Why Casper forked Rust's rand crate

5 hours ago

The author forked Rust's 'rand' crate to create 'urandom', aiming to reduce papercuts and improve discoverability of everyday operations.
'urandom' places the high-level consumer API on a single 'Random' wrapper struct, eliminating the need to import multiple traits for common operations like random range, choose, and shuffle.
'urandom' seals its low-level 'Rng' trait, preventing external generators and allowing the crate to optimize algorithms (e.g., Xoshiro) for specific outputs like f32 and f64, achieving up to 31% higher throughput in benchmarks.
Uniform sampling in 'urandom' uses a unified, unbiased lazy threshold calculation, which is simpler and often faster than 'rand's split approach with separate biased and unbiased paths.
'urandom' guarantees reproducibility of deterministic generators across architectures and SemVer-compatible releases, unlike 'rand' which may change output in minor updates.
The trade-off for 'urandom' is less flexibility (no arbitrary generators) and a smaller ecosystem compared to 'rand', but it offers improved performance and ease of use for common tasks.

CodeCrucible is a new LLM-driven static application security testing (SAST) tool for vulnerability discovery, but the focus is on the design blueprint rather than the tool itself.
The key design choices involve four questions: compaction (how to fit code into context window), identification (how to prompt for vulnerabilities), relevance screening (how to filter false positives), and determinism (how to ensure stable outputs).
Whole-repo concatenation packs the entire repository into a single LLM call, enabling cross-file reasoning, and works better than expected despite earlier assumptions that it would not scale.

<antirez>5 days ago

V8's Math.random() PRNG is weak for generating unique IDs, as discovered by Mike Malone.
Unique IDs require either stateful mechanisms (like counters with persistence) or a large random space with uniform distribution to make collisions negligible.
Using a large enough random space (e.g., 160 bits) makes collision probability astronomically low, even at high generation rates.

Transition from Julia to Rust for scientific computing due to Julia's runtime issues in large codebases.
tenferro-rs is a Rust-native dense tensor stack supporting autodiff, einsum, FFT, and explicit CPU/CUDA backends.
Rust's ownership and types enhance correctness at compile time, making it practical for AI-written code.

The author delays building a local AI inference rig due to fast market changes and high hardware prices.
China may restrict export of top AI models, including open-weight ones like Qwen and GLM-5.2, threatening Western access.
Open-weight models from China enable affordable local and serverless AI; a ban could limit this access.

Pseudorandom number generators (PRNGs) produce clumpy values, which can be undesirable for exhaustive exploration of a space.
Quasirandom number generators (QRNGs) generate uniformly dense points, improving Monte Carlo simulations by avoiding grid search artifacts and converging faster (error proportional to 1/n vs 1/√n).
The author implemented a Rust crate for quasirandom sequences, using a generalized golden ratio method to generate d-dimensional points.

CodeCrucible is a new LLM-driven static application security testing (SAST) tool for vulnerability discovery, but the focus is on the design blueprint rather than the tool itself.
The key design choices involve four questions: compaction (how to fit code into context window), identification (how to prompt for vulnerabilities), relevance screening (how to filter false positives), and determinism (how to ensure stable outputs).
Whole-repo concatenation packs the entire repository into a single LLM call, enabling cross-file reasoning, and works better than expected despite earlier assumptions that it would not scale.

<antirez>5 days ago

V8's Math.random() PRNG is weak for generating unique IDs, as discovered by Mike Malone.
Unique IDs require either stateful mechanisms (like counters with persistence) or a large random space with uniform distribution to make collisions negligible.
Using a large enough random space (e.g., 160 bits) makes collision probability astronomically low, even at high generation rates.

Transition from Julia to Rust for scientific computing due to Julia's runtime issues in large codebases.
tenferro-rs is a Rust-native dense tensor stack supporting autodiff, einsum, FFT, and explicit CPU/CUDA backends.
Rust's ownership and types enhance correctness at compile time, making it practical for AI-written code.

The author delays building a local AI inference rig due to fast market changes and high hardware prices.
China may restrict export of top AI models, including open-weight ones like Qwen and GLM-5.2, threatening Western access.
Open-weight models from China enable affordable local and serverless AI; a ban could limit this access.

Pseudorandom number generators (PRNGs) produce clumpy values, which can be undesirable for exhaustive exploration of a space.
Quasirandom number generators (QRNGs) generate uniformly dense points, improving Monte Carlo simulations by avoiding grid search artifacts and converging faster (error proportional to 1/n vs 1/√n).
The author implemented a Rust crate for quasirandom sequences, using a generalized golden ratio method to generate d-dimensional points.

Hasty Briefsbeta