Abdi Builds

Geda Built NONOX with no laptop. No team. No funding. Just one phone, 8,500 lines of Java, and a dream that refused to die. NONOX is an Android focus app that blocks distractions, locks temptations, and helps you build deep-focus sessions. Now it’s making its way to Google Play, but it needs testing first. Try it out and help him get it ready: https://nonoxapps.com

Software development is making its way into high schools.

AI makes coding faster. It also makes bad engineering decisions faster.

Apply if u r interested in hackathon

I Guess the secret is doing what you love and staying busy with it.

I used to overthink everything on campus. This summer, I barely overthink, I even forgot I was like that. Hopefully, I stay this way when I’m back on campus.

AI generated but I found it very important.

Building a System: 20 Steps 1. Define the problem —

Be clear about what you're solving and for whom. Tip: Write the problem in 2–3 sentences before coding.

2. Scope the MVP —

Keep only what is necessary to prove the idea. Tip: Remove features that don't directly support the core problem.

3. Define requirements —

Specify what the system must do and how it should behave. Tip: Separate functional and non-functional requirements.

4. Map user flows —

Understand how users move through the system. Tip: Start with the most important user journeys.

5. Design the data model —

Define entities, relationships, fields, and constraints. Tip: Get the database structure right before building heavily around it.

6. Choose the architecture —

Decide how the major parts of the system communicate. Tip: Start with the simplest architecture that meets your requirements.

7. Define APIs/contracts —

Establish how components exchange data. Tip: Define inputs, outputs, errors, and permissions before implementation.

8. Set up the foundation —

Configure the repository, environment, database, linting, etc. Tip: Automate repetitive setup early.

9. Build vertically —

Complete one feature across frontend, backend, and database. Tip: Avoid building the entire frontend before the backend.

10. Use AI deliberately —

Let AI accelerate implementation, research, and problem-solving. Tip: Give AI small, well-defined tasks with enough context.

11. Test continuously —

Test normal cases, edge cases, and failures as you build. Tip: Don't leave testing until the end.

12. Review AI output —

AI-generated code still needs engineering judgment. Tip: Never keep code you can't explain.

13. Handle failures —

Design what happens when things go wrong. Tip: Think about invalid input, timeouts, missing data, and service failures.

14. Add observability —

Make the system's behavior visible through logs, errors, and metrics. Tip: If you can't see what's happening, debugging becomes guesswork.

15. Review security —

Protect data, users, APIs, and infrastructure. Tip: Check authentication, authorization, validation, secrets, and common vulnerabilities.

16. Deploy early —

Get the system into a real environment before it's “perfect.” Tip: Production exposes problems your local environment hides.

17. Get real feedback —

Observe how people actually use the system. Tip: Real user behavior is more reliable than assumptions.

18. Refactor with evidence —

Improve areas that actually cause problems. Tip: Don't prematurely optimize or over-engineer.

19. Document decisions —

Record important architectural and technical decisions. Tip: Document the “why,” not obvious code.

20. Iterate the system —

Improve architecture, features, UX, and reliability continuously. Tip: After every iteration, ask: What did we learn, what broke, and what should change?

Me: Why does this function need 40 lines? Me: Let me simplify it. Me 2 hours later: Now it needs 73.

Late-night grinding. Who’s awake?

Then you fix one annoying bug and suddenly you’re motivated again.

Nothing humbles a developer like opening their own code from 6 months ago😁

AI changed the developer workflow. Your learning strategy should change too.

One underrated developer skill: Knowing when not to build something.

Getting real users is harder than building. I expected AbdiMarket-P2P to get more users after launching… turns out shipping the app was the easy part. Getting people to actually use it? Whole different boss fight. If you see this, give it a shot. Buy something, sell something, and let me know what you think. I need actual humans using this thing, not just me refreshing the dashboard. https://abdimarket-p2p.vercel.app/?refresh=1

Finally original AbdiMarket is available for live demo I started building AbdiMarket as an e-commerce platform, but halfway through, I thought a P2P marketplace would be more useful for users. So I switched direction, built and launched AbdiMarket-P2P first. Then I came back to the original version—and now almost finished it. Two versions. One project. A lot of lessons from building, changing direction, and actually shipping. 🔗 https://abdimarket.vercel.app

Improved feature you can now contact seller without logging in. But u've to login to sell, review and rate, report, view wishlist... Check it out what's improved. https://abdimarket-p2p.vercel.app/?refresh=1

I often see new developers build a few projects and immediately call themselves founders. I respect the ambition, but I think many people rush into it. I think your odds improve when you first build solid technical skills and get close to real problems—through freelancing, a job, research, open source, or real users. Don't start with, "What startup should I build?" Start with, "What problem do I understand really well?" Once you've solved that problem enough times, you'll be in a much better position to turn it into a product that helps many people. There are always exceptions, but great startups are usually built on deep problem understanding, not assumptions.

I overthought the MVP features for my upcoming project... a lot. Then I realized my MVP was trying to become V10. Now I just ask: - Does it solve the core problem? - Is it actually a must-have? - Will it help validate the idea? If not, it goes to the "Later" list.

Just finished adding the Forgot Password feature. 🔐