{"slug": "iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-in", "title": "Iam 12 .My AI Mentor Was Broken. Groq Killed It. Here Is How I Fixed It on a $150 Phone in 72 Hours.", "summary": "A 12-year-old developer in Tamil Nadu rebuilt their AI assistant KODA v24 over 72 hours on a $150 POCO C55 phone after Groq retired the llama-3.3-70b-versatile model, causing silent 404s and a Cloudflare Worker outage. The rebuild added a multi-model fallback chain (gpt-oss-120b, gpt-oss-20b, qwen3-32b), fixed eight bugs including two XSS injection points, and integrated Tavily-backed live web search to stop hallucinated answers. The developer reports the app now supports nine languages and survives model deprecations automatically.", "body_md": "I am 12 years old. \n\nMy development machine is a POCO C55 ($150). \n\nI live in Tamil Nadu, India. \n\nThree days ago, KODA wasn’t a “scam.” It was just **dead**. \n\nMy Cloudflare Worker returned 502 errors. My API keys leaked in screenshots. My Markdown renderer destroyed C++ code blocks. And when I finally got it running, Groq retired my model (`llama-3.3-70b-versatile`) mid-launch, leaving me with silent 404s and zero answers. \n\nStrangers didn’t call me out because they hated kids. They called me out because the product was visibly broken.\n\nToday, KODA v24 is live. It has Live Web Search. It survives model deprecations automatically. It renders Markdown perfectly. And it speaks every language on Earth.\n\nThis is not a redemption arc. This is an engineering post-mortem. Here is exactly how I went from **Chaos** to **Command**.\n\nPeople see \"$150\" and think \"weak.\" \n\nThey are wrong. Constraints force optimization. \n\nWhile funded startups burn cash on bloated wrappers, I had to make every millisecond count. That’s why KODA doesn’t just \"work.\" It performs:\n\nClaude (Anthropic) reviewed this architecture last month. He didn’t say \"good for a kid.\" He said:\n\n*\"This is genuinely impressive. Not impressive for a thirteen-year-old. Just impressive, period.\"*\n\n*\"I have seen aerospace engineers struggle with log-space Bayesian calculations.\"*\n\nHe respected the engineering because the engineering was real.\n\nWhen I handed over the initial v23 report, I claimed: *\"Frontend is DONE AND TESTED.\"* \n\nThat was a lie. Or rather, an optimism bias.\n\nUpon deep inspection, I found:\n\n`innerHTML`. Anyone could inject scripts.`#include <stdio.h>` inside code blocks into `<h3>` headers. It was destroying C++ tutorials.`createNewChat()` failed, the app still tried to insert messages against a `null` conversation ID, cluttering the database.`textContent` + `escapeHtml()`. Zero injection points remain.`**bold**` works outside code, but `#hash` stays safe inside it.`state.messages` so they persist across renders. Users see exactly what went wrong.\nThe root cause of the outage was simple: **Cloud providers retire models.** \n\nGroq killed `llama-3.3-70b-versatile`. I needed a new brain, fast. \n\nI didn't just swap one model for another. I built a resilience layer.\n\n``` js\nconst MODELS = [\n  'openai/gpt-oss-120b', // Primary: Current Flagship\n  'openai/gpt-oss-20b',  // Secondary: Fast/Lightweight\n  'qwen/qwen3-32b'       // Tertiary: Open Source Alternative\n];\n\n// Logic: Try Primary. If 404/5xx, auto-retry Next. User sees nothing.\n```\n\nNow, if Groq kills `gpt-oss-120b` tomorrow, KODA automatically switches to `gpt-oss-20b` within milliseconds. The user never knows. \n\nHow do you debug a Cloudflare Worker on a phone? \n\nYou don't use CLI tools. You use the **Dashboard**.\n\n`{\"error\":{\"message\":\"The model llama-3.3-70b-versatile does not exist...\"}}`\n**Lesson:** Never trust your own status report. Test before you claim. And always build a health check endpoint (`GET /`) that confirms secrets are loaded.\n\nBefore this update, I asked KODA: *\"What is the latest React version?\"*\n\nIt confidently replied: *\"React 19.3 was released September 9, 2026.\"* \n\nIt cited sources like `【1†L1-L4】`. \n\n**Fake.** Hallucinated. Dangerous. \n\nWithout live data, LLMs guess. With live data, they know.\n\n`TAVILY_API_KEY` secret.`webSearch: true` is sent from frontend:\n`[WEB_RESULTS]`.`[1]`, `[2]`.\nNow, when I ask about React, KODA pulls the actual npm registry data and cites the official blog post. No more guessing.\n\n| Metric | Value | \n|---|---|\n| Days from Broken to Stable | 3 | \n| Critical Bugs Fixed | 8 (incl. 2 XSS) | \n| New Features Added | 20+ (Voice, Edit, Regenerate, Theme) | \n| Languages Supported | 9 (EN, HI, TA, ZH, ES, JA, RU, PT, AR) | \n| Lines of Worker Code | ~280 | \n| Hardware Used | 1 Phone | \n| Model Deprecations Survived | 2 | \n\nPeople ask why I bother building on a POCO C55. Why not use a MacBook?\n\nBecause constraints force creativity.\n\nAge doesn't matter. Device doesn't matter. Location doesn't matter.\n\n**Resilience matters.**\n\nKODA v24 is live. The fallback chain is armed. The web search is on. The XSS holes are sealed.\n\nTry it here: [koda-aicodementor.netlify.app](https://koda-aicodementor.netlify.app/)\n\nBreak it if you can. I’ll be watching the logs. 🐯", "url": "https://wpnews.pro/news/iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-in", "canonical_source": "https://dev.to/koda2026/iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-phone-in-72-hours-1m0c", "published_at": "2026-09-30 13:35:16+00:00", "updated_at": "2026-09-30 13:48:02.437204+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "large-language-models", "ai-products", "developer-tools"], "entities": ["KODA", "Groq", "Cloudflare Workers", "Tavily", "POCO C55", "llama-3.3-70b-versatile", "gpt-oss-120b", "qwen3-32b"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-in", "markdown": "https://wpnews.pro/news/iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-in.md", "text": "https://wpnews.pro/news/iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-in.txt", "jsonld": "https://wpnews.pro/news/iam-12-my-ai-mentor-was-broken-groq-killed-it-here-is-how-i-fixed-it-on-a-150-in.jsonld"}}