Congratulations @elonmusk for being the World's First Trillionaire 🔥 You've been a motivation to me and a constant reminder that "if you can dream it, you can do it" Congratulations 👏👏

I assisted a colleague in building a donation website for her Parish. 1. I designed the backend API with MD5-hashed idempotency keys combining session token, amount, and timestamp to prevent duplicate charges. 2. Integrated Paystack payment gateway with kobo conversion, debugged 500 → 400 → success in real time. 3. Added server-side payment verification so no one bypasses the thank-you page , saved donor records to datyabase after verified payments. 4. I Secured and gatekept all API endpoints Set up ngrok/localtunnel to test webhooks locally. 5. Configured Paystack webhook for reliable background payment confirmation. Built admin dashboard to view all donation records, events scheduling and making blog posts. Deployed to Vercel with proper Express routing via vercel.json for review before switching to real paystack integration. Stack: Node.js · Express · Paystack · Vercel

Rule of thumb : Cookies : notes for the browser Sessions : notes for the server with a key given to the browser .

Spent today fixing a "notifications won't turn on" bug. It turned out to be 6 bugs stacked on top of each other . Every fix just uncovered the next error underneath. Here's the full dig 👇 1. Layer 1: the FCM token wouldn't even mint. 403 PERMISSION_DENIED: Requests from referer http://localhost:8888/ are blocked Firebase's API key had an HTTP-referrer restriction that silently blocked localhost. Added the referrer, moved on. 2. Layer 2: token mints fine now → saves straight into a 500. The push_subscriptions table was missing a column the code assumed existed. Classic "schema drifted from the code" gap. 3. Layer 3: added the column → another 500. there is no unique or exclusion constraint matching the ON CONFLICT specification The upsert needed a UNIQUE index on fcm_token to resolve ON CONFLICT. There wasn't one. Added it. 4. Layer 4: index added → still 500. This was the real boss:violates foreign key constraints "push_subscriptions_user_id_fkey" A foreign key to auth.users was rejecting every single user trying to subscribe. 5. Why would a valid, logged-in user fail a foreign key? Because these are Clerk users, they don't exist in Supabase's auth.users table. They ride on a synthetic mapped UUID. The FK never stood a chance. Dropped it. 200 OK. Notifications finally saved. ✅ 6. Then two bonus bugs from the same root cause mismatched data shapes: Clerk vs GitHub auth returned different user objects, so names rendered blank on some pages → a daily_activity view the UI queried had never been created (clean 404) 8. Added skeleton skins for each page loading states The lesson: The symptom ("can't enable notifications") had nothing to do with notifications. It was an auth-model mismatch wearing a generic error message as a disguise. Read the real error; the one in the network tab and the DB logs not the one the UI hands you. Half the battle was just refusing to trust the top-level error string and digging one layer deeper, six times in a row. visit math-mentor.netlify.app today 🔥

What I did today as well on MathMentor 🥹 1. I fully integrated clerk auth (dev mode) redesigned the login page with a 'last used' feature 2. implemented a sync in the general database with clerk_users independent table via uuid instead of creating extra tables across all fields 3. I improved the UI and made it more mobile friendly, 4. improved the FCM logic, though push notifications got through but pop-up stopped showing,would fix that . 5. prepping for the flutter mobile app ,, I'm using this to build my dev skills and learn more

Same energy over here 👀 Been deep in MathMentor today: 1. I Tried building the full auth flow which worked but kept running into a race condition with GitHub login, so switched to Clerk with GitHub Google OAuth while keeping anonymous accounts with recovery keys intact. Two auth systems, one clean flow. 2. Fought with FCM push notifications all week, streak reminders, re-engagement, admin broadcasts. Getting closer to bulletproof. 3. Tightened the AI solver on Python questions. Turns out telling an LLM "don't contradict yourself" 18 different ways still isn't enough 😅 I haven't pushed the recent updates yet, still fixing some edge cases, but you can check the past update via math-mentor.netlify.app Different stack. Same obsession with getting it right.

Skeleton is the better choice always, it gives the user a sense of belonging, what to expect,, the spinner just makes them eager for nothing especially on a very bad network. I hope I said it well

401: You don't have the right details to go in , step back 403: ooh I know you, but still step back

Bitget Stocks 2.0 is live, and we heard your voice. More stock tokens? When 24/7 trading? Now it's time to show us your wishlist and win a share of $2,000 USDT! Which U.S. stock ticker should come to Bitget Stocks 2.0 next? Repost comment your ticker and reason with #BitgetStocksUpgrade

Mr Bean was pure talent 😂

fact ✅

golang 🥲

npm audit

I mean they could use GitHub codespace naming logic 🥲 you'd see things like funny-bug-finder, legendary-screaming-potato 😂😂😂🤣

“After getting what you manifested, ask for discipline to keep it and wisdom to multiply it.”

not really true as he said, yeah adopting AI makes you work over redundant tasks easily while focusing on other things, so at the end you work on more tasks , the goal is to focus on more important things while the AI works on problems or redundant tasks .. more creativity, more time to work on more tasks 👌

I don't have a MacBook and work station 😌

A good decision requires good information derived from raw facts ✅

✅

A 10GB file shrinks to 2GB after zipping because compression algorithms eliminate redundancy and repeated patterns rather than deleting the actual data.