Aurora Algorithm (@aurora.algorithm)

#programming #algorithm #rust #nodejs #python

Are semicolons optional in JavaScript? I deleted every statement-ending semicolon (18,241) from 100 popular npm packages and ran each package's own tests. 68 still passed, 20 had no semicolons to delete, and 12 broke. The code needed 43 of them: 22 before a line starting with (, 18 before a line…

Is Protobuf faster than JSON? The same 1,000 shop orders written to bytes and read back, one core (P-core 4 of an i7-13700F, Windows 11), 5.0 s warm-up + 5.0 s measured, median of 5 rounds. Go 1.27.3, Rust 1.99.1, Java 25.0.5. - Size: a page of 50 orders is 25.6 KB as JSON and 11.4 KB as Protobuf…

#rust #postgres #golang #programming #backend

Claude Code keeps no memory between messages, so every message sends the whole conversation again: the system prompt, your CLAUDE.md, every file Claude read, every test log and every reply. Prompt caching makes the unchanged part cheaper (cache reads are billed at 0.1x the base input price, and…

Linux runs Android phones and most websites, but it isn't Unix. The kernel's own README calls it a clone of Unix, written from scratch. Here's where that clone came from: Bell Labs in 1969, a teacher's Minix, then a student in Helsinki in 1991. - Unix started at AT&T's Bell Labs in 1969, and AT&T…

Go answered 22 times as many requests as Python in our test, and at 1,000 requests a second the two were 0.3 ms apart. The same orders API built twice: FastAPI 0.162.2 with SQLModel 0.0.48 on Python 3.14.10, the way FastAPI's full-stack template writes it (sync handlers, a Pydantic response model…

Is Postgres faster than MySQL at updates? Postgres 18.6 and MySQL 8.4.9 got the same single-row updates on one table of 1,000,000 trips with 10 secondary indexes: 64 connections from oha 1.16.0, 20 s per run, Windows 11 on one i7-13700F, Postgres with 4GB of shared buffers and MySQL with a 4.0 GB…

In Claude Code, the effort level sets how hard Claude works on your task, and thinking time is only one part of it. It also sets how many files Claude reads, how much it checks its own work (like running your tests after an edit), and how far it goes on a long task before it checks in with you. At…

Linux kernel vs distribution: what "Linux" actually means When people say Linux, they mean one of two things. The first is the kernel, the program between your apps and the hardware; apps reach it through system calls such as write(). The second is a distribution, which packs the kernel with GNU's…

How much faster is C# than Java? The same API built twice, Spring Boot 4.1.2 on Java 25 and ASP.NET Core 10 on C# 14, each in Docker on its own 4 cores of an i7-13700F, loaded by oha with 64 connections. Every number is a run you see on screen. - JSON response, a small object built from the id…

How many days did it take Linus to write Git? Linus Torvalds says Git "really came together in just about ten days or so", though he'd been mulling over the ideas for a while. He wrote it for the Linux kernel, which had used the proprietary BitKeeper for three years until its free version was…

Is JWT encrypted? Paste one into jwt.io and read the email (JWS vs JWE) JWT explained with a real login: you send your email and password, the server sends back a JWT, and your browser shows it on every request to prove who you are. It looks encrypted, but a normal JWT is signed (a JWS), so jwt.io…

32-Bit vs 64-Bit: What's the Real Difference? You put 4 GB of RAM into a 32-bit Windows PC, and System Properties reported only about 3 GB usable (commonly 2.75 GB to 3.5 GB depending on the devices installed). Where did the rest of your RAM go, and why does 64-bit Windows still have a `Program…

Microsoft's Azure CTO says stop new C++ projects. Why Rust? Sources: - Mark Russinovich on X, 2022-09-19, as reported by The Register: https://www.theregister.com/2022/09/20/rust_microsoft_c/ Narration: AI voice (ElevenLabs) - https://try.elevenlabs.io/aurora-algorithm #cpp #rust

WebSockets vs Server-Sent Events vs long polling: how a server pushes to the browser #websockets #server-sentevents #sse

Fastest way to load 1,000,000 rows into Postgres? INSERT per row vs batches vs COPY We loaded the same 1,000,000 rows into the same Postgres table four ways from Node with node-postgres, 5 runs each, order rotated every round. #postgres #postgresql #database #programming

How does a database index make a lookup ~400x faster? The B-tree, page by page #database #index #btree #programming

Every QR code is just 0s and 1s. Here is how your phone reads them into a link. #qrcode

Why Docker exists, what a container really is, and why a push that changes only code rebuilds just the top layer. #docker

Difference between C and C++ (for beginners) Difference between C and C++ for beginners: where they came from, what the difference feels like in code, and why both are still used. - C: Dennis Ritchie made it at Bell Labs in 1972 to write the Unix operating system, so it stays close to the…

How does DNS work? How google.com becomes an IP address Your computer asks a resolver, a server that looks up IP addresses for you. That's usually your internet provider's, or a public one like Google's 8.8.8.8. The resolver then asks three servers in turn: - A root server (for example…

Is Drizzle really the fastest Node ORM? (Benchmark: pg vs Kysely vs Drizzle vs Prisma) #postgresql #database #programming

Is regex slow? Myth or fact: "regex is slow, use string methods instead"? #regex #javascript #python #programming

Why Cloudflare replaced nginx with Pingora? Cloudflare's production figure (quoted from their blog): one large customer's reuse went from 87.1% to 99.92%. Source: https://blog.cloudflare.com/how-we-built-pingora-the-proxy-that-connects-cloudflare-to-the-internet/ Our Pingora is a minimal proxy…

Why setTimeout doesn't run on time: the JavaScript event loop, step by step setTimeout(fn, delay) doesn't run fn exactly when the delay is up. The delay only means "not before". We follow one example Node program step by step: the call stack runs one function at a time, setTimeout hands the timer…

A race condition in its most familiar form, the lost update: two requests read the same counter, both add 1 in the app, both write the same value back, and one increment disappears. We ran it on Postgres: 64 Python threads, each with its own psycopg connection, added 1 to one counter row 10,000…

What does gRPC actually send? JSON vs protobuf, byte by byte #json #grpc

UUID v4 vs UUID v7 vs bigint primary keys: measured on 5 million rows

Does try/catch make your code slow? Use try freely for error handling but don't throw for normal control flow in hot loops #javascript #python