I spent this week fixing something that's been bugging me for months. My TikTok reposting workflow kept trying to post long videos to Twitter and Blue Sky, which don't accept them. The result was error messages and failed posts.

Out of 47 videos I tried to cross-post last month, 23 failed because they were over Twitter's 2:20 limit. I was manually checking each failure and reposting shorter clips by hand. That's not automation. That's just slower manual work.

But the real problem wasn't the failure rate. It was what I found when I dug into why.

Most automation failures aren't logic failures. They're data failures. You write the perfect if-statement, test it, and it works. Then it breaks. You spend two hours convinced your code is wrong when the real problem is the data your code was reading.

That's exactly what happened here.

My TikTok reposting workflow is pretty straightforward. An RSS feed checks my account every hour, grabs new videos, uploads them to Google Drive, then sends them through Blotato to post across platforms. The problem was videos over two minutes would fail on Twitter and Blue Sky.

The fix I tried first looked solid. Google Drive's API includes video duration in the metadata. I built nodes to pull that data and check the length before posting. Tested it. Worked perfectly. Closed my laptop.

Next day: same errors.

Google Drive only populates duration metadata reliably for about one in five uploaded videos. The rest come back empty. I had built filtering logic on top of data that existed 20% of the time. My "reliable" solution was failing 80% of the time.

I scrapped the whole thing and looked upstream instead.

I already had JavaScript running in my workflow to pull the TikTok video URL. TikTok's own data included the duration. It was flowing through my system the entire time. I just wasn't looking at it.

Ten lines of code. Duration in seconds, minutes, and formatted time. If the video is under two minutes, it posts everywhere. If it's longer, Twitter and Blue Sky get skipped.

The whole mess could have been avoided if I had asked one question before building: is this data source consistent, or just occasionally present?

Microsoft Goes All-In on AI Agents That Actually Do Work

At Microsoft Build, the company laid out their vision for AI agents that work inside your actual business environment. They're updating Windows with local AI reasoning, launching new developer hardware, and connecting agents to real business data through services like Web IQ and Fabric IQ. GitHub's new Copilot app now pulls from multiple AI models.

Joe's Read: Small business owners using Microsoft 365 will see AI agents handling routine tasks directly in their existing workflow within the next year.

OpenAI's Codex Now Reaches Beyond Coders

OpenAI reports Codex hit 5 million weekly users, with 20% being non-developers. They're building role-specific versions for sales, analytics, and other business functions that connect to tools like Salesforce and HubSpot. It's not just about writing code anymore.

Joe's Read: Sales teams and analysts without coding skills can now automate parts of their workflow that previously required a developer.

Google's Gemma 4 Runs AI Tasks on Your Laptop

Google released Gemma 4, designed to handle advanced AI tasks directly on laptops without cloud computing. This means faster response times and no data leaving your machine for certain AI work.

Joe's Read: Small businesses concerned about data privacy can now run powerful AI locally instead of sending sensitive information to cloud servers.

Unwrap

Unwrap consolidates customer feedback from surveys, reviews, support tickets, and social media into one dashboard. It uses AI to categorize feedback and alert you to sentiment changes. Companies like Stripe and DoorDash use it to track what customers actually think.

Caveat: It won't replace reading individual customer messages, just helps you spot patterns you might miss manually.

https://cal.com/unwrap/quick-connect

An API that can return a value is not the same as an API that always returns a value. Google Drive's documentation says it provides video metadata. It does. Just not reliably, and not for uploaded videos the way it does for native recordings.

The lesson isn't "don't trust APIs." The lesson is: test your data source at scale before you build logic around it. One successful test isn't proof. Twenty consistent tests might be.

If your automation keeps breaking in the same place, stop debugging the code first. Go upstream. Check whether the data it's reading actually shows up every time. More often than not, that's where the real problem is hiding.

n8n — The automation tool I use to connect apps, trigger workflows, and stop doing things manually. If there's a repetitive process in your business, this is where you start fixing it.

VoiceInk — A local AI dictation tool for Mac that transcribes your voice with near-perfect accuracy and runs entirely on your device, meaning nothing you say ever touches a cloud server.

Blotato — Handles the full content distribution side of your business: drop in a topic and it generates platform-specific posts, or feed it existing content and it repurposes it across formats. TikTok videos become tweets, podcasts become blog posts. Includes a scheduling calendar, visual creation tools for carousels and infographics, and publishes natively to 9 platforms with no per-post fees.

Beehiiv — What you're reading right now is published on Beehiiv. If you're thinking about starting a newsletter or moving off a clunky platform, this is the one I'd recommend. 20% off your first 3 months with my link.

Google Workspace — Beyond email and Docs, a Business Standard plan includes Gemini Pro built into every app, NotebookLM Plus, and access to the enterprise versions of the whole suite. Better value than a standalone Gemini subscription when you're already paying for Google anyway. 14-day trial and 10% off your first year.

Descript — Video and podcast editing that works like a text document. You edit the transcript and the media follows. Cuts filler words, cleans up audio, and handles captions automatically. 50% off your first two months on the Creator Plan.

A working automation that breaks silently is worse than a broken one. At least the broken one tells you immediately. Build on data you can count on.

Free resources here.

PS: If you want automation that actually works consistently without the trial and error, book a discovery call and we'll build the solution together in a live session.

Cheers,

Joe