A slightly faster slimmer curl
Back in August I mentioned the new curl performance tests we have running. I figured it could be apt to return to this topic just to mention some of the initial outcomes.
Lobsters4 min read
performance
44 stories
Speeding up the Unreal editor by… not using unnecessarily secure hash methods
One good trick when trying to look for ways to optimize software is to stay away from areas that are very likely already quite well optimized, and instead focus on areas that might have been overlooked, either because they are too obscure or don't show up during regular profiling sessions.
Lobsters5 min readJSON Tiles: Fast Analytics on Semi-Structured Data (2021)
Developers often prefer flexibility over upfront schema design, making semi-structured data formats such as JSON increasingly popular. Large amounts of JSON data are therefore stored and analyzed by relational database systems.
Lobsters1 min readHow shredding JSON is giving Logfire 1000x query speedups
How dynamic shredding makes the vast majority of queries in Logfire a thousand times faster, with early adopter customers reporting more than 1000x improvement on query performance.
Lobsters14 min readMaking np.searchsorted up to 25× Faster in NumPy 2.5
We explore how to speed up binary search by batching independent searches with NumPy's vectorized operations. We then reformulate the algorithm so all searches progress together with only $O(1)$ additional memory, port it to C++, and achieve up to a 25× speedup over NumPy 2.4's implementation.
Lobsters13 min readGithub Migrates Copilot Runtime to Rust with AI-Assisted Rewrite
GitHub migrated more than 800,000 lines of Copilot runtime code from TypeScript and Node.js to Rust in about 14.5 weeks using AI-assisted development. The incremental migration used N-API interoperability, automated testing, human review, and 128 pull requests while continuing to ship releases.
InfoQ - AI/ML/Data3 min readIntroducing on-demand CPU and memory profiling with flamegraphs for Workers and Durable Objects
On-demand CPU and memory profiling for Cloudflare Workers and Durable Objects is now available. Now you can generate interactive flamegraphs directly in production to quickly track down memory leaks and performance bottlenecks.
Moltworker Blog9 min readThe Performance Cost of RwLock in Our Read-Heavy Workload
How per-item RwLock reads became expensive in our read-heavy workload, how using lock-free has improved performance, and where a simple RwLock remained a good fit.
Lobsters6 min readrat's minimal register allocator
rat is my smallish compiler backend (with a semi-working C99 frontend). Its x86-64 code generator translates the intermediate representation (IR) into x86-64 instructions. These use an unlimited number of virtual registers (vregs).
Lobsters11 min readBurn 0.22.0: Faster Builds, Easier Extensions, and Smarter Autotuning
Burn 0.22.0: Faster builds without backend generics, easier custom operations, adaptive memory management, smarter autotuning, GPU graph replay, and new fine-tuning and deployment workflows.
Lobsters14 min readJanet on x32: 32-bit Pointers, 64-bit Speed, 25% Less RAM
RAM’s expensive again. Can existing software run faster with 50% less RAM via an unloved compiler flag?
Lobsters12 min readSpeed up your Selenium scripts: Lightpanda now speaks WebDriver
Lightpanda 1.0 speaks WebDriver, making it the only browser that natively speaks every web automation protocol: CDP, WebDriver and MCP. Run your Selenium scripts on it by changing the driver setup: up to 4.4x faster than headless Chrome, with 7x less memory.
Lightpanda Blog9 min readVector Database Memory Sizing and Cost Math: How Much RAM Do 1 Million Vectors Need?
Vector Database Memory Sizing and Cost Math: How Much RAM Do 1 Million Vectors Need? Estimate vector database memory, disk, and cost with embedding dimension, precision, metadata, index overhead, replicas, and capacity planning.
Dev.to AI12 min readpgvector HNSW Filtering: My RAG Asked for 10 Rows and Got 0
The support ticket was one line long: "Your assistant says our docs don't mention refunds. We have an entire page called Refunds." I checked. The page was there. It was chunked, embedded, and sitting in Postgres with the right tenant_id.
Dev.to AI7 min readYour LLM Bill Jumped After You Added Context: Find the Cache Miss Before You Downgrade the Model
If your LLM API spend climbed after you added retrieval, a longer system prompt, or tool definitions, the cause is almost always input tokens being reprocessed at full price on every request — not the model you picked.
Dev.to AI7 min readCustomization: Optimizing Compiler Technology for SELF, a Dynamically-Typed Object-Oriented Programming Language (1989)
Dynamically-typed object-oriented languages please programmers, but their lack of static type information penalizes performance. Our new implementation techniques extract static type information from declaration-free programs.
Lobsters1 min readUpstream Rust maintenance report
As noted in my previous report, I am currently working on the open source Rust toolchain as a Sovereign Tech Fellow. Every two months, I’m putting out a report of my open source work done in that period. This is the second installment of this series.
Lobsters41 min read2026 Birthday week: network performance update
Cloudflare now ranks as the fastest provider across 74% of the top 1,000 global networks. By incorporating background telemetry from Cloudflare Challenge Pages, we have expanded our real-user measurement scale while maintaining user privacy.
Moltworker Blog7 min read
Showing 18 of 44 stories
Load more storiesAiByts Weekly Brief
Get the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts