To become extremely good at System Design, you need to master concepts from the ground up.
Not just surface-level knowledge.
Deep understanding.
Here is a clear roadmap of concepts you must learn extensively, grouped by category.
Save this thread.
Delete Account is not a moral event.
It is a schema decision.
If your “delete user” job touches the ledger, the chart, the transcript, the claim file, or the invoice, you did not ship privacy. You shipped a compliance incident with a friendly button.
“Right to be forgotten” has a hard edge. Close the identity. Do not rewrite history.
What must survive deletion, in 5 industries:
I’m curious how other engineers would approach this:
You’re building an ingestion layer that needs to pull data from a third-party platform as close to real-time as possible.
The catch? You’re not fully confident in the third party’s engineering quality or event implementation.
Would you go with:
→ Polling
→ Event-driven/webhooks
→ Hybrid: events for speed + polling for reconciliation
And more importantly, why?
What would you choose in production, and what failure scenarios would influence your decision?
Would love to hear how you’d design it.
If you want to build your personal brand on X and attract clients, Do this asap!
First, screenshot your X profile.
I mean the part with your:
→ Profile picture
→ Header
→ Bio
→ Link
→ Pinned post
→ Some of your recent posts
Try it. I know it takes a minute, but actually do it.
Now go to ChatGPT or Grok and paste this prompt 👇
Two bugs, same root lesson: reading code tells you what it's supposed to do. Watching it run tells you what it actually does.
When those disagree, the browser is right and your mental model is wrong.
A dropdown menu stopped working after I added a library. Not "buggy" — completely dead. No console errors either.
Spent an hour reading code before I gave up and just watched it happen in DevTools instead.
Same method — DevTools, not code reading. Watched the click event actually fire.
Found a second click listener, somewhere else in the codebase, closing the dropdown immediately after it opened. A leftover from an earlier version, never removed.
Odoo is an open-source enterprise resource planning (ERP) platform that helps teams manage operations.
And choosing how to host it comes down to how much control you want vs the operational overhead.
In this article, @AbdulTa62185243 compares self-hosted and managed Odoo hosting, including the tradeoffs around maintenance, security, customization, and cost.
https://t.co/UQPhDsKEl2
@IamAroke Only experienced CORS errors and it dealt with me for months because the backend dev was in Ghana and I in Lagos, omo the project no later fly o. Merge conflict🤧😭
Don’t give AI the whole project at once to build.
A good workflow is:
1. Product requirements
→ What are we actually building?
2. System architecture
→ What are the major components and how do they interact?
3. Domain model
→ What are the entities, relationships and business rules?
4. API/contracts
→ What does each service expose and consume?
5. Design system
→ How should the product look and behave?
6. Project conventions
→ How is code organised and written?
7. Implementation plan
→ Break the system into independently implementable units.
8. AI implementation
→ Give AI one bounded unit with the relevant context.
9. Tests
→ Verify the implementation against the contract.
10. Review/refactor
→ Make sure the implementation actually conforms to the architecture.
The key distinction is:
AI should answer “How do we implement this?” far more often than “What should our architecture be?”
You can absolutely use AI for architecture too, but you should make the final architectural decisions yourself.
AI is very good at generating alternatives, identifying trade-offs and spotting gaps; it becomes dangerous when its first generated solution silently becomes the architecture of your product.