惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

酷 壳 – CoolShell
酷 壳 – CoolShell
D
DataBreaches.Net
C
Check Point Blog
雷峰网
雷峰网
小众软件
小众软件
GbyAI
GbyAI
美团技术团队
P
Proofpoint News Feed
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
罗磊的独立博客
大猫的无限游戏
大猫的无限游戏
WordPress大学
WordPress大学
MyScale Blog
MyScale Blog
The Cloudflare Blog
阮一峰的网络日志
阮一峰的网络日志
Apple Machine Learning Research
Apple Machine Learning Research
Y
Y Combinator Blog
Jina AI
Jina AI
爱范儿
爱范儿
Last Week in AI
Last Week in AI
MongoDB | Blog
MongoDB | Blog
I
InfoQ
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
博客园 - 司徒正美

Hacker News: Show HN

PurrrrrFocus: Pomodoro Timer App - App Store Workflow Engine — Multi-Step Orchestration for Bun RapidPhoto: Pro Photo Editor App - App Store GitHub - DheerG/swarms: Achieve extraordinary results with claude code across a variety of tasks SPICE simulation → oscilloscope → verification with Claude Code — Lucas Gerads Show HN: VCoding – A 5 MB native Windows IDE with no dynamic dependencies Show HN: LLMs don't hallucinate because they're bad at math, it's the format GitHub - Agent-FM/agentfm-core: AgentFM is a peer-to-peer network that turns everyday computers into a decentralized AI supercomputer. AgentFM lets you run massive AI workloads directly across a global mesh of idle CPUs and GPUs. Show HN: Tracking Top US Science Olympiad Alumni over Last 25 Years GitHub - Potarix/agent-hub: One place to talk to all your agents Show HN: Runtime security for AI agents(injection,tool abuse, data exfiltration) GitHub - dubeyKartikay/lazyspotify: Terminal Spotify client for macOS and Linux GitHub - the-banana-tool/king-louie: Easy to use GUI Personal AI Assistant. Win/Linux/Mac. Show HN I made my vacation rental bookable by AI agents–no Airbnb, 0% commission GitHub - basteez/jsf-autoreload: maven plugin to enable hot reload on jsf projects uvm32/hosts/host-gdbstub at main · ringtailsoftware/uvm32 GitHub - labsai/EDDI: Config-driven engine that turns JSON into production-grade AI agents. Multi-agent orchestration, 12+ LLM providers, MCP/A2A protocols, RAG, persistent memory, and enterprise compliance (EU AI Act, GDPR, HIPAA). Built on Quarkus. GitHub - glitchnsec/fortyone-oss: AI Executive Assistant Platform Quickstart | Alien GitHub - muxshed/shed: One stream in, or many. Every destination, simultaneously. No cloud middleman, no per-channel fees, no limits. GitHub - ocrbase-hq/ocrbase: 📄 PDF/IMG ->.MD/JSON Document OCR API for PaddleOCR and GLMOCR. Self-hostable. GitHub - impactjo/home-memory: MCP server that lets your AI assistant remember everything about your home. GitHub - Sets88/dbcls: DbCls is a powerful terminal database client that supports various databases GitHub - neptun2000/heor-agent-mcp GitHub - SeanFDZ/macmind: Single-layer transformer in HyperTalk for the classic Macintosh RollQuation: Math Puzzles - Apps on Google Play GitHub - dropbox/witchcraft Show HN: Agent-cache – Multi-tier LLM/tool/session caching for Valkey and Redis GitHub - opentalon/opentalon: OpenTalon is an open-source platform built from the ground up in Go as a robust alternative to OpenClaw LinkedIn™ 职位抓取工具 - Chrome 应用商店
GitHub - studiohaze/topaz: A small, closed, Unicode-first...
yo_tafo · 2026-06-16 · via Hacker News: Show HN

A small, closed language for application intent — it reads like Python or TypeScript, and it compiles to a self-contained native binary.

You write the intent. The compiler writes the Rust.

Topaz is written by people and by agents and checked by the toolchain. For each intent there is exactly one canonical form, locked in the Topaz v5.2 specification. The canonical documentation lives at topaz.ooo; this repository is the compiler, runtime, and conformance corpus behind it.

Four pillars

  • A small, closed surface. One way to say each thing. The language decides policy — Result for recoverable failure, optionals for absence, defer for cleanup — so people and agents don't re-decide it per file.
  • Unicode-first identity. Identifiers are Unicode from the lexer up; domain words stay in your language instead of becoming romanized approximations. No silent normalization — what you wrote is what compares equal.
  • Agent-ready by design. A specification small enough to hold in full, machine-checkable profiles, and documentation that is itself verified against the toolchain.
  • Templates with intent. sql, sh, and path strings are structured templates whose parts and interpolations stay separate — safety as grammar, not discipline.
function greet(name: string, language: string) -> string {
    return match language {
        case "한국어" => "안녕하세요, {name}님!"
        case "Русский" => "Привет, {name}!"
        case _ => "Hello, {name}!"
    }
}

let 사용자 = "김토파즈"
print(greet(사용자, "한국어"))
print(greet("Topaz", "Русский"))

사용자 is an ordinary variable.

Reaching Rust without writing it

Rust is the substrate; Topaz is the surface. topaz run executes a program through the reference interpreter, and topaz build lowers it to Rust and compiles a self-contained native binary — offline, from a runtime closure the compiler carries itself, with no Topaz runtime to install on the target:

topaz run   examples/hello.tpz          # execute via the interpreter
topaz build examples/hello.tpz \        # lower to Rust, build a native binary,
      --out-dir out --run               #   and run it — prints "Hello, Topaz!"

The emitter is held to the interpreter by a differential harness: every fixture runs through both engines and must agree, so the binary behaves exactly as the reference does.

Status

The Topaz v5.2 language is locked, and this toolchain implements it end to end — a Unicode-first front end and module resolver, a whole-unit static type checker (topaz check), the reference interpreter (topaz run), and native code generation (topaz emit / topaz build). The toolchain is pure Rust with zero external (crates.io) dependencies and forbid(unsafe_code) across the workspace.

v5.2.0 is the first public release. It is an early release — there are no API-stability or performance promises yet, only that the toolchain does what the locked specification says. topaz.ooo is the source of truth for what runs today and the gates that prove it.

Install

Prebuilt binary. Download topaz for your platform from the GitHub Releases page and put it on your PATH.

From source. With a Rust toolchain installed:

git clone https://github.com/studiohaze/topaz
cd topaz
cargo build --release          # the driver lands at target/release/topaz

topaz run, check, and emit work from the standalone binary; topaz build shells out to cargo through rustup, so that one command needs a Rust toolchain present.

Commands

Command What it does
topaz run <entry> Execute a program through the reference interpreter (--root <dir> for a multi-file unit).
topaz check <entry> Parse, resolve, and type-check a whole compilation unit.
topaz build <entry> --out-dir <dir> Lower to Rust and build a native binary (--release to optimize; --run to execute it).
topaz emit <entry> Lower a single-module program to Rust source — to stdout, or with --out-dir <dir> a cargo run-able crate.
topaz parse <file> / topaz dump-ast <file> Parse and report diagnostics, or print the AST.
topaz version [--verbose] Print the toolchain version.

The language version defaults to 5.2; pass --language-version 5.1 for the frozen single-file mode.

More

  • topaz.ooo — the canonical, verified documentation (introduction, guides, language reference) in English, Korean, and Russian.
  • examples/ — small runnable programs (each parses, type-checks, runs, and topaz builds).
  • docs/PHILOSOPHY.md — why Topaz exists and where it does not compete.
  • spec/ — the vendored, read-only language specification (v5.1 frozen, v5.2 locked).
  • docs/design/ — the Compiler Design Record ledger.

Building from source — the gates

This compiler is developed by AI agents under an adversarial, dialectic review protocol against the locked specification (see docs/design/CDR-005.md). The CI gates are the contract; run them locally with:

cargo fmt --check
cargo clippy --workspace --all-targets -- -D warnings
cargo test --workspace               # 635 tests, including the corpus + differential harnesses
cargo run -p topaz_cli -- check-corpus   # the golden parse/exec gates

Community

Contributing, security, conduct

License

Topaz is licensed under the Apache License 2.0. It includes data derived from the Unicode Character Database; the required attribution is in NOTICE.