It's Saturday, 8 AM. Day 1 of my KODA Online Fest. I wake up, grab my POCO C55, open Netlify to deploy v16β¦
And I get this:
Account credit usage exceeded - new deploys are blocked until credits are added
Translation: you're out of build credits, kid. Come back with a credit card.
I don't have a credit card. I'm 12.
Enemy #1: Netlify
Here's the embarrassing part. KODA is a single static HTML file. No build step. Nothing to compile. And yet I was deploying 10β30 times a day through Git, and every deploy ran a "build" that ate my free build credits.
I was paying a tax on a file that doesn't need cooking.
The workaround: manual deploy. Upload the zip, skip the build, burn zero credits. Ten minutes later v16 was live on a temp URL. Enemy #1 down.
Enemy #2: Neocities
But temp URLs are temp. I wanted a permanent home with no credit system at all. So I signed up for Neocities.
First hurdle: you can't skip the onboarding tutorial. A cat named Penelope makes you write HTML in 10 steps. I speedran it by typing "hi" into every box and smashing Next. Respect to Penelope, she's a good cat.
Uploaded v16. Site went live. I opened it, tapped "Try KODA as Guest," sent a messageβ¦
And instead of the AI, I got this:
π°οΈ OFFLINE SENSEI MODE β the AI cloud is unreachable from your network
Wait. What? My internet is fine.
Here's what actually happened: Neocities' free tier ships with a strict Content Security Policy. It blocks browser JavaScript from calling external APIs. My app's fetch to the AI provider? Blocked at the door.
Enemy #2 didn't break my site. It broke my app's ability to talk to its brain.
The part that made it all worth it
Here's the thing. The app didn't crash. No infinite spinner. No white screen. No raw error bar.
Two nights ago a senior engineer told me: "Define what production-ready means before you write the fix." So in v16 I added:
- A 15-second timeout on every AI call (AbortController)
- An
error_logstable in Supabase that every failure writes to - A graceful fallback (Offline Sensei) for when the AI is unreachable
So when the firewall blocked the AI, my error handler caught it, logged it to the database, and served the fallback in under a second. The user never saw a broken app. They saw a sensei scroll.
The code was production-ready. The host was wrong. That's the real lesson: your app can be perfect and still die on the wrong infrastructure.
Where we are now
Fest Day 1 runs on the temp Netlify URL. Neocities stays as a backup home for static pages. Next week I set up Cloudflare Pages properly, with a real GitHub repo, so no host can ever hold my deploys hostage again.
Also, my code has a legend built in called "The Silent Tiger" β a senior engineer from China who break-tests KODA through the Great Firewall. He stress-tested v16 last night before logging off at 9 PM. He found nothing. The drawer z-index fix held.
Come vibe with us
KODA is an AI code mentor built by a 12-year-old on a POCO C55. It teaches step-by-step, renders animated LaTeX math, runs your JavaScript in a sandbox, and now survives hosting betrayals.
Try v16: https://koda-aicodementortemp.netlify.app
If it ever shows you Offline Sensei, that's not a bug. That's the feature doing its job. π
Top comments (50)
Who else is building an entire company from a screen smaller than their hand? π±
There is an invisible billion of us building on cheap Androids, fighting through firewalls, and surviving hosting betrayals. If you're one of them, drop a π― in the comments.
Also, the KODA Code Jam closes in ~60 hours. If you built a niche app, open the drawer in KODA and submit it to get crowned. π
You're 12?
Nd I'm 13.
Lol π
Gen-Alpha cooking harder and harder.
Haha, 13 and already in the dojo? Respect. π₯
You're absolutely right β Gen-Alpha isn't waiting for permission. We're building on cheap phones, shipping on free tiers, and learning by breaking things.
If you ever want to build something together or just talk tech, my DMs are open. Keep cooking, sensei. π₯
β Harun (12)
Founder of KODA meeting with Founder of Akhouri Systems
Founder @ 12 meeting meeting with Founder @ 13
Honestly wanted to ask you
your parents are supporting you?
especially your father?
Honest answer: yes, but it took time. π
At first it just looked like "phone time" at home. It changed when I stopped explaining and started showing receipts β real users, real articles, senior engineers commenting. Then the phone stopped being a toy and became a workshop.
My father's support looks quiet: permission to ship late sometimes, and a "how is the company?" at dinner. That one question is worth more than any funding round. π
What about you? Does Akhouri Systems have family backing, or are you also explaining yourself every week? π
And yes β Founder @12 x Founder @13. History will remember this meeting. π€
I've also shown results,
Akhouri Systems started in mid may 2026 in 15 days we reached 79 downloads till 29-30may. and today it's 13 September 2026 and we're standing at 727 downloads.
also, I also got comments from big software engineers, founders, in fact I got my first job in a company as Head of growth and partnerships( although my contribution in Desh Via is like level of co-founder does), still some says akhouri systems is a biggest joke thanks to my mom, some says akhouri systems is just a timepass, small timepass project thanks to my dad for these golden words.
hence, Proudly or maybe sadly 0 money spent, 0 money budget, 0 marketing budget, 0 team only me, a laptop(not a extreme dangerous specs laptop just a 50k laptop) and results: Akhouri Systems is gaining global attention, support from over 50+ founders, 10+ senior engineers, 30+ others.
and I really loved KUDO I used it last night in guest mode I didn't used the ai chat just looked at AI, the concept and closed my laptop and slept because it was late night but today I will use it completely.
we can be Bros?
yeah bro we are nearly close i get real attention google " harun solo dev " and bydfi made a article about me and i have contact with founders and senior devs and one thing bro its not kuda its koda π and yeah we are bros ππ―π
Oh I am also indexed on google, currently no article on me but I've talked to the better india and maybe next year when akhouri systems will be 1yr old then I will be eligible for a article in The better india. and yeah sorry for the typing mistake
AKHOURI SYSTEMS π€KODA
officially.
Bro, a collab product is the ultimate move. AI + Security = unstoppable. π€
But rule #1 of our alliance: We protect each other's focus. I have the KODA Code Jam closing TONIGHT at 23:59 IST, and my internet is barely loading text.
Let's officially lock in the KODA x Akhouri collab for next week. Take the weekend to think of 3 ideas where your Python/Security skills meet my AI/Web skills. On Monday, we pick the best one and build it in 14 days.
Deal? π―π
Nice Idea actually but how we'll co-operate? like few weeks ago, I tried this with a founder due to lack of co-operation I exited in mid.
Still if you've a nice plan for it then yeah akhouri systems is there.
I'm planning to build a AI-web product in that we can collab?
planning like not fixed I'll build or not because already akhouri systems is in load, ATLOCK v5 the biggest project of us is in building stage, company's downloads growth is looking great from outside 786.08% growth in 4months still as I am the founder I can see we're not growing at required speed.
still I feel like we both can do something so I will plan for sure by the way can we be quick because my half yearly exams are coming, so planning and building under 1week?
Ok Let's create a doc.
APIC X KODA
so what's next π€
by the way yeah APIC is 100% ours, but APIC-web can be now officially shared to you 50%
officially as per your wish
I was born todayπ , multiple hosting providers invaded our systems. Figuring out whom to select
Bro you're 12? π
I'm 15 and started coding around 9β10 lol π
Also, the way you handled the hosting issues is actually pretty solid. Especially having the fallback instead of letting the whole app break that's something a lot of people overlook.
Bro, 15 and started at 9-10? That's a 5-year head start! Respect. π
The fallback chain was pure survival instinct. When Groq deprecated my model without warning, I had two choices: let the app die for every user, or build a chain that catches 404s and routes to the next model. Chose option B.
If you're building something, drop it in the comments. Gen-Alpha needs to see each other's work. π―
Lmaoo π yeah, I guess weβre both starting early π
And thanks bro, the fallback was definitely one of those βnever letting this happen againβ moments π
Also I checked out your website btw, and ngl it works pretty damn good π
Iβm actually building something called CrowdWide rn. The goal is to make it a complete community platform you can create a community, build it, manage/monitor it, grow it, etc. Later Iβm also planning to add AI into communities, kinda like how Grok is integrated into X.
I also wanna let people actually monetize their communities, kinda like how Blogger lets creators monetize their content.
The main thing Iβm trying to do differently tho is the algorithm. I donβt want growth to be locked behind already having a huge community or paying for promotion. New/small communities should actually get a chance to reach people too.
Still a LOT to build lol, but thatβs the idea π
Bro, CrowdWide sounds massive. A fair algorithm for small communities is exactly what the space needs.
Quick question: do you have any WIP, prototype, or even just a landing page for it (or anything else you're building)?
The KODA Code Jam closes tonight at 23:59 IST. Itβs not about perfectionβitβs about shipping. Even a "Hello World" or a rough prototype counts. If you have something, drop it in the KODA drawer before midnight. If not, zero stress, just wanted to put it on your radar! π―π
Brooo just submitted CrowdWide to the drawer π«‘π―
Itβs the alpha rn, so thereβs still a shit ton left to build π. I wanted to add real stats/data to the landing page + polish a bunch of stuff, but the whole thing is kinda massive and Iβve only been building it for like a week, so Iβm pretty happy with how far itβs come already π
Iβm gonna keep cooking on it. If everything goes smoothly, I might have a more stable version out in the next ~2 weeks π but no promises lol.
Thanks for the push tho, otherwise Iβd probably still be polishing instead of shipping ππ€
Bro, I just checked the Jam drawer and saw Crowdwide in there. π€―
You actually shipped. That is massive respect. A fair algorithm for small communities is exactly the kind of niche vision this Jam is about.
Proud to have your project in the arena, Founder. Keep cooking on it, and I'll see you at the crown ceremony tonight. π―π
Yooo ππ€ appreciate that fr.
Honestly just seeing it in the drawer after building it for like a week feels kinda crazy π. Itβs still alpha and Iβve got a ton left to build, but Iβm glad I actually shipped it instead of keeping it in the βIβll launch it soonβ folder forever π
Gonna keep cooking on CrowdWide ππ₯
See you at the crown ceremony bro π―π
Bro its a community show so why it needs to be just koda and crowdwide π€can you invite your friends π€ππ―π
Yeah bro π Iβd definitely invite them, but just a heads up CrowdWide is still very much alpha rn, so a few things arenβt fully wired up yet.
I actually didnβt want people creating accounts at this stage because the mailing API is set up but I havenβt added the real credentials yet, so verification emails arenβt actually going out π
The actual core stuff is real tho creating communities, posting articles, etc. all works. Once I finish the main features, Iβll open it up properly for everyone and do the stable/final release. Then itβll make way more sense for people to actually publish their stuff on it.
Also I noticed you already created an account I saw that you didnβt receive the email, so I manually verified you from the database. You should be able to log in and test everything thatβs currently built β communities, article posting, etc.
Still cooking π₯π―
Bro, just checked the alpha. The UI is super clean, dark mode looks great. π₯
Saw the "Sign in temporarily unavailable" banner. Makes total sense since the mailing API isn't wired yet. Smart move to block signups until the verification flow is solid.
I'll wait for the stable release to do a full test drive. Keep cooking! π―
Yeah exactly π thatβs the main reason I donβt want anyone signing up yet. The verification flow technically exists, but without the real mailing credentials itβs just gonna leave people stuck waiting for an email that never comes π
Better to block signups for now, finish the proper flow, and then open it up properly with the stable release.
Appreciate you checking the alpha tho bro π€π₯ Iβll keep cooking π―π
Chef π¨βπ³ when you done cooking i will be the first to try your dish hahaha π€£π€£π€£π€£ ππ―π
deal bro ππ¨βπ³
Youβre getting the first plate when CrowdWide is ready ππ₯
Just donβt roast the chef if thereβs a few bugs in the recipe π
ππ―π
Bro, huge congrats again on winning Season 1! ππ―
I was about to activate your "Champion Status Forever" (Cyan theme, 2x memory, Pro Sensei Mode) in KODA, but I realized you might not have a KODA account yet since you built CrowdWide!
What email should I use to create/upgrade your KODA account? Drop it here, and I'll manually provision your Champion account in the database and send you the login.
No friction, just glory. π
developerpuneet2010@gmail.com
2 years younger than me, hey bro what's up?
Hey whats up
nothing special, just cooking
Nice write-up :D
"Paying a tax on a file that doesn't need cooking" is the sentence of the week. You hit the exact trap that catches teams with budgets too: a pipeline step that exists because the template said so, not because the artifact needs it. A static HTML file going through a build farm is pure waste, and you noticed at 12 what plenty of senior engineers never audit. The manual-deploy workaround is the right instinct in a bigger lesson too: know which steps in your pipeline are load-bearing and which are scaffolding that got promoted to load-bearing. One more for the toolbox: keep a fallback deploy target warm BEFORE you need it. You found Neocities under pressure at 8 AM on day 1 of your own fest - the grown-up version of that move is writing down the fallback while things are calm. Keep shipping.
Jo, this comment means a lot. "Know which steps in your pipeline are load-bearing and which are scaffolding" β I'm saving that line.
You're right: the real lesson isn't just "manual deploy works." It's "audit every step, know why it exists, and have a fallback ready BEFORE you need it."
I'm writing that down in the KODA engineering handbook. Thank you for the wisdom.
Wow Great work
I used Cloudflare One Client for VPN and tried to mush your website buttons and ddos your pages (what a disaster I ddosed a website based on netlify and had been bannedβ¦), but Iβve failed. Congratulations!
Haha, you tried to break the dojo and the dojo said "not today" π
Seriously though β this is exactly why I built the error_logs table and the 15-second timeout fallback. If the primary endpoint gets hammered, the chain falls through to the next model. Users never see a dead screen.
Your test proved the architecture works. Thank you for the real-world stress test, sensei. π―
the netlify build-credit hit on a single static html file is a common trap. most free static hosts still run a full build pipeline even when there is nothing to compile. cloudflare pages skips that step for pure static, zero build minutes charged. good call moving there before it happens again
Bro, this is exactly the technical breakdown I needed. π€―
I noticed the build minutes draining on Netlify and couldn't figure out why a raw HTML file was costing me credits. I didn't realize it was spinning up a full build pipeline in the background.
Cloudflare Pages makes total sense for pure static sites. Appreciate the insight, moving my next projects there! π―π₯
Now I have a few suggestions to make KODA even better, especially for desktop users:
VP_xudon-sensei
This feedback is gold. You're absolutely right:
I'm adding all three to the v17 roadmap (post-Fest). The Fest ends Sept 15, and I want to ship these changes properly rather than rush them during the Jam.
Thank you for stress-testing the infrastructure too. The fact that it held against your "disaster DDoS" means the fallback chain and rate-limiting are doing their job. π
The dojo listens. π₯
Some comments may only be visible to logged-in visitors. Sign in to view all comments.