A Silicon Valley start-up founded by former OpenAI researcher Diogo Almeida is attracting significant investor interest with its newly launched artificial intelligence tool, which aims to offer a more cost-effective alternative to leading large language models (LLMs) such as those behind ChatGPT and Claude. The company, TypeSafe AI, released its first product, Jev, last week and has already seen its valuation rise from $200 million to potential offers exceeding $10 billion.

Jev is designed primarily for software developers and focuses on automating programmatic, repetitive decision-making tasks rather than generating text, images, or explanations. The model specializes in “classifying” tasks within software applications, such as routing support tickets, approving or blocking requests, and underwriting insurance or credit risk. TypeSafe AI positions Jev as more efficient and significantly less expensive to operate compared to general-purpose LLMs, which are typically optimized for human interaction and natural language generation.

Almeida has explained that the concept for Jev originated during his time at OpenAI, when he questioned whether chatbots were the ideal tool for all AI applications. He noted that many AI calls might be machine-to-machine rather than human-facing, suggesting the potential for specialized models tailored to high-volume, straightforward decisions. Unlike LLMs that use substantial computing power to process complex chains of reasoning, Jev reportedly employs a probabilistic approach that quickly selects from a limited set of fixed responses. This method, similar to traditional deterministic machine learning systems, reduces computing demands and virtually eliminates hallucinations, where models generate inaccurate content.

James Hardiman, a general partner at DCVC, the venture capital firm that led TypeSafe’s most recent funding round, stated that the start-up is already profitable and highlighted the drastically reduced costs of its model, which he described as “orders of magnitude” cheaper than top-tier LLMs. The model’s name references Jevons paradox, an economic principle suggesting that greater efficiency can lead to increased overall consumption, underlining TypeSafe’s expectation that lower-cost AI could drive broader usage.

Despite the enthusiasm, some experts have raised questions about how novel Jev truly is. Anastasios Angelopoulos, CEO of Arena, an AI model evaluation platform, observed that the technology appears similar to existing zero-shot classifiers, which are standard tools in machine learning and not radically new. Furthermore, TypeSafe has maintained a degree of secrecy around the training methods behind Jev, limiting external assessments of its uniqueness and performance.

The start-up’s emergence reflects a broader concern in the tech industry about the expense of deploying AI solutions at scale. If TypeSafe’s specialized, cost-efficient approach finds wide adoption, it could pressure the business models of larger AI companies reliant on computationally intensive LLMs. The rapid investor interest and viral social media response suggest that the market is closely watching this potential shift within the AI landscape.