Show HN: Ante, a coding agent in a single binary that runs offline
Ante presents a self-contained, offline coding agent in a tiny Rust binary, promising impressive resource efficiency and model-agnostic operation. HN's interest is piqued by its technical design and bold claims for an 'agentic era' tool. However, the decision to ship a closed-source binary with opt-out telemetry sparks significant debate among the security-conscious community.
The Lowdown
Ante is a new coding agent from Antigma Labs, designed to be a lightweight, self-organizing "ghost in your shell." It’s distributed as a compact ~15MB Rust binary, offering full offline capability by integrating a managed llama.cpp engine for local inference. The project distinguishes itself by focusing on the "harness, not the model," aiming for an agent that is verifiable, affordable, and runnable anywhere, forming the basis for "cellular-native" agents.
- Resource Efficiency: Ante claims significantly lower resource consumption compared to commercial alternatives like Claude Code, using up to 7x less peak memory and 9x less CPU.
- Offline Operation: It natively supports local LLM inference via embedded
llama.cppand GGUF models, requiring no internet connection or API keys for core functionality. - Model Agnostic: Users can bring their own API keys or local models, switching between over 12 providers, including OpenAI, Anthropic, and various local GGUF models.
- Modular and Extensible: The agent supports multi-agent orchestration, custom skills, persistent memory, and various operating modes like interactive TUI, headless scripting, server daemon, and gateway for bots.
- Alpha Preview & Source Code: Currently in alpha, the core harness is shipped as a prebuilt binary, with the developers planning to progressively open-source components and address concerns around "how open source should work in the agentic era."
In essence, Ante positions itself as a robust, efficient foundation for the next generation of AI agents, emphasizing performance and local control. While its technical ambitions are clear, its current closed-source nature and default telemetry settings have drawn critical scrutiny from the Hacker News community, highlighting a tension between rapid development and community expectations for transparency and control in the AI tools space.
The Gossip
Source Code Scrutiny: "Binary Blob" Brouhaha
A dominant concern among commenters is the distribution of Ante as a prebuilt, closed-source binary for a tool that requires significant system access. Users expressed reluctance to grant "god access" without the ability to review the source code or compile it themselves, citing supply chain security risks and a desire for modification freedom. The author acknowledged these concerns, stating they are working on shipping source code and addressing the complexities of open-sourcing in the "agentic era."
Telemetry Troubles: Opt-Out Outrage
Many users were strongly critical of Ante's opt-out telemetry, finding it contradictory to the product's emphasis on offline functionality and local control. Commenters felt that telemetry, especially for a tool with potential access to sensitive code, should be strictly opt-in. The author `ubermon` received this feedback positively, noting it was a carry-over from preview development and implying changes might be considered.
Defining Development: AI's Role in "Building"
A lively philosophical debate arose regarding whether using an AI agent to generate code or build projects still constitutes "building it yourself." Analogies ranged from hammers to ghostwriters, carpenters, Roomba vacuums, and even goat farming. Some argued AI is merely a tool, like any other, while others contended that complex AI agents are more akin to skilled labor or co-creators, potentially diluting the user's claim to authorship. This discussion explored the changing nature of creative and technical work with advanced AI assistance.