Arc Notes Weekly #72: Offspring
This week, we're pulling back the curtain on Vercel's serverless sorcery, marveling at the infrastructure wizardry that conjures up seamless scalability and breakneck speed.
This week, we're pulling back the curtain on Vercel's serverless sorcery, marveling at the infrastructure wizardry that conjures up seamless scalability and breakneck speed. We'll also exorcise the demons of database soft deletes, exposing the anti-pattern's dark secrets.
Oh, and we've traded our cozy corner of the web for a shiny new Substack address - but more on our big move later this week.
Enjoy this week's round-up!
— Mahdi Yusuf (@myusuf3)
👋🏾 You are reading Architecture Notes - Your Sunday newsletter, which curates best system design and architecture news from around the web. We would appreciate you sharing it with like-minded people or a kind word on Twitter.
Articles
Why do CPUs have multiple cache levels?
Ever wonder why CPUs have multiple levels of cache? It's like how an office worker has a desk for active files, a cabinet for recent files, and an archive for old ones - each optimized for a different purpose to keep the CPU humming along as fast as possible!
Behind the scenes of Vercel's Infrastructure
Peek under the hood of Vercel's serverless deployment infrastructure to discover a marvel of engineering enabling seamless scalability and blistering performance. From intelligent anycast routing to dynamic image optimization, see how every request flows through a meticulously designed system abstracting away complexity - empowering developers to ship their best work with unrivaled speed and reliability.
Making a Postgres query 1,000 times faster
Imagine having a Postgres query that gets slower and slower each time it runs, until it's taking over 20 seconds! Sounds like a nightmare, right? Well, that's exactly what happened to one engineer at Mattermost. But through some clever sleuthing with EXPLAIN ANALYZE, a bit of luck, and a Postgres feature called "row constructor comparisons", they managed to turn that 24 second monstrosity into a speedy 34 millisecond marvel. That's a 1000x speedup!
Integrate API users 50% faster
Creating a frictionless API experience for your partners and customers no longer requires an army of engineers. Speakeasy’s platform makes crafting type-safe, idiomatic SDKs for enterprise APIs easy. That means you can unlock API revenue while keeping your team focused on what matters most: shipping new products. Make SDK generation part of your API’s CI/CD and distribute libraries that users love at a fraction of the cost of maintaining them in-house. Try Speakeasy for free.
GPT-4o’s Memory Breakthrough!
GPT-4o's memory puts all other LLMs to shame in the new "needle-in-a-needlestack" benchmark, which buries a single query in thousands of limericks. While GPT-4 and Claude get hopelessly lost in the haystack, GPT-4o's laser-focused attention achieves near perfection - a major breakthrough that has the AI world buzzing about what OpenAI did to supercharge their latest model's memory.
You probably don’t need microservices
Think you need microservices? Think again! This dev's wild tales of companies drowning in complexity will make you question everything you thought you knew about architecting software at scale. A must-read before you jump on the microservices bandwagon!
Avoiding the soft delete anti-pattern
Soft deletes - the "best practice" of marking database records as deleted instead of actually deleting them - are an insidious anti-pattern that will haunt your codebase. This deep dive tears down the misleading complexity they create, from crippling constraints to querying quagmires, and shines a light on saner alternatives like lifecycle modeling, temporal tables, and outsourcing history to your data warehouse.
Projects
Bend
Imagine Python-like code that automatically runs 50x faster on your GPU, no threads or locks required. That's the magic of Bend, a new language that makes massively parallel programming feel like just writing normal code. If you thought GPU programming had to be low-level, think again!
Winamp is releasing its source
Cool. What’s old is new again…maybe.
Oracle goes vegan: Dumps Terraform for OpenTofu
Who didn’t see this coming.