023. Its August launch of Harvey Tenet, an in-house legal model, made the shift from using general-purpose models to controlling more of the model stack concrete.

The move comes as investors scrutinize whether open-weight model developers can turn technical interest into durable revenue, as detailed in coverage of the US open-weight model ecosystem. Harvey, Abridge, Ramp, and Rogo give that ecosystem an important customer-side rationale: lowering the cost and dependency associated with frontier-model APIs.

First-order effects

  • Harvey, Abridge, Ramp, and Rogo can reduce their exposure to frontier-lab model costs by running open-weight models or training models tailored to their own workflows.
  • Harvey's Tenet strategy gives its legal product a proprietary model layer rather than relying solely on GPT-4-class external models.

Second-order effects

  • Frontier labs face pressure to justify premium API pricing and differentiated capabilities when specialized software companies can substitute internally trained or open-weight alternatives for parts of their workloads.
  • Open-weight model providers gain a clearer route to adoption through vertical software companies that need lower-cost, controllable inference rather than a general-purpose model endpoint.

Third-order effects

  • If vertical AI vendors keep moving model development in-house, competitive advantage shifts from access to a frontier model toward proprietary training data, workflow integration, and the ability to operate models economically.
  • The open-weight ecosystem's commercial test becomes less about selling models directly and more about enabling software companies to own strategic parts of their AI stack.

The trend: Vertical AI companies are bringing model capability closer to the application layer to reduce frontier-lab dependence and protect unit economics.

Discussion

  • @jkubicki Josh Kubicki on x
    As will law firms to cut reliance on Harvey and Legora.
  • @rebeccatorrenc5 Rebecca Torrence on x
    Harvey's AI costs got so high its gross margins plunged from 50% to -50% in 6 months. Its response: build its own model using open-weight AI. Startups from Abridge to Rogo are following suit to cut costs and reduce their reliance on OpenAI and Anthropic. My latest w/ @nmasc_👇
  • @tedunderwood.com Ted Underwood on bluesky
    Another story about institutions deciding they can build on open models rather than pay premium to a frontier lab. h/t @ver.ooo
  • @sungkim Sung Kim on bluesky
    It's a bit funny that Harvey, a legal AI company that offered seat based pricing, saw its gross margin go from 50% to negative 50% as customers actually started using the product and agentic AI began consuming more and more tokens.  —  www.bloomberg.com/news/article...