70 Billion Raised! DeepSeek Code Is Coming, ACM Gold Medalist Cui Tianyi at the Helm
DeepSeek has raised 70 billion yuan and is developing its own AI coding product, DeepSeek Code. Senior researcher Deli Chen posted a recruitment for a Code Harness team, and former TSY Capital co-founder Cui Tianyi may lead the team.
Article intelligence
Key points
- DeepSeek raised 70 billion yuan, prioritizing AI research over commercialization.
- DeepSeek Code confirmed in development, hiring for Agent Harness team.
- ACM gold medalist Cui Tianyi likely to head the Harness team.
- Alternative tools like DeepSeek-TUI are available now.
Why it matters
This matters because deepSeek raised 70 billion yuan, prioritizing AI research over commercialization.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
DeepSeek has secured a massive 70 billion yuan in funding, with management signaling to potential investors that the priority remains breakthrough AI research rather than short-term monetization. Amid this financial milestone, the company's long-rumored AI coding tool, DeepSeek Code, is now confirmed to be in active development.
The confirmation came via a recruitment post by DeepSeek senior researcher Deli Chen, who explicitly called for building a "Code Harness" team from scratch. The post highlighted three keywords: Harness team, Code Harness, and DeepSeek Code. This aligns with the broader AI industry trend where the concept of "Harness"—a system that orchestrates, executes, and corrects AI agent workflows—has become a focal point as large language models approach performance ceilings.
Further lending credibility is the likely appointment of Cui Tianyi, former co-founder of TSY Capital, as the head of the Agent Harness team. Cui joined DeepSeek in March 2026 and brings an impressive background: a bachelor's degree in computer science from Zhejiang University, six gold medals at ACM ICPC Asia regional contests, and nine years at Jane Street in Hong Kong and New York. His transition from competitive programming to quantitative finance and now to AI agent development underscores DeepSeek's ambition to build a world-class coding assistant.
DeepSeek has been steadily ramping up hiring, with over ten positions listed in a single day, including roles for agent deep learning researchers and data strategy engineers. The company's own job postings describe the core concept as "Model + Harness = Agent," emphasizing the need for candidates deeply familiar with AI coding tools like Claude Code, Cursor, and GitHub Copilot, as well as agent frameworks like Manus and OpenClaw.
While DeepSeek Code is still in its formative stages and not yet available to the public, several alternatives exist. Notably, DeepSeek-TUI, a Rust-based terminal UI tool modeled after Claude Code, has gained 32.9k stars on GitHub. It supports file read/write, shell execution, web search, Git management, and sub-agent orchestration, all powered by DeepSeek V4. Additionally, the official awesome-deepseek-integration repository lists numerous code-related tools, including AIMatrices, a lightweight framework for building AI applications quickly.
In summary, DeepSeek is making a determined push into the AI coding assistant space, backed by substantial funding and top talent. Although the official product may take time to mature, the ecosystem of compatible tools is already expanding, offering developers a taste of what's to come.