Captions

Gorav Misra came to Captions after spending years at Snap, where he witnessed firsthand how a singular visionary leader (CEO Evan Spiegel) combined obsessive attention to product detail with a small, empowered design team to create one of the last breakthrough social networks. At Snap, Gorav eventually led the design engineering function—a unique hybrid role merging design and engineering—which enabled rapid prototyping and validation of new features at scale. That experience crystallized a core insight: the best consumer products come from deep user understanding, fearless innovation, and the ability to move fast without getting bogged down in process.

When Gorav launched Captions, the timing was perfect. Generative AI had just cracked video manipulation in ways that made impossible things possible. The insight was deceptively simple: most people want to create videos but can't because they lack the skills, equipment, or time. AI could solve that instantly.

Captions was built with speed and scrappiness as core principles. Rather than trying to build a "complete" product, Gorav applied the same lean methodology he'd refined at Snap: ship small, validated features fast, learn from user response, and iterate. The engineering philosophy is brutally simple—every engineer ships a marketable product or feature every single week. Not a partial feature. Not an infrastructure task. A complete, launchable, testable product that users might subscribe to the app just to use.

This sounds chaotic, but it's actually disciplined. The discipline comes from ruthless scope cutting. When designing a feature like "add images to videos," the team doesn't start by building background removal, hue/saturation controls, cloud import, and native pickers. Instead, they ask: what's the absolute minimum? Native camera roll picker. No UI. That's it. If that's not useful, nothing built on top of it matters. Users come in, try it, and complain about exactly what's missing. Next week, those complaints become features.

For longer-term technical debt and infrastructure work, Captions dedicates Q4 annually as "infrastructure quarter." Gorav reframes technical debt as strategic leverage—the way startups operate faster than big companies. Big companies pay down debt immediately or pay back debt from their startup days. Startups should take on debt strategically, pushing problems to future engineers who will have more resources to solve them. The key metric: if you're paying more than 1-2% interest per day in bugs and crashes, you've taken on too much. If you hit 80-90% interest, you're just keeping the lights on.

Captions went viral through a combination of product virality and social proof. One early breakthrough came from the "eye contact" feature—an AI capability (built with NVIDIA) that shifts a creator's eyes to look at the camera instead of at a script. This solved a real problem: people recording videos often look off-screen, breaking engagement. The social media post demonstrating this feature went massively viral, getting remade in multiple languages and accumulating millions of views. That single demo validated the entire premise: AI video features, when genuinely useful, spread organically.

The virality wasn't accidental. Captions' product team spends significant time on social media watching trends, measuring what resonates, and identifying what has organic spread potential. They use virality as a signal of user demand—if something is getting shared, reshared, and remade, there's genuine user interest. This directly informed roadmap prioritization.

What worked was relentless focus on solving real problems, not chasing hype. While the AI space is flooded with gimmicky features, Captions stayed anchored on democratizing video creation. The team maintains two roadmaps: a public roadmap (user-requested features like background removal, undo/redo, longer videos) and a secret roadmap (internally generated ideas nobody asked for, designed to change user behavior fundamentally).

The secret roadmap comes from quarterly all-hands brainstorming sessions involving everyone—engineers, designers, PMs, marketing, recruiting. The company harvests signal from all the noise (social media trends, model advancements, user behavior shifts) and synthesizes it into ideas that nobody else is thinking about yet. Eye contact was one such idea. These are the features that drive outsized growth.

What also worked was the weekly ship cadence. Shipping constantly keeps the product fresh, keeps the company in the news, and most importantly, keeps users engaged and commenting. When users complain, the team knows they care enough to stick around. Constant iteration also means more shots on goal—many features won't work, but enough will that you maintain momentum.

Captions now has over 10 million users and has raised over $100 million. The company is operating at a scale where the original constraints—a tiny design team running everything through the CEO—no longer work. But Gorav has preserved the core of what made Snap successful: deep, intuitive user understanding, empowered small teams, and a relentless commitment to innovation over incremental feature work.

The company is also fully leveraging AI development tools (Cursor, Devon) to multiply engineer productivity, allowing the weekly ship cadence to scale. As Gorav notes, this is something a startup can do that big companies can't—they can pull in an AI agent engineer (Devon) in minutes; big companies need 30 lawyers in a room first.

The broader lesson Gorav brings to startups in the AI era is that the fundamentals haven't changed: identify a user problem, apply technology to solve it, find people with that problem. What's changed is that AI has expanded what's technologically possible. The real skill is knowing what to build and what not to build, staying focused amid constant noise, and maintaining the bias toward action that lets startups compete with bigger, slower organizations.