Conway Research Secured Funding for Local AI

The startup attracted major venture capital to develop on-device AI assistants running on Mac and iPhone hardware.

Updated on Oct. 2, 2026 in Artificial Intelligence

Isometric editorial illustration featuring a central metallic computer chip inside a geometric housing, representing localized artificial intelligence hardware. — Conway Research has raised a new round of venture capital to scale its local AI assistant, Underdog, which runs exclusively on Apple hardware. AI Illustration. Upload story photo >

Live Poll

Do you trust AI assistants more when they run entirely on your own personal devices?

Conway Research has secured a fresh round of investment from Andreessen Horowitz, Khosla Ventures, and Hummingbird VC. The company simultaneously launched an AI assistant called Underdog that processes data locally on Apple hardware.

Why it matters

By executing models directly on user hardware, Conway Research aims to eliminate privacy risks associated with cloud processing and shift compute costs away from providers. This approach leverages the power of Apple Silicon to maintain data security without external server dependency.

The firm's Woof model features 4 billion parameters while occupying less than 2.5 GB of storage. Additionally, the system reaches processing speeds of 730 tokens per second on MacBook devices.

The players

Conway Research

This startup focuses on developing local AI inference engines designed to function directly on consumer hardware.

Sigil Wen

He is the founder of Conway Research and a former Thiel Fellow who gained industry attention for his viral essay on Web 4.0.

Andreessen Horowitz

It is a prominent venture capital firm that provides funding to technology companies and is a lead investor in Conway Research.

Apple

It provides the hardware ecosystem and Silicon architecture that powers the local AI models developed by the startup.

The details

Founded by Sigil Wen in 2026, the company utilizes custom inference engines for Apple Silicon to facilitate local data processing. Its flagship Underdog 27B model reportedly outperforms Claude Opus 4.6 on specific internal benchmarks.

2025: Sigil Wen served as a Thiel Fellow.
February 2026: Sigil Wen published a widely read Web 4.0 essay.
2026: Conway Research was founded.
October 2, 2026: Sigil Wen announced the successful funding round.

The Tech Race

The move toward local processing represents a challenge to the established dominance of massive cloud-based LLM architectures. By optimizing software for specific chipsets, developers are increasingly positioning their tools as high-speed alternatives to centralized, server-side AI models.

Users can expect faster AI responsiveness and improved data privacy since information remains on the device instead of being sent to external servers. This deployment model may allow consumers to run sophisticated AI tools on existing Mac and iPhone hardware without requiring a subscription-based cloud service.

The takeaway

The trend toward local inference suggests that the next generation of AI could be defined by hardware optimization rather than raw server-side compute. Consumers prioritizing data privacy should look for tools that keep processing confined to their own personal devices.