Crazy numbers this week. I went from 500 followers to nearly 1,700. It's awesome to connect with founders and builders. My timeline is much better now! Here's to the next 1,000. It's still Day 1! 🚀🚀🚀

Building AI products requires a completely different way of thinking than SaaS. You don’t ship a feature, you enable customers to build their features. Workflows are a great example: it you're building AI, you don’t create specific workflows, you create the ability for customers to create their own workflows. SaaS knowledge doesn’t directly translate into AI SaaS knowledge.

I use Opus 5.5 for 90% of all tasks now. It’s incredibly creative, smart and great at instruction following. What’s your go-to model now and why?

So when are we officially starting to call it “Super Intelligence”?

If Claude started doing ads now, what would OpenAI say in their super bowl ad?

I never bought SaaS from paid ads. Here are the latest products I bought and where they came from: → Origami: X organic content → Mailforge: Reddit post → Clerk: Reddit post → Nango: ChatGPT recommendation → Composio: saw in another product For early stage, there’s no alpha in paid ads. Literally every other way to let people know about your business is better.

How much per month are you spending in AI tooling?

We need tasteful interactions again. You guys are bringing it back.

If this is true, it’s pretty outstanding.

The one thing I tell myself when I feel hopeless is that it’ll pass. In 7 days, what I’m going through probably won’t matter. What do you do on those moments?

There’s a concept called “right to win”: the reason why your product will win. What’s the reason why your product will eventually win?

Peak usability: my Grok Bot launched a cloud agents based off an issue with a Posthog feature flag. How long till I need only 1 chatbot to orchestrate literally everything?

The great upside of AI is to be able to work in many projects at once. The great downside of AI is to be able to work in many projects at once. It's either a productivity hack or a way to dilute your attention and lower your span. Which one do you put yourself in?

Codex & Claude are awesome, but Cursor only focuses on coding. Because of it, they crate specific workflows that help the development lifecycle so much. I barely need to think about release management, testing an documentation when I'm in Cursor.

I need to be more ambitious with the tasks I give my agents. They can accomplish so much more than anyone imagines. Are you being ambitious with the tasks you give your agents?

What's the value in brainstorming with AI if it'll just tell you from its past data? There's no novelty there.

My process to shipping a feature these days 👇 → Talk to customers (a lot) → Use @deck_io to find the most impactful problems to solve → Think about the problem, design a solution for MCP and API → Launch in API and MCP Post first launch → Deck learns from API/MCP feedback then generates a project and specs to @claudeai Design to explore UI → once I’m happy with the UI, pass to Claude Code to implement the design How much has your process changed with AI?

It’s cheaper to build and market anything. Why are then VCs needed?

Would you buy a product when it clearly looks like AI slop?