You’re in a backend interview. They ask: Your API gets 10x more traffic overnight. What’s the first metric you check? Here’s how to answer 👇

Talk is cheap, send patches.

If a CTO is self-hosting and has not set up redundancy and fail over, he may be under qualified. But 37 Signals---they are self-hosting with multiple servers in different regions. CTO 101. You don't need cloud to have redundancy.

Never call an API in a transaction. In relational databases, it's best practice to keep transactions as short-lived as possible. Longer transactions lead to more resource contention. If each transaction acquires row locks, being longer-lived means higher likelihood of lock waits and execution delays. Then there's MVCC. Postgres keeps multiple versions of rows for MVCC, and long-running transactions keep old rows from becoming autovacuum candidates. MySQL handles this with an undo log. When there are lots of long transactions, the log has to be kept around longer, log length and performance. If you're using a connection pooler like PgBouncer in transaction pooling mode, long transactions can easily result in pool exhaustion producing failures on your clients. The database is the most core part of your stack. Keep it as contention-free as possible.

Talked to lots of Rails shops & devs this week and strengthened my confirmation bias that the Rails frontend sucks. Why? Because most of them use Rails as a backend, but on the front end they use React, Vue.js. In the middle they do the API toil & think “that’s just how it is”

When N+1 queries in SQLite actually save the day:

Behind the products I love are those who pour their hearts into what they do. The most beautiful thing is that both @dhh and @mitchellh have kids, families, and interesting hobbies. They build great things while living a whole life. Thank you for great products & inspiration.

A SUA #FURIACS É CAMPEÃ MUNDIAL NO CS2! CHEGAMOS NO TOPO COM O TÍTULO NA FISSURE PLAYGROUND 2! 🏆 ISSO É POR NÓS, POR VOCÊS, PELO BRASIL! RESPEITEM O CS BRASILEIRO! 🇧🇷 🥇 @FalleNCS 🥇 @tvoy_molodoy 🥇 @yek1ndar 🥇 @yuurih 🥇 @kscerato 🥇 @siddecs 🥇 @krizzencsgo 🥇 @HepaCS É #DIADEFURIA NESSA PORRA VAMOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

Presente

Still one of the most insane lines I've read in any ML paper.

I built a shitty Dropbox in 30 minutes. I removed everything from Dropbox, cancelled my account and put 'File Browser' by Github on my Beelink Mini NAS. Omarchy is killing my fear of computers. Thanks @dhh

X just open-sourced the code for the “For You” algorithm Everything you need to know in 1 tweet and 1 image: 1. The Raw Inputs (Data) Everything starts with three big buckets of data: - Social graph → who you follow, who follows you, your network connections. - Tweet engagement → likes, retweets, replies, bookmarks. - User data → your profile, behavior, what you click, what you ignore. This is the “fuel” that powers the whole system. 2. Feature Engineering The raw data runs through different models/features that try to make sense of it: - GraphJet → real-time graph of tweets, retweets, follows. - SimClusters → groups people and tweets into communities of interest (like “AI Twitter” or “NBA Twitter”). - TwHIN → Twitter’s “heterogeneous information network,” basically a map of user-to-tweet relationships. - RealGraph → predicts the strength of connections between users. - TweepCred → credibility scoring. - Trust & Safety signals → moderation, blocks, mutes, NSFW filters. Think of this layer as the “filters” that figure out what’s worth even considering. 3. Candidate Sourcing (Home Mixer) This is where tweets get pulled in as “candidates” to potentially show you: Search Index (find relevant tweets directly) - CR Mixer, UTEG, FRS (different mixers pulling tweets from clusters, graphs, and recency) They pull from different pools so you don’t just see the same circle of people. 4.Heavy Ranker Now the candidates run through a big neural network (the heavy ranker). It scores tweets based on predicted engagement: likelihood you’ll like, retweet, reply, spend time reading. 5. Heuristics & Filtering After the ML ranker, some hand-coded rules kick in: Social proof → is it liked by people you follow? - Author diversity → don’t show 10 tweets in a row from the same person. - Visibility / T&S → block spam, NSFW, muted accounts. - Content balance → avoid overwhelming you with one type of content. - Feedback fatigue → avoid showing you the same tweet repeatedly. 6. Mixing Finally, the ranked tweets get mixed with: - Ads (obviously). - Who to follow suggestions. 7. What does this mean for you? Probably not much. - Pick a niche - Write valuable tweets - Reply to people in your niche with value - You'll grow and build an audience X is a beautiful place to grow an audience to meet interesting people and get insight for what businesses to build for It's worth being here. Enjoy.

Esse é o ctop: o htop dos containers. Monitoramento em tempo real do que tá sugando CPU, memória e rede no seu Docker. Simples, visual, direto ao ponto.

Jim Rohn's call for massive action is one I've heeded many times in my career. It's incredible what doors and opportunities that open when you show up with a relentless spirit to make things happen.

DHH just declared Docker is annoying as a dev environment and used AAA game rendering an image in 8ms as a baseline for what's reasonable wait time for things

What a great idea! Learning to operate your OS with the keyboard takes a minute or two, but by the time you have it down, it's hard to image computing any other way. Manually dragging overlapping windows around quickly seems downright barbaric!

Omarchy 2.0 is ready!! New ISO, AUR-free installation, Chrome micro-fork, Starship prompt, new icon, and 400 other changes from 45 committers 🤘 github.com/basecamp/omarchy/…

It's important to remember that "technology influencers" can be bad role models. Many encourage people to be "fans of technology" instead of actual developers of technology. Hot takes are easy, writing good code is not.

Thrilled for @37signals to be the first corporate sponsor of Hyrpland! @vaxryy has built something incredibly special, and I think it's key we accelerate to offer developers an easy, compelling, and sexy-as-hell alternative to computing with Windows or macOS 🤘

“When researchers forced ChatGPT users to write without AI, they performed worse than people who never used AI at all. It's not just dependency. It's cognitive atrophy.” I have a bet this is why in-person coding interviews on the whiteboard will be back. A hedge against this