DeepSeek Code: The Ultimate Form of AI Programming — From Chatbox to Terminal

TL;DR: DeepSeek's funding surpasses 70 billion RMB, all-in on AI coding. DeepSeek-TUI, a terminal-based coding agent, hits 8,700 GitHub stars. A new Agent Harness team targets Claude Code. When AI programming evolves from "Q&A" to "on-site collaborator," the terminal becomes its home turf.

1. 70 Billion Rounds of Ammunition, All In on AI Coding

In May 2026, DeepSeek confirmed total funding exceeding 70 billion RMB. More importantly, leadership declared: full focus on breakthrough AI research, abandoning short-term commercialization.

The first target: AI programming. DeepSeek is building a Harness team, hiring Agent Harness PMs and engineers, explicitly benchmarking against Anthropic's Claude Code.

2. DeepSeek-TUI: The Terminal Coding Agent

DeepSeek-TUI is the vanguard. Built in Rust by developer Hunter Bown, this terminal-based coding agent has earned 8,700 GitHub stars.

The fundamental difference from chat-based AI coding:

• Chat AI: Like a remote consultant — you send a message, it returns a code snippet, you copy-paste
• Terminal Agent: Like an on-site collaborator — it reads files, understands code, executes commands, runs tests directly

DeepSeek-TUI capabilities: - Optimized for DeepSeek V4 - Terminal-native interaction with long context - Visible reasoning process - File editing, shell command execution, task management, sub-agent coordination - Fully local, extremely low cost

3. The Harness Architecture: Engineering Foundation for Coding Agents

The Harness architecture is the critical engineering framework in today's agent space, systematically addressing bare models' shortcomings in memory, code execution, and tool calling.

The technical progression of coding agents: 1. Bare model: Direct code generation from LLMs (most primitive) 2. IDE plugins: AI assistance embedded in editors (e.g., Cursor) 3. Terminal agents: AI directly operates in the engineering environment (e.g., Claude Code, DeepSeek-TUI) 4. Full Harness: Agent runtime with memory, toolchain, and execution environment (ultimate form)

DeepSeek is moving from step 3 to step 4.

4. Terminal Agents Need Terminal Hardware

The 24/7 operation requirement of terminal coding agents differs fundamentally from chat-based AI. You don't need to watch it constantly — it autonomously reads project files, runs compilations, executes tests, analyzes results, and only notifies you at critical decision points.

This means you need dedicated hardware independent of your dev machine to run coding agents continuously. This is exactly Kaihe AI Box's use case: low-power 24/7 operation, physically isolated from your main workstation, giving agents a stable work environment.

KaiheAiBox · AI Agent Tracker