Happy Monday! ☀️
Welcome to the 84 new hungry minds who have joined us since last Monday!
If you aren't subscribed yet, join smart, curious, and hungry folks by subscribing here.
📚 Software Engineering Articles
🗞️ Tech and AI Trends
👨🏻💻 Coding Tip
Use
SKIP LOCKEDto claim queue jobs across PostgreSQL workers
Time-to-digest: 5 minutes
Airbnb runs hundreds of MySQL clusters that handle millions of queries per second. Synthetic benchmarks helped engineers measure throughput, but they could not show how an upgrade would handle Airbnb's real queries, transactions, and application assumptions. The team built a system to record production traffic and replay it against a test database before making a change.
Earlier, Java, Python, and Ruby applications each logged queries through separate tooling. Engineers struggled to maintain those pipelines and reconstruct the transaction context needed for a faithful replay.
The challenge: Airbnb needed to capture real workloads without changing every application, preserve transaction behavior, and compare results without exposing production data.
Implementation highlights:
Capture at ProxySQL. Engineers recorded queries at the database proxy that already sat on the traffic path, giving them one capture point across application languages.
Move logs off the proxy. A Log Mover sidecar shipped ProxySQL's logs to object storage before local disks filled up.
Rebuild the workload. A Log Processor decoded the logs, grouped queries by database cluster, reconstructed transactions, and split traffic into five-minute buckets.
Preserve comparison behavior. The team rewrote inserts to retain production auto-increment IDs and accounted for queries whose row order had no explicit guarantee.
Replay at controlled load. Distributed workers used scheduled start times and adjustable speed factors to reproduce concurrency and test capacity.
Results and learnings:
Upgrade regression. A MySQL 5.7-to-8.0 replay exposed a query whose latency rose from 0.03 to 2.6 seconds.
Capacity limit. Adding 80% more write traffic raised commit latency from about 6 to 34 milliseconds in one test.
Correctness and privacy. Replay caught result differences that throughput tests missed; engineers encrypted captured queries, restricted access, and kept replay in production-equivalent environments.
Replay gives engineers a way to test the workload their applications send, including assumptions hidden in query results. Production users make poor volunteer test runners.
ARTICLE (proof by GPU)
OpenAI shares progress in mathematical research
ARTICLE (large, sparse, fast)
Reflection introduces Beam, a 501B-parameter open-weight model
ESSENTIAL (platform called back)
Why don't more developers "use the platform"?
ARTICLE (trust, then query)
The Agent Said It Was Done. The Database Disagreed
ESSENTIAL (delete that intro)
Anti-Patterns in Software Blogging
ARTICLE (build it, bin it)
Ephemeral testing
ARTICLE (code got cheap)
When code is cheap, judgement becomes the job
ARTICLE (agents need a home)
We built our own cloud agents runtime. Here's what we learned
ARTICLE (math in a box)
calc-input, a custom input element that accepts mathematical formulas
ESSENTIAL (all aboard errors)
Railway-oriented programming
Want to reach 200,000+ engineers?
Let’s work together! Whether it’s your product, service, or event, we’d love to help you connect with this awesome community.
Brief: OpenAI's Decisions API beta returns structured classifications, choices, and scores for text or images to support routing and review tasks.
Brief: Anthropic put Dashboards in beta for paid plans with BigQuery, Databricks, and Snowflake connections, while Motion's animated-explainer beta targets Team and Enterprise plans.
Brief: Google will add JPEG XL decoding to Chrome 155 using the pure-Rust jxl-rs library.
Brief: Mistral opened public preview of Large 4, a multimodal mixture-of-experts model with 1 trillion total parameters and 49 billion active per token.
Brief: Google made Spanner Omni generally available for deployments outside its cloud, using software-based time synchronization in place of dedicated clock hardware.
Brief: GitHub used Copilot to migrate its runtime from TypeScript to Rust, producing roughly 800,000 lines of Rust code.
Brief: Shopify moved five Checkout Blocks extensions to Polaris Web Components and cut transferred JavaScript by 40–85% across those extensions.
This week’s tip:
Claim queue jobs with SKIP LOCKED
When several workers poll jobs, have PostgreSQL lock one queued row and skip rows other workers hold.
Wen?
You run multiple workers against a PostgreSQL-backed queue.
You need each worker to claim a different job without a separate broker.
If you're walking down the right path and you're willing to keep walking, eventually you'll make progress.
Barack Obama
That’s it for today! ☀️
Enjoyed this issue? Send it to your friends here to sign up, or share it on Twitter!
If you want to submit a section to the newsletter or tell us what you think about today’s issue, reply to this email or DM me on Twitter! 🐦
Thanks for spending part of your Monday morning with Hungry Minds.
See you in a week — Alex.
Icons by Icons8.
*I may earn a commission if you get a subscription through the links marked with “aff.” (at no extra cost to you).