Hey! This is cool, and actually in the same spirit of my own tool. Not sure if you'd be interested in combining your vision of a repository of architecture models, with my platform (blueprintr.io)?
Author here. I kept bouncing between paywalled courses and link dumps while prepping for system design interviews, and wanted one place where each design is worked end to end and you can click into the architecture instead of squinting at a PNG.
Up front: most of the writing is LLM-generated. I picked the topics, set the structure and reviewed the pages, but the prose isn't hand-written.
The architecture diagrams are real React Flow canvases rather than images — they pan, zoom and go fullscreen, and every node is clickable, showing what it's for, its trade-off, and a link to the concept module explaining it. Node colour is by kind across every diagram, so the same colour always means the same thing.
The numbers are interview-grade rather than measured, so treat them as a starting point. Corrections are the most useful thing anyone can send — that's the main reason it's open source.
Give a shit and write your own text, or even comments. Generated spam is worthless to post. If you can't put in the effort of typing a few human words about 'your' project then how can the actual thing possibly be trusted?
Hey! This is cool, and actually in the same spirit of my own tool. Not sure if you'd be interested in combining your vision of a repository of architecture models, with my platform (blueprintr.io)?
Seems like a match made in heaven :)
Author here. I kept bouncing between paywalled courses and link dumps while prepping for system design interviews, and wanted one place where each design is worked end to end and you can click into the architecture instead of squinting at a PNG.
Up front: most of the writing is LLM-generated. I picked the topics, set the structure and reviewed the pages, but the prose isn't hand-written.
The architecture diagrams are real React Flow canvases rather than images — they pan, zoom and go fullscreen, and every node is clickable, showing what it's for, its trade-off, and a link to the concept module explaining it. Node colour is by kind across every diagram, so the same colour always means the same thing.
The numbers are interview-grade rather than measured, so treat them as a starting point. Corrections are the most useful thing anyone can send — that's the main reason it's open source.
Tell your agent to spell out abbreviations with tooltips. For example, in https://atlas-sysdes.vercel.app/docs/05-notification-system it talks about APNs, and FCM without defining them.
Give a shit and write your own text, or even comments. Generated spam is worthless to post. If you can't put in the effort of typing a few human words about 'your' project then how can the actual thing possibly be trusted?
Great work. Thanks for inspiration.
Is it static collection or the designs were dynamically generated from concepts?
Really nice! It’s also a good reminder that there is always more to discover. Even when you think you know it all.
Checked it out and I must say this is really incredible work. Wow!