F*: A general-purpose proof-oriented programming language

5 hours ago

F* is a proof-oriented programming language combining dependent types with SMT solving and tactic-based interactive theorem proving.
It compiles by default to OCaml and can be extracted to F#, C, Wasm, or assembly using various tools.
F* is open source under Apache 2.0, available on GitHub, and installable via multiple methods including binaries, OPAM, Docker, Nix, or from source.
Learning resources include an online book, a Low* tutorial, and course materials from seasonal schools.
F* is used in industrial and academic projects such as Project Everest, HACL*, ValeCrypt, EverCrypt, and EverParse, with production deployments in Firefox, Linux kernel, Python, Azure, and more.
Extensive research on F* covers semantics, effects, security, cryptography, systems, parsing, and program verification, with many published papers.

Fuse is a statically typed, purely functional language with higher-kinded types and ad-hoc polymorphism.
It compiles to the GRIN whole-program optimizer, producing LLVM-generated native code.
Features: statically typed (based on System F with higher-order polymorphism, algebraic data types, generics, traits).

Converts JSON to human-readable GitHub Flavored Markdown with deterministic output.
Runs in browser, Node, Go, and command line; TypeScript and Go implementations produce byte-identical output.
Output is a one-way projection using canonical formatting: # Results heading, LF endings, no trailing spaces.

transcribe.cpp is a ggml-based open-source transcription library supporting various ASR models with numerical validation and WER testing.
It addresses challenges in distributing cross-platform speech-to-text applications, offering GPU acceleration via Vulkan, Metal, CUDA, and TinyBLAS.
The library supports over 16 ASR families (60+ models), streaming/batch transcription, and is a near drop-in replacement for whisper.cpp.

Cyberscript10 hours ago

Inkling-Small is a general-purpose multimodal model that accepts text, image, and audio inputs and generates text outputs.
It can be used with libraries like Transformers, vLLM, SGLang, and Docker, as well as through APIs and third-party providers.
The model uses a 42-layer decoder-only transformer with a sparse Mixture-of-Experts (MoE) architecture, totaling 276B parameters but activating only 12B per token.

Hasty Briefsbeta