Most developers use AI today as a code generator layered on top of existing languages (Python, TypeScript, Go). That works—but it breaks down as soon as you move beyond snippets into autonomous systems:
• Non-deterministic behavior is hard to reason about
• Side effects are implicit and unsafe
• Verification is bolted on after the fact
• “AI safety” lives in policy, not semantics
If AI is writing, refactoring, and operating code, the more interesting question is:
If AI is doing all the coding, why are we using languages designed for humans?
A possible answer is AILANG: A Language Designed for AI Coders.
AILANG is a small, deterministic, functional language designed for AI-authored systems.
It focuses on:
• Determinism (pure by default)
• Explicit effects (IO, cloud, infra must be declared)
• Machine-legible structure (low token noise, stable ASTs)
• Verification-friendly semantics
This makes it suitable for autonomous agents that plan, execute, and audit real work—especially in production environments like Google Cloud.
Install via Gemini Extensions gemini extensions install https://github.com/sunholo-data/ailang_bootstrap.git
or Claude Code Plugins:
/plugin marketplace add sunholo-data/ailang_bootstrap
/plugin install ailang
Then ask your favourite AI to load up the ailang CLI (binary executable) and introduce the language to you. It comes with its own system prompt, examples, hello world and documentation. AILANG is intended to be 100% coded by AI, with you the human providing the direction and inspiration.
Website designed by AI as well! https://ailang.sunholo.com
Why This Might Interest You
Features aim to maximise the value of the iterative loop of AI coding (e.g. use less tokens than python to complete same task) and minimise search space on what is good (e.g. structured errors, static code verification AI can use to iterate quickly) - it also address concern around vibe coding for issues of security and compliance by baking in to the compiler verifiable paths:
• Effect typing instead of trust
Agents cannot touch infra, data, or APIs unless the effect is explicitly granted. Includes contracts and budget system (e.g. you can only call this API twice)
• Replayable, auditable execution
Deterministic programs → interpretable traces → explainable agent behavior.
• Designed for agents, not prompts
Syntax and semantics are optimized for LLM generation and self-refactoring.
• Run natively via ailang run, or compile it to llightning fast binary - currently supports Go (more languages on roadmap, but compiler is also written in Go)
AILANG is intended to integrate with existing services rather than replacing your stack. It is not trying to be “another general-purpose language.” or another "AI framework".
It is a coordination and reasoning layer for AI-operated systems.
Status & Call for Feedback
The project is live, open, and evolving:
• Language core implemented
• Go backend in production use for gaming engine, data pipelines and UI backends.
• Active work on verification and always-on agents - deployable to GCP for coordination across coding agents.
If you are working on:
• Autonomous agents
• Infra automation
• Policy-constrained systems
• AI safety via semantics (not filters)
I’d be interested in technical feedback to alpha test it and help get to v1.0.0 (currently v0.6.2)
Top comments (0)