Valerium ∇

Today I'm coming out of stealth to launch the doc2doodle beta. I've been working on it for the last 3 months. The idea is simple: give us a document, and we'll turn it into a doodle video. Claude can do this, but it uses a lot of credits and isn't cheap. So I built a template-based rendering system, with AI handling image and library searches. It's live now. The first 1,000 customers get 25% off their subscription. doc2doodle.com

A build-in-public update should show the decision behind the metric, not just the metric. If you're building in #buildinpublic, let's #connect!

Customer feedback is useful when you can tie it to a real attempt. 1. Ask what they were trying to finish 2. Watch where they get stuck 3. Write down what they do next What do you watch for when a user gets stuck?

I'd test an AI model on the customer's messy input, not the clean demo prompt. Real files arrive with gaps Instructions contradict each other The output needs editing That's where the product either earns trust or creates work. What real input do you use to test your AI feature?

The most useful build-in-public post might be the feature you cut. It shows what you chose not to build. Share the cut or the launch? If you're building in #buildinpublic, let's #connect!!

Build-in-public screenshots make a feature look finished. The browser tab labeled "localhost" has other plans.

AI founders, what is plan B if the model you built around ships inside ChatGPT?

Some founders don't hate fundraising. They hate explaining why the business needs money before they know what the money will prove. If you're building in #tech, let's #connect!!

Which SaaS task do you still handle manually?

You announce an AI feature before it works reliably. People try it. It fails on their real task. Now the demo has set expectations your product can't meet. The launch post can wait until the workflow holds up. If you're building in #tech, let's #connect!!

What was the first thing you charged for in your SaaS?

SaaS founders today are basically: 90% shipping 5% talking to customers 5% hoping the launch post finds buyers If you're building in #tech, let's #connect!!!

Show annual pricing upfront. Hiding it makes the monthly option harder to compare.

What happened to the first AI feature you shipped?

Founders should write the first version of the sales page themselves. You learn what customers actually buy before you hand the story to someone else. If you're building in #buildinpublic, let's #connect!!

Which SaaS metric is hardest to improve? - activation - retention - expansion

Founder problem: You can ship fixes so quickly that the onboarding screenshots are outdated before a new user sees them. If you're building in #tech, let's #connect!!

Claude drafts the feature spec. I spend an hour removing features it confidently invented.

I would not call an AI feature defensible just because it uses a strong model. If another SaaS can add the same prompt and API, customers have little reason to stay with yours. The edge has to come from the workflow, context, or distribution around the model.

Nobody warns you what building in public leads to. - Drafting launch posts - Explaining a rollback - Screenshot archaeology Very public. The changelog needs a changelog. If you're building in #buildinpublic, let's #connect!!