3 AI trends I learned from my founder (+ how to apply them)

AI is everywhere—your inbox, brainstorming sessions, and late-night doomscrolling.

But outside the tech bubble? Crickets. No one’s talking about it at holiday family dinners.

That’s how you know we’re still early.

At Arc, we’re not waiting for AI to “arrive.” We’re already building with it. And it’s completely changing how we work—faster campaigns, smoother workflows, and better results with fewer resources.

I sat down with Weiting Liu, Arc’s founder (and resident AI optimist), to talk about what’s next, what he’s hearing from the YC community, and how we’re actually using AI to get more done.

1. AI speeds up building and testing (but it’s not perfect)

AI isn’t creeping in—it’s sprinting. And it’s never been easier (or cheaper) to go from idea to product.

Launching a startup is 10x easier than 2 years ago (according to YC founder Rohi Mittal). Tools like Replit let you spin up functioning apps in hours, not weeks.

Takeaway: AI gets you from idea to MVP faster than ever. But don’t expect it to do everything for you—yet.

2. AI gets you to V1 faster, but humans still finish the job

AI is great for rough drafts—copy, visuals, or even landing pages. But you'll still need humans when it comes to the final 20%.

Weiting learned this firsthand:

Same. I used Replit to test a “Wall of Love” customer testimonials page idea. It got me 80% there—but I still needed our developer, Linda, to debug the final step before deploying. That’s still 10 times faster than if I had the developer create the whole page!

We’ve also been using these tools in our marketing team:

• Copy.ai for faster drafts, content briefs, and templates

• Spiral to repurpose content, such as LinkedIn posts into Twitter threads

• Lex for AI-powered word editor

• PhantomBuster for automating social media outreach

These tools save time, but none work on full autopilot!

Takeaway: Treat AI as your first draft assistant—it’s fast but needs real human marketers to refine and launch (and debug).

3. AI wins by going niche—focus on specific problems

The best AI tools aren’t trying to do everything. They’re solving particular problems for specific industries.

Weiting shared a few examples:

• Casetext spent 10 years as a non-AI legal tech company before pivoting to AI-powered legal assistants. Six months later, they sold for $650 million.

• Vapi, an AI voice startup Weiting, met at YC camp, pivoted into AI-powered customer service agents late last year, and raised $20 million this month!

At Arc, we’re applying AI to improve our core product experience to hire top talent faster in these ways:

• AI filters and tags talent profiles to match candidates faster.

• AI-generated job descriptions attract better-fit talent faster.

Takeaway: The best AI products solve one problem really well—go niche instead of trying to do it all. Even if that’s looking at what issues you should solve internally with AI.

Where to start right now

AI isn’t slowing down, and neither should you.

1. Go niche. Find a specific problem AI can solve and focus on, like writing better job descriptions, not overhauling your entire recruiting process.

2. Test quickly. Build rough versions, test fast, and tweak as you go. Be satisfied when AI gets you 80% of the way— even if humans handle the final 20%.

3. Stay human. AI handles the heavy lifting, but people are still driving the strategy and closing the loop.

If you’re looking to hire the world’s top 2% talent, check us out over at Arc.dev.