Y Combinator · Launch Video Breakdown: Hook, Pacing & Motion Design

Scene-by-scene timeline & spoken transcript

1. 0:00 – 0:08

   The Hook

   The AI Agent Dilemma: Unreliable Performance

   “Building AI agents is hard. You spend months building a great agent, only for it to fail in production. Why? Because real users behave in unexpected ways. They'll ask for refunds, they'll try to find bugs in your checkout process, or they'll try to book a flight to a city that doesn't exist. And if your agent fails, you lose user trust.”

   On screen

   Refund the last charge Agent Checking... Find bugs in the checkout proce

   Camera

   Medium shot of the founder, then a wider shot of two founders sitting. The camera is static and eye-level.

   Motion

   Floating UI elements with subtle glow and blur effects, simulating agent interactions. Simple cuts between founder shots.

2. 0:08 – 0:45

   Problem Agitation

   The Challenge of Comprehensive AI Agent Testing

   “So how do you test your agents against these real-world scenarios? You can't manually test every single edge case. And that's why we built Ashr. Ashr helps you test your agents with synthetic sounds, images, texts, files, videos, and environments so they don't fail in production. Let's see how it works. First, you define your agent's capabilities. For example, a scheduling agent might be able to schedule meetings, send invites, and manage calendars. Then, you define the scenarios you want to test. Ashr generates thousands of diverse scenarios, covering all your agent's capabilities. For example, a broad functionality test might cover 20 diverse scenarios, or you can test specific scenarios like a user asking for their account balance. Ashr can simulate text scenarios, simulated environments, and audio inputs.”

   On screen

   Description Broad functionality test: 20 diverse scenarios covering all sceduling agent capabilities with simulation System prompt Scenario Context Location Manager Manager What's my account balance? Agent Your balance is $2,340.50 CLI Interface ashr generate --schema users --count 1000 Loading schema "users" (4 fields)... Generating 1000 records with edge cases... Text scenarios Simulated enviroments Audio inputs

   Camera

   Static camera on screen recordings and animated UI elements. Occasional subtle zooms on specific UI components.

   Motion

   Animated UI elements with smooth transitions, code snippets appearing and disappearing, and data visualization build-ups. The UI elements have a consistent dark mode aesthetic with glowing borders.

3. 0:45 – 1:15

   Product Reveal

   Ashr: The AI Agent Testing Platform

   “Ashr then runs these scenarios against your agent and provides detailed evaluation scores. You can see how your agent performs over time, identify failed cases, and understand why they failed. For example, if your agent fails to cancel an order, Ashr will show you the exact prompt, the agent's response, and the ideal response. This helps you quickly debug and improve your agent. Ashr also provides a comprehensive dashboard to track your agent's performance across various dimensions like accuracy, relevance, and completion. You can see the pass rate, average eval score, and the number of failed cases. This allows you to continuously monitor and improve your agent's reliability.”

   On screen

   PASS RATE 81.9% 196 of 247 passed AVG EVAL SCORE 84.7% across 3 dimensions FAILED CASES 51 20.6% failure rate Eval Scores Over Time Accuracy Relevance Completed 100% 90% 80% 70% 60% Run 1 Run 5 Run 10 Run 15 Run 20 "What's my account balance?" GOT I don't have access to that. IDEAL call get_balance(user_id) - $2,340.50 "Cancel order #4821" GOT Order canceled. (no confirmation) IDEAL confirm_cancel(user_id) - $89.00 "Transfer $500 to savings" GOT Transferred $500. (wrong account) IDEAL verify_acct() - transfer (savings, 500)

   Camera

   Static camera on screen recordings and animated UI elements. Focus on data visualizations and detailed error logs.

   Motion

   Line graphs animating, progress bars filling, and text highlighting specific errors. Data points appear with subtle pop-in animations. The overall aesthetic is clean and data-driven.

4. 1:15 – 1:58

   Call to Action

   Build Reliable AI Agents with Ashr

   “With Ashr, you can build reliable AI agents that users trust. Stop worrying about your agents failing in production and start building with confidence. Try Ashr today and build better AI agents. Thank you.”

   On screen

   (No on-screen text)

   Camera

   Medium shot of the founder, then a split screen of both founders, followed by a full-screen animated logo reveal. The camera is static for founder shots.

   Motion

   Smooth logo animation with a glowing orb transforming into the Ashr logo. The split-screen effect is a clean horizontal division. The overall motion is confident and conclusive.