The latest articles on DEV Community by sanhita dawn (@sanhita_dawn_a1e0df0f1714). https://dev.to/sanhita_dawn_a1e0df0f1714 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3991118%2F908a5996-77c3-4c79-93d5-ea60c2921405.png https://dev.to/sanhita_dawn_a1e0df0f1714 en sanhita dawn Fri, 19 Jun 2026 07:33:08 +0000 https://dev.to/sanhita_dawn_a1e0df0f1714/i-read-about-system-design-for-months-it-didnt-make-me-better-at-it-2bil https://dev.to/sanhita_dawn_a1e0df0f1714/i-read-about-system-design-for-months-it-didnt-make-me-better-at-it-2bil <p>_For the longest time my system design prep looked like this: read a blog post about how Twitter's timeline works, nod along, feel smart for ten minutes, forget the actual tradeoffs by the next day.</p> <h2> The Problem Was Never The content </h2> <p>It took me embarrassingly long to realize the problem wasn't the content — it was that I was never actually building anything. I could explain what a load balancer does in an interview, but I'd never had to decide between round-robin and consistent hashing under an actual constraint, watch that decision blow up my latency, and fix it.</p> <h2> Reading and Doing Are Different Skills </h2> <p>System design is one of those skills where reading and doing are almost unrelated activities. You don't learn to debug a production incident by reading a postmortem. You learn it by being paged at 2am once. Obviously nobody's paging you for practice, but you can get close to that feedback loop — build something, see it fail under load, figure out why, fix it.</p> <h2> What Actually Changed When I Started Building </h2> <p>Once I started treating practice that way — actually wiring up architectures and watching them get scored on latency, cost, and correctness instead of just sketching boxes and arrows — things stuck a lot better. Same with schema design. I used to "design" a database by drawing five tables and calling it done. Actually being forced to define keys, constraints, and relationships and then seeing where they broke under realistic queries taught me more in an afternoon than three articles on normalization ever did.</p> <h2> The Gap Nobody's Filling:Gen AI System Design </h2> <p>The other thing nobody really talks about: most system design practice out there is still 100% interview-focused — URL shorteners, rate limiters, the greatest hits. Almost nothing covers designing actual GenAI systems — RAG pipelines, AI agent architectures, vector databases, model routing. If you're working anywhere near LLM products right now, that's a real gap, and it's the part of system design practice that's grown the most this year while most prep content hasn't caught up.</p> <p>If anyone's looking for something hands-on along these lines, I've been using <a href="https://scaledojo.dev/" rel="noopener noreferrer">ScaleDojo </a>— interactive labs for HLD, LLD, API design, and GenAI systems, free to start. Not affiliated, just found it useful enough to mention._</p> architecture career learning systemdesign