Wordless AI Startup TypeSafe Raises $870M at $7.5B Valuation
TypeSafe just raised $870M for Jev, an AI model that outputs probabilities instead of text to automate enterprise workflows instantly.
Funding · Source: TechCrunch
What happened
TypeSafe AI just closed an $870 million Series A funding round at a reported $7.5 billion valuation. Andreessen Horowitz led the massive fundraise. Sequoia Capital, DCVC, and other angel investors also joined the round. This capital injection comes just weeks after the startup launched its flagship model, Jev, on September 15. The model went viral almost instantly.
Jev is built on a transformer architecture, but it refuses to write a single word. Instead of generating natural language text or code, it outputs probabilities. The company calls these calibrated decisions. When an application asks Jev a question, the model responds with a simple yes or no, a selection from a provided list, or a specific score. The model also provides a confidence number with its answers. This helps applications mitigate the impact of AI hallucinations.
The enterprise adoption rate is staggering. TypeSafe claims a third of Fortune 500 companies are already using Jev. Co-founded in 2024 by former OpenAI researcher Diogo Almeida, former Meta engineer Sasha Sheng, and entrepreneur Erik Gafni, the team built a custom architecture. Almeida previously noted that human language is not useful for automation because computers speak a different language. TypeSafe trained Jev using a new method called reinforcement learning for calibrated decisions.
Key facts
- $870 million — Series A funding raised by TypeSafe AI
- $7.5 billion — Reported valuation of the company
- 1/3 — Share of Fortune 500 companies reportedly using Jev
- 700 milliseconds — Time it takes Jev to process user requests
- 200x — How much faster Jev is compared to some frontier LLMs
Why it matters
The era of parsing chatbot output to trigger software actions is ending. Large language models are great at talking to humans. However, computers need structured decisions. When an enterprise application uses a traditional language model, developers must write extra code to condense the natural language text into a standardized format. Jev skips the text generation phase entirely. By outputting structured data directly, it removes the need for complex data preparation. This simplifies the code base and drastically reduces the risk of software errors.
The second-order effect is a massive drop in latency and compute costs for enterprise automation. TypeSafe claims Jev processes requests in under 700 milliseconds. This custom technology makes the model up to 200 times faster than some frontier language models. It is also reportedly up to 100 times more cost-efficient. When artificial intelligence becomes this cheap and fast, companies can afford to put a decision engine behind every single micro-interaction. Developers can configure Jev to rate cybersecurity alerts based on severity or quantify the urgency of a support ticket instantly.
For builders
Stop parsing text for software automation
If you are building workflows, stop forcing language models to output natural language. Switch to models designed for calibrated decisions to cut latency and reduce parsing errors. Developers who cling to text generation for backend tasks will lose to faster competitors.
Use confidence scores to handle hallucinations
Jev outputs a confidence number alongside its structured decisions. Builders must use these metrics to build reliable fallback mechanisms. You can automatically route low-confidence automated decisions to human operators while letting high-confidence tasks run instantly.
Cheaper compute unlocks high-volume tasks
With costs reportedly 100 times lower than frontier models, high-volume sorting tasks are now economically viable. Founders can build highly profitable products around high-frequency data tagging, routing, and scoring. Enterprises will pay a premium for reliable automation at scale.
My take
We spent years teaching machines to talk like humans, but computers do not care about poetry. TypeSafe realized that enterprise software just needs a fast, reliable answer. Building an artificial intelligence that stays completely silent is the smartest pivot in the industry right now. The future of automation is not chatbots, it is invisible decision engines.
Original reporting: TechCrunch. This is my rewrite and opinion.