No-code personal agents, powered by GPT-4.1 and Realtime API
OpenAI
Genspark built a no-code AI agent product in 45 days using GPT-4.1 and OpenAI's Realtime API. It's already pulling in $36M in annual revenue, which is stupid fast for a launch.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Forty-five days. That's the entire build window Genspark gave itself to ship a personal-agent product, and somehow the thing is now generating $36 million in annual recurring revenue. For context, plenty of funded startups spend that long just picking a logo. Genspark skipped the drama and went straight to shipping, leaning on GPT-4.1 for the reasoning work and OpenAI's Realtime API for the voice layer.
The pitch is simple: no-code agents that regular people can spin up without touching a line of Python. Users describe what they want handled, and the agent goes and does it, talking back and forth in real time rather than making people wait on a spinner. That's the part Realtime API is doing the heavy lifting on — it strips out the lag that used to make voice AI feel like talking to a bad phone line from 2003.
GPT-4.1 is doing the planning and task execution underneath, deciding what steps an agent needs to take to actually finish a job instead of just answering a question. That distinction matters more than it sounds. A chatbot that explains how to book a flight is trivia. An agent that books the flight is a product people will pay for, and pay for immediately, which the revenue number suggests happened here.
What's notable isn't the tech stack itself — GPT-4.1 and Realtime API are available to anyone with an API key — it's the speed of assembly. Genspark didn't build a foundation model, didn't build infrastructure, didn't spend months on evals. They wired existing pieces together and found a market that was ready to pay before the ink dried. That's the actual story: the barrier to building a serious AI business isn't research anymore, it's just execution speed.
OpenAI is clearly using this as a showcase for what its APIs can do outside a chat window, and it's a good one. But the bigger signal is for anyone still treating no-code agent tools as toys. $36M in ARR in a month and a half says the toys are starting to work.
My take — AI-written commentary, not fact-checked reporting
I'll believe the $36M number when it survives a full renewal cycle, because early ARR on hyped AI products has a habit of evaporating once the novelty wears off and support costs show up. That said, this is a genuinely useful data point against the 'wrap around GPT-4 isn't a real business' crowd — wrappers built fast and aimed at a real workflow can print money, at least for now. The bigger story here is that OpenAI keeps winning by being the boring, reliable infrastructure layer while everyone else races to build the flashy thing on top, which is exactly the position you want if you're the one selling the picks and shovels.
Read more about this at: OpenAI