I don’t have answers for this, but I do think sound human judgment is more important than ever. The truth is that we still very much work for people. Agent-driven workflows make it easy for design system teams to lose track who they work for. But, assuming that they didn’t, they turned out to be so oblivious to common humanity that their product might just as well be called sinister.
But then I remembered how people treat people. My immediate thought about this was to leave people stuff to people. Lean on OS/platform defaults whenever possible.
Interviewers love candidates who can explain not just what scales a system, but why it scales. In real systems, you’ll often use a hybrid approach depending on read/write ratios. A great System Designer understands how data behaves, how queries perform, and how to structure information so it remains accurate, fast, and easy to retrieve.
Defining the Core Entities
- This is one of the most reliable ways to stand out, because many candidates know components but cannot tell a coherent story.
- Implement optimistic UI updates that show expected results immediately while confirming with servers asynchronously, making the interface feel responsive even when network conditions are poor.
- But people are notoriously bad at understanding chaining events.
- Generalizing this concept, various large-scale applications can be developed using components like API gateway, load balancer, CDN, messaging queue, cache, database, etc.
- Some systems maintain summarized versions of old posts in hot storage while keeping full content in cold archives.
- The thing everything else gets built on.
Let’s assume each postID is 10 bytes, to store 200 posts per user we’d need 2KB per user. Each entry in the PrecomputedFeed table is keyed by userId and its value is a list of post IDs in reverse chronological order, limited to a small number of posts (say 200 or so). Is your user with 100k follows really going to notice if their posts are appearing a couple minutes late? Your instinct here should be to think about ways that we can compute the feed results on write or post creation rather than at the time we want to read their feed. We have a working system that allows a small number of users to follow a small number of users with a small number of posts and view their feeds! Since we already have a GSI on the post table which sorts posts by creation timestamp, we can use that directly to return only posts older than the cursor.
Users in between might use either model based on additional factors like posting frequency or the activity levels of their followers. Modern news feed systems at scale use hybrid approaches that apply different models based on user characteristics. Instead of pre-computing feeds, the system fetches recent posts from all accounts a user follows, applies ranking, and assembles the feed in real-time. The choice between pushing content to followers proactively versus pulling it when requested fundamentally shapes a feed system’s performance characteristics, cost structure, and complexity.
Elastic by default: serverless + events
Interfaces encourage speed even when judgement sic, care, or reflection would serve people better. Sure, building components and tokens have arguably become faster with AI. My gut says that many design systems have some version of the first two. How many have one that’s actually used? How many design systems have a north star?
Read vs write patterns and their impact
I don’t think this is what the people who make design systems set off to do. But now, with AI tooling, code being the source of truth makes it easier than ever for design and engineering to be a https://dominicanrental.com/mozhno-li-razvernut-nejroset-na-svoem-servere.html REAL, single team. This includes a thorough understanding of distributed systems, load balancing, caching strategies, and other advanced concepts necessary for building robust, scalable systems.
🖤 Design Systems
If you persist an append-only log of changes, you can rebuild derived state after failures. Make the data model explicit enough to explain your reads and writes. This is a great interview walkthrough because it starts simple and naturally grows into caching and async work. Once you can name the hot path, you https://commonpost.info/the-quantum-leap-major-tech-consortium-announces-q-day-breakthrough/ can reason about which dependency dominates p95 latency and which resources saturate first. Performance problems are usually architectural problems before they are “code” problems.