Automating 9 years of manual Instagram posting in one week, as a non-developer.
A posting pipeline built from scratch with Python, a Telegram bot, Claude AI and a cloud server, so a 9-year-old car modification account runs without a phone.
The background
Nine years of finding reels, writing captions and posting by hand.
I've been running @modifiedwhipss on Instagram since 2017. Finding modified Indian car reels, downloading them, writing captions, crediting creators, and posting, all manually, every time.
Claude, ChatGPT and Gemini were already part of my day job in marketing, but I'd never built anything with them. So I gave it a week, 6 hours a day.
The build (and the problems)
It did not take one day.
- Instagram blocks every cloud server. 401 errors everywhere. Sessions, cookies, proxies: all blocked. Two days gone just trying to download a reel.
- Video format issues. Instagram delivers video and audio as separate streams. Merging them required a tool that refused to install on the server.
- The first AI captions were bad. Generic, robotic, wrong tone. Better prompts mattered more than the model.
- The bot crashed when two instances ran simultaneously. Fixed with a lock file, but it took hours to diagnose.
How it works now
Paste a link. Get a caption. Approve. Post.
Paste a link in Telegram
Drop any Instagram reel link into the bot and it downloads the video automatically.
Claude AI writes the caption
Credits the original creator, adds the follow line, matches the account tone.
One-tap approval
Review the caption and tap approve. Edit if needed, or just type /postnow.
Posts to Instagram
The bot posts the reel with caption and tags in seconds, without touching a phone.
The real learning
Telling brands to automate is easy. Building it is a different job.
Things broke in ways no plan predicted, and the 2am error logs taught more than any course. If a non-developer can get a pipeline like this working in a week, the barrier for marketers is lower than it looks.
Stack