Industry / FundingXifeng2026-04-14
View Original

Li Di's Harness Agent Startup Secures Major Backing from Kai-Fu Lee and Qi Lu

Original: Quantum Bit · Xifeng · 2026-04-14 Category: Industry / Funding


Overview

According to Quantum Bit, a harness agent startup led by Li Di completed two rounds of financing within 4 months, securing a 3-5 year financial runway. Notably, Kai-Fu Lee and Qi Lu both invested heavily in the company — a rare convergence of two of the most influential figures in the AI space betting on the same target.

Why This Matters for Harness Engineering

The signal from this funding far outweighs the dollar amount:

1. Harness Has Evolved from Technical Concept to Investment Thesis

The term "Harness Engineering" was formally introduced by OpenAI's Codex team in early 2026. Within just 2-3 months, it evolved from a technical blog concept into a startup category attracting significant capital. This is one of the fastest concept-to-funding cycles in AI Agent infrastructure.

2. Consensus Among Top Investors

Kai-Fu Lee (Innovation Works / 01.AI) and Qi Lu (MiraclePlus) represent two distinct investment perspectives — AI application layer and infrastructure layer respectively. Both backing the same Harness company suggests strong market consensus on "Agent Runtime" as a foundational infrastructure layer.

3. 3-5 Year Runway = Long-Term Infrastructure Bet

This isn't 12-month fast money, but 3-5 years of reserves. Both investors and the team view Harness not as a hype-driven project, but as an infrastructure layer requiring long-term commitment.

The Landscape Context

This company's emergence further validates that the Harness Engineering ecosystem is rapidly stratifying:

Layer Representative Projects Positioning
Open-Source Frameworks OpenClaw, Cline, Aider, OpenCode Community-driven Harness Runtimes
Commercial Platforms Devin, Cursor, Windsurf Developer-facing commercial Agent IDEs
Infrastructure Li Di's team (this article), CREAO AI Enterprise-facing Agent infrastructure
Ecosystem Standards Harness Engineering Guide Open-source knowledge systems

Li Di previously led Microsoft Xiaoice (later spun off as an independent company) and has deep experience in conversational AI and multimodal interaction. Moving from conversational AI to Agent Harness infrastructure is a natural technical evolution.

Key Takeaway

Harness Engineering is no longer just an engineering practice — it's becoming a startup category with its own investment logic. When Kai-Fu Lee and Qi Lu both place their bets, it typically means a direction has shifted from "early signal" to "inevitable trend."


Editor's note: Based on Quantum Bit's April 14, 2026 report. See original article for details.