Y CombinatorEmergent: How Six Months of Tinkering Led To A $100M ARR Company
At a glance
WHAT IT’S REALLY ABOUT
Emergent’s rapid rise: tinkering, agents, and global AI execution
- Emergent is an AI-native platform that lets non-programmers chat to build, deploy, and maintain real, monetizable software end-to-end.
- The company grew to 8.5M+ users, 10M+ apps, and $100M+ annualized run-rate roughly nine months after launch, with revenue primarily from the US and Europe.
- Its technical edge came from starting as a coding-agent research effort, topping the SWE-bench benchmark, and building a multi-agent system with memory, testing, design, and orchestration.
- Mukund’s prior startup experience—especially scaling Dunzo—shaped Emergent’s customer obsession, operational rigor, and emphasis on focus after learning what distracted Dunzo from doubling down.
- A key founding strategy was “tinkering” and “living at the edge”: betting on fast model progress, skipping soon-to-be-solved problems (like brittle JSON output), and repeatedly rewriting the system as new model classes emerged.
IDEAS WORTH REMEMBERING
5 ideasBuild for “ship-ready” outcomes, not demos.
Emergent differentiated by completing the full software lifecycle—backend, database, testing, hosting, deployment, and maintenance—because users ultimately pay for working products, not prototypes.
Use benchmarks as a compass when the product is unclear.
Attacking SWE-bench created measurable progress, focused the team during rapid pivots, and generated core innovations (memory, agent communication, test-time compute) later embedded into the product.
Tinkering can be a deliberate startup strategy.
Six months of unpressured experimentation with rapidly evolving models helped surface non-obvious insights about what would soon be possible and which bets would compound.
“Live at the edge” by betting on near-future capability, not current limits.
Mukund’s team assumed exponential improvement, skipped point-solutions that would be obsoleted by the next model iteration, and built toward full software-engineering automation.
Expect to rewrite the system repeatedly in fast-moving AI markets.
Emergent rewrote its architecture multiple times as new model classes arrived, treating each new model as a reason to reimagine workflows, agent roles, and what’s feasible in the next six months.
WORDS WORTH SAVING
5 quotesEmergent is a platform that allows anybody without any programming knowledge to be able to build software that you can actually ship, uh, that your users can use, that you can monetize.
— Mukund Jha
If you remove all the software companies from, um, you know, Nasdaq and, and S&P, you'll see it's been just a flat line. And, and we started thinking, okay, what if we can bring this power to almost everybody in the world?
— Mukund Jha
I told my co- co-founder that, "Hey, I think now we are too big to fail."
— Mukund Jha
I actually got this luxury of six months of like just pure tinkering on things that I really liked with no sort of objective in mind.
— Mukund Jha
Building a company for, for India, a local company versus building a global company is actually exactly same effort.
— Mukund Jha
High quality AI-generated summary created from speaker-labeled transcript.
