#performance
// 18 transmissions tagged with #performance
Mitchellh blog breaks down SIMD for performance‑critical code
Mitchellh.com’s July 22, 2026 article explains SIMD fundamentals, instruction sets, and practical pitfalls, giving engineers concrete guidance for speeding up compute‑heavy workloads. The piece also sparked a lively Hacker News thread on real‑world adoption.
GLM 5.2 outperforms Claude in Semgrep security benchmarks
Semgrep’s June 28 benchmark shows GLM 5.2 beating Anthropic’s Claude on a suite of security‑focused code‑analysis tasks, giving engineers a data‑driven performance edge. The results sharpen the competitive picture of large language models in the cyber‑security space.
DeepSeek open-sources inference optimizations, 60–85% faster generation
DeepSeek released a paper detailing open‑source inference optimizations that deliver 60–85% faster text generation, giving engineers concrete techniques to speed up LLM deployments.
OpenAI previews GPT‑5.6 Sol with new API and performance target
OpenAI released a preview of GPT‑5.6 Sol, a next‑generation model, together with a new API endpoint and a concrete performance target for developers to plan against [OpenAI Blog].
Garbage Collection Handbook releases second edition in 2023
The Garbage Collection Handbook’s second edition, released in 2023, updates its coverage of automatic memory‑management techniques for performance‑critical systems.
ClickHouse adds Silk fiber runtime to boost query scheduling
ClickHouse released Silk, a fiber‑based runtime that reworks query scheduling and execution, aiming to lower latency and raise throughput for real‑time analytics and large‑scale data processing.
Half-Life 2 runs in a browser via WebAssembly
A developer team compiled the full Half-Life 2 game to WebAssembly, delivering a browser‑based version that matches native performance. The demo is live at hl2.slqnt.dev.
Cloudflare launches SOCKMAP, a new TCP splicing mechanism
Cloudflare unveiled SOCKMAP, a kernel‑level TCP splicing feature that cuts latency and streamlines connection handling for high‑performance networking stacks. The code is open for testing and review.
Microsoft's new Outlook adds 10‑second delay versus Outlook Classic
The latest Windows version of Outlook introduces a performance regression, with common tasks taking up to 10 seconds compared to the instant response of Outlook Classic, raising concerns about productivity and competitiveness.
Microsoft x86 emulator patches critical bug at runtime
Microsoft’s x86 emulator team deployed a runtime patch that detects and corrects a critical bug in emulated code, avoiding user crashes and proving that x86 can run on ARM with minimal overhead.
Google releases DiffusionGemma, a model that generates text four times faster
Google’s DiffusionGemma model cuts per‑token latency by a factor of four while preserving text quality, opening the door to real‑time NLP workloads on modest hardware.
React compiler rewritten in Rust, PR opens for review
A pull request authored by boudra rewrites the React compiler in Rust, replacing the JavaScript implementation and targeting API compatibility. The PR is under review on GitHub.
Linear's speed dissected in a technical breakdown
A performance.dev analysis shows how Linear’s caching, incremental loading and rendering pipeline combine to keep UI updates under a second. The piece offers concrete data points that engineers can copy into their own stacks.
Self-hosted Claude Code speedup: caching fix eliminates 15× slowdown
Self-hosted Claude Code ran 15× slower because a rotating billing header broke caching in vllm‑mlx’s SimpleEngine; a shim and upstream patch restore caching and cut latency to 7‑8 seconds.
Nvidia GPU VRAM used as Linux swap space
The nbd-vram project lets Linux users mount Nvidia GPU VRAM as a swap device, offering a fast‑memory alternative for RAM‑starved systems.
Fast mode for Opus 4.7 on AI Gateway cuts latency 2.5x at 6x cost
Vercel's AI Gateway now supports fast mode for Claude Opus 4.7, delivering 2.5x faster output token generation with full model intelligence, priced at $30 input and $150 output per 1M tokens.
Postgres 18's async I/O subsystem hits 3x on sequential scans
Postgres 18 introduced an asynchronous I/O subsystem that issues parallel I/O requests instead of waiting on each one. Benchmarks show up to 3x gains on seq scans, bitmap heap scans, and vacuum.
Bun 2-3x ahead of Node on RPS. The runtime question is now serious.
2026 benchmarks put Bun at 30-50K RPS on standard HTTP workloads vs Node's 13-20K. Deno 2 sits in the middle around 22K. With Deno's full Node-compat and Bun's drop-in story, the runtime decision is no longer academic.