The Social Media Machine Guide

What This Guide Is

This guide is not a technical manual. It is 90 days of lessons from running an automated social media system across two brands — what worked, what broke, and what I would do differently if I were starting today.

The system posts 10 times per day across LinkedIn and Instagram for SoftHire Systems and @buildwithtacha simultaneously. My daily involvement is a 10-minute Monday review. Everything else runs without me.

I built it using Python, Claude API, and browser automation. The implementation requires real technical setup — if you want to build this yourself, you will need either technical skills or a developer. What this guide gives you is something more durable than code: the strategy, the lessons, and the honest account of what it actually takes to run a system like this well.

• Posts per day — 10 — 5 LinkedIn + 5 Instagram
• My daily input — Zero — fully automated
• Weekly review time — 10 minutes on Monday morning
• Monthly API cost — Under $5 (Claude API for caption generation
• Time running — 90+ days and counting

The goal of reading this guide: you should leave understanding exactly how to think about this kind of system — the content strategy, the quality control habits, the failure modes, and the decision to automate at all. Whether you build it yourself, hire someone, or decide it's not right for your business, you'll make that decision clearly.

The System — What It Does and Why It Works

The system has four layers. Understanding what each layer does matters more than understanding how each layer is built — because the how changes as platforms update, tools improve, and your needs evolve. The what stays constant.

• Layer — What it does — Why it matters
• Content Brain — Claude API generates captions and hashtags from a brand voice system prompt. Every post is written in my voice, at scale. — Without a brand system prompt that knows your voice precisely, automated content sounds like automated content. This layer is where quality lives.
• Visual Assets — A Python script using the Pillow library renders branded images — stat cards, quote cards, carousel slides — from templates. — Visual consistency at scale requires templates, not one-off design. The template is built once. Every post after that is a render.
• Content Queue — All posts live in a schedule file synced to Google Drive. I review from my phone on Monday. Posts with issues get flagged before they publish. — The queue is the only manual touchpoint in the entire system. It is not optional. Automation without human review is how errors compound.
• Publishing Engine — Browser automation opens Chrome, logs into each platform, and posts the scheduled content at the scheduled time. — This is the layer that has the most failure modes. Platform changes break it. Sessions expire. The troubleshooting section covers all of them.

Why It Works

The system works because each layer has a clear job and a clear failure mode. When something goes wrong, I know exactly which layer to look at. When I want to improve output quality, I know exactly which layer to adjust.

Most automation attempts fail not because the technology is hard, but because the architecture is muddled. The content generation, the scheduling, and the publishing are three different problems. Treating them as one is why most home-built social automation breaks down after two weeks.

"Automation doesn't replace strategy. It amplifies whatever strategy you already have — good or bad."