Y Combinator

Mukund & Madhav: How Solving Verification Enabled 7M Apps

Suitebench's multi-agent verification enabled long-horizon autonomy; non-technical builders now ship production software without writing a single line of code.

MUMukundguestJFJared FriedmanhostMJMadhav Jhaguest

Mar 16, 2026·39m·Watch on YouTube ↗

📖 CHAPTERS

1. 1

   Emergent’s breakout growth: 7M apps in 8 months and what that signals

   Jared frames Emergent as one of YC’s fastest-growing companies and asks what drove the inflection. The founders position Emergent as unlocking software creation for domain experts previously blocked by technical barriers.

2. 2

   Twin-founder origin story and the pain point that started it: engineering velocity vs testing

   Mukund and Madhav share their background across Google, Amazon deep learning, and building Dunzo in India. Mukund’s experience managing ~300 engineers highlighted software testing as the major bottleneck to shipping quickly.

3. 3

   From AI testing agents to general coding agents: the verification insight

   They applied to YC with automated testing, then realized solving verification enables long-running autonomy. That insight led to pivoting toward a general coding agent capable of broader software engineering tasks.

4. 4

   Early market timing and getting ahead: Suitebench and foundational agent discoveries

   In early 2024 the ecosystem was nascent—Cursor early, Devin newly launched, Lovable not yet. The team focused on topping Suitebench, which forced rapid learning about multi-agent orchestration, memory, routing, and test-time compute.

5. 5

   Pivoting to non-technical builders: leaving enterprise and shipping a lightweight beta

   Initially they pursued enterprise adoption but found it slow to integrate and iterate. Meanwhile, tools like Lovable and Bolt proved consumer demand, prompting Emergent to repackage its stronger coding agent for non-technical users.

6. 6

   Why second movers can win in AI: model shifts, re-imagining the product, and distribution

   Madhav argues each model generation creates new openings, letting later entrants learn and rebuild assumptions. Emergent differentiated by focusing on production shipping (not just front-end prototypes) and accelerated distribution via influencer marketing.

7. 7

   Building for production, not prototypes: infra choices, stack design, and “agent experience”

   Emergent invested in its own infrastructure and an end-to-end platform that mirrors real engineering teams (testing, deployment, security, hosting). They chose a backend/frontend architecture meant to support serious apps and emphasized empathy for both users and agents.

8. 8

   Staying ahead of foundation models: autonomy, harness design, and multi-model orchestration

   They describe intentionally not over-optimizing for temporary model weaknesses (e.g., JSON parsing) and instead trusting improvements in future model generations. Emergent aims to extract additional performance via harnesses, verification loops, and combining multiple models’ strengths.

9. 9

   Live demo: prompt-to-app workflows, mobile vs web, and non-technical-friendly abstractions

   They demo creating a podcast/interview practice app and show examples built by customers worldwide. The workflow includes clarification questions, sensible defaults, and removing friction like managing API keys.

10. 10

    Dogfooding to the extreme: building an internal Asana/Jira-like tool and versioning without Git

    Emergent uses its own platform to build internal systems, including a full project management tool, saving thousands per month. They discuss collaborative feature requests, controlled releases, and optional GitHub integration while offering simpler versioning for non-technical users.

11. 11

    Hiring and operating lean across SF + Bangalore: ownership, problem-solving, and customer empathy

    They run a small team (core engineering ~12) with most members in Bangalore and a small SF presence. Hiring emphasizes ownership and problem-solving, and everyone—engineers included—does customer support and user calls regularly to stay close to needs.

12. 12

    Is SaaS dead? Personalized, agentic software and the shift in how workflows get consumed

    They argue today’s SaaS model faces pressure from two directions: agents consuming SaaS workflows and customers demanding customized tools. Emergent is seeing meaningful demand for “agentic apps” where users embed agents into their own software.

13. 13

    The future: longer-horizon agent work, verification as the moat, and the ‘niche-of-niches’ economy

    They forecast agents running for many hours (eventually 24/7) with multiple agents coordinating under an overseeing agent. Emergent’s R&D focus is verifiers and fine-tuned verification layers, while customer stories highlight a world where tiny niche apps become viable businesses.