- Sounds basic, but this is a good process to fix any bug in code https://overreacted.io/how-to-fix-any-bug
- Ensure you have a reliable, repeatable way to observe the bug
- Define and confirm the exact steps to consistently reproduce the bug
- Simplify and adjust the repro case while validating that it still exhibits the bug
- Incrementally remove code or factors, checking each time that the bug remains
- Continue reducing until you isolate the minimal case or root cause
- A bug in DynamoDB’s system that manages DNS updates caused its main server address in the US-East-1 region to break. Basically, a timing glitch made it write a blank DNS record, and the automatic repair system didn’t fix it https://thundergolfer.com/blog/aws-us-east-1-outage-oct20
- This single point of failure took down more than 140 AWS services and exposed systemic issues inherent to large-scale distributed systems
- Distributed systems are fragile, with latent bugs that can have a massive real-world impact
- Simple root cause analyses are inadequate; failures result from a combination of faults such as race conditions, poor control mechanisms, and human factors
- Even the most advanced cloud providers can suffer compound outages due to system complexity, operational practices, and metastable failure modes
- Modern software reliability practices should focus on control problems and interconnected system design, not just bug hunting
- Mixing database calls, network requests, and other external interactions directly with your core logic can lead to code that’s difficult to test, reuse, and understand. Instead, consider writing a functional core that’s called from an imperative shell https://testing.googleblog.com/2025/10/simplify-your-code-functional-core.html, https://www.destroyallsoftware.com/screencasts/catalog/functional-core-imperative-shell
- A functional core should contain pure, testable business logic that’s free of side effects (such as I/O or external state mutation). It operates only on the data it’s given
- An imperative shell is responsible for side effects, like database calls and sending emails. It uses the functions in your functional core to perform the business logic
- ChatGPT’s new browser, Atlas, is more than just a Chromium clone https://openai.com/index/building-chatgpt-atlas
- It uses OWL (OpenAI’s Web Layer), which abstracts away Chromium’s UI, letting Atlas engineers build much faster without modifying Chromium directly
- It runs Chromium in a completely separate background process, making the UI instantly responsive and crash-proof
- It uses SwiftUI and Metal for a modern, animated user interface on top of Chromium’s engine
- If you build your life around convenience, you start living too abstractly. You begin to think fine things belong to brands or rare occasions, and forget that luxury can exist in the ordinary. We know people who can afford anything, but drink bad coffee not because of taste, but because of neglect. Real luxury isn’t logo-based; it’s fork-based, cup-based. Every day is all there is. The things you touch daily should mean something to you https://map.simonsarris.com/p/viscerality
What I’ve been reading (W44, 2025)
Fixing bugs, systemic failures in software, better code, ChatGPT Atlas internals, and finding pleasure in everyday things