Happy to be here

I no longer have to worry about designing IG video covers. Here’s how I achieved it. STEP 0 - GOAL My desired goal was to create a simple process for creating Instagram video covers. One that I’d not have to revisit for a long time. STEP 1 - STRATEGY I started by mapping out the kinds of content I wanted to create and grouping them into themes. I decided that for recognition, each theme should have a consistent style. STEP 2 - DESIGN I designed 8 different covers on Figma using real content to ensure I created complete design assets. STEP 3 - TACTIC The first hack was to create a reusable template out of each design. I removed only the variables and left every other constant like logo and background design. STEP 4 - WORKFLOW USING ICM - In my project, I created a ‘content/backgrounds’ folder to hold all the assets (original designs and templates for reference) - Claude reads the blog and drafts several headline options and I pick one or write mine. (It follows some sets of rules to do this) - It then looks up the theme the blog belongs to and picks the matching template. - Opens the template in an invisible web browser (using puppeteer) and designs the cover. - Saves the cover design in a ‘content/socials/covers’ folder For this workflow I continued to refine my instructions until each AI generated design looked like the original cover design. Again ICM shines here with simple folders and .md files running the show. WHAT MADE THIS WORK? 1. Control the level of consistency. The eight Figma frames mean every cover is on-brand by construction. Nothing about the look is left for Claude to decide. 2. Leave exactly one thing variable. Here it is the headline. The fewer moving parts per asset, the more reliable the automation. 3. Put the human gate on the decision, not the output. Approving words as text is fast and honest. Approving a generated image means judging the entire cover design instead of the headline. 4. Write the rules down, with failed examples. The failed headlines are worth more than any list of principles, because they show what "wrong" looks like in practice. Each one was added once it was caught.

Using ICM: - You don’t load everything, every time just to complete a simple task. - Inspection is easy if you can read. - If something is not right you can literally edit your files before you continue. - Keeping your workspace neat is nothing technical. To reorder simply rename the folders. Want to add a stage to your process? That’s a new folder away. Want to handoff? Simply copy the folder. - There’s no need for extra documentation because the instructions are the documentation. All of these benefits are strategic for a simpler workflow that anyone can operate and for the right type of work it’s almost all you need. ICM handles 90% of your strategy, but you have to show up with your 10%. Your unique workflow will require you to plan and strategize on how to achieve your desired goals. For instance, if you want a content creation workflow, what’s your goal? Where will you source your content from? How will you keep the engine running? What tactics will you use to do more with less? These and more I believe is your 10%. You don’t have to have everything figured out before you start, you can grow as you go and ICM allows that level of flexibility. But understanding that you need to have some level of clarity on the problem you are solving, your strategy, and tactics will help you move faster and work better with folders and files.

Hey guys, I'm not sure if this has been talked about in here before but I found this alternative to ElevenLabs in GitHub , I'm aware several people here already use ElevenLabs so just dropping this here if it helps

Before you take a retainer or sell that dashboard, watch this one. https://youtu.be/CWbJIBgiLvc It walks through the scoping and discovery steps that, in my mind, have to happen first. Because the dashboard is the easy part. With AI you can have it on screen in an afternoon and the client will say yes on the spot. What it costs is underneath: the CRM it has to read, the records nobody trusts, the workaround one person does every Thursday that nobody else knows about. The underlying data is the job. There's an old Bell Labs joke about this, the ninety-ninety rule: the first 90% of the code takes 90% of the time, and the last 10% takes the other 90%. My version from the field is that 90% of your time goes to the 10% of cases nobody wrote down. If you scope the happy path, the rest of it is free work. So the video is about what I do before there's a price. A paid diagnosis, fixed fee, where I walk a day with the people who touch the records and follow the work rather than the wish list. What was the last thing you fixed by hand. Where do the hours go. Which workaround did you stop noticing. Then the three ways you can take the job: replace the system, pull the data out into something trustworthy, or investigate the discrepancies and hand back findings they can use with the vendor they already have. I go with the third one first. Small number for the client, they walk away with evidence either way, and it's the only one where you know what you're promising before you promise it. Which is the part I actually care about. Underpromise and overdeliver is how you keep a client, and you can't do it without scoping. Almost everyone I've worked with had been burned before: late, over budget, or the vendor never finished. So you're untrusted on day one, and the only way through is to finish the small thing you said you'd finish. Question for the room, since most of you build folder-first: when you go into someone else's system, how do you find the Thursday workaround? The thing that lives in one person's head and in none of the files. That one decides whether the quote was right.