In July 2025, Genspark launched Super Agent, a no-code platform for building personal AI agents—powered by GPT‑4.1 and OpenAI’s Realtime API—that carry out complete workflows via prompts, without coding. These agents can handle tasks like phone calls, slide decks, animated videos, and data research, orchestrating over 80 integrated tools and nine specialized models. Within 45 days, Genspark achieved $36 million annual recurring revenue (ARR), shipping eight agent features—all driven organically with a 20-person team. In partnership with OpenAI, they leverage GPT‑4.1’s million-token context window, JSON‑structured outputs, and voice-based real-time interaction. This article delves into Genspark’s evolution, architecture, capabilities, business performance, and the broader implications for AI-driven workflows.
In October 2023, OpenAI's image generation model DALL-E 3 was integrated into ChatGPT Plus and ChatGPT Enterprise. The integration was using ChatGPT to write prompts for DALL-E guided by conversation with users.[50][51]
Originally launched as an AI-powered search engine aimed at structuring and synthesizing information for business users, Genspark found that users weren’t satisfied with mere answers—they sought actionable outcomes. According to Kay Zhu, co‑founder and CTO, users wanted deliverables like pitch decks, emails, follow-ups, or creative outputs—not just summaries.
In April 2025, Genspark pivoted decisively, embracing agentic AI. Their new product, Super Agent, launched as a no-code platform enabling users to instruct an AI via natural language and get fully executed workflows—ranging from calls to slides, summaries to videos. The shift reflected both evolving user expectations and rapidly expanding AI capabilities, including multimodal inputs, longer context, and improved reasoning.
Super Agent orchestrates nine specialized LLMs and over 80 integrated tools to accomplish user prompts. GPT‑4.1 serves as the core reasoning engine, coordinating tasks across models tailored for research, summarization, image and video generation, voice interaction, and formatting. The system dynamically routes sub-tasks to appropriate components for efficiency and accuracy.
Key capabilities of GPT‑4.1 that enable Genspark’s agentic workflows include:
A 1 million-token context window, allowing agents to ingest entire reports, data sets, or long documents in one pass.
Improved instruction-following, enabling precise structured outputs in JSON.
Automatic prompt caching, reducing API costs, latency, and redundant processing across multi-step workflows.
These features support workflows like analyzing large documents, generating structured slide decks, or orchestrating voice conversations with dynamic context retention.
One standout feature is Call For Me, enabling agents to make real-world phone calls using speech‑to‑speech powered by OpenAI’s Realtime API. The system uses two layers:
A Realtime agent for immediate voice interaction.
A shadow model that monitors ongoing dialogue and guides strategy via message queues.
This architecture allows agents to handle live human conversations—like scheduling, cancellations, or confirmations—with minimal glitches, even in noisy or ambiguous call environments.
Users can type simple prompts such as:
“Call my dentist to reschedule.”
“Create a pitch deck about solar energy.”
“Turn this recipe into a short animated video.”
Super Agent handles the rest: generating slides, stylized graphic covers, voice calls, or scene-by-scene scripts with video delivery. Videos, spreadsheets, or presentation decks are produced end-to-end without manual intervention.
Agents can generate animated short videos, visual slide covers, voice interactions, and document summaries. GPT‑image‑1 is used for image generation. The combination of multimodal inputs and outputs removes the friction of integrating separate tools.
Thanks to GPT‑4.1’s strict JSON output format, agents produce clean, machine-readable data—ideal for downstream tools like spreadsheets or APIs. Outputs can be fed directly into presentations, databases, or analysis workflows, removing the need for manual reformatting.
Within 45 days of launching Super Agent, Genspark reached $36 million ARR, despite having a 20-person team and zero paid advertising. Growth was entirely organic, driven by product virality and user-led expansion. Eight major agent features were launched in just 70 days.
Users praised Super Agent for making automation accessible, with no coding barrier—turning complex tasks into seamless workflows through plain-language prompts. LinkedInReddit
Genspark credits close collaboration with OpenAI—including regular contact with solution architects—as a core enabler of rapid scaling. OpenAI’s APIs, documentation, and multimodal model capabilities expedited development and debugging, helping Genspark ship faster than expected.
OpenAI highlighted Genspark’s launch on its official “Stories” page, branding Super Agent as a signature example of practical AI-powered automation for business and personal users.
Super Agent represents a major step in making agent-based automation usable by non-technical users. No coding, no API keys, no configuration—just plain-language prompts. This democratization opens AI agentic workflows to individuals, startups, small businesses, and creatives. RedditReddit
Instead of juggling multiple specialized tools (video editors, slide software, voice APIs), users can rely on Super Agent as a single unified agent, capable of combining research, generating assets, and executing live interactions seamlessly.
Prompt caching and strict structured outputs minimize latency and costs, enabling professional-grade automations even on low budgets. The agent architecture supports rapid deployment and scalability.
Resignation calls handled autonomously.
Animated social media videos produced from a single prompt.
Lead generation and data summarization workflows that deliver insights and formatted outputs.
These use cases highlight how agents can serve as extended personal assistants across domains.
While impressive, the model is not without limitations:
Agents making live phone calls or client communications must behave reliably. Misunderstandings, hallucinations, or tone mismatches in voice interaction could lead to reputational or operational issues.
Handling sensitive data—like contacts, documents, or proprietary content—requires clear data governance. Users need assurances about data retention, model training access, and security.
Although GPT‑4.1 is powerful, ambiguous or imprecise prompts can produce incorrect outcomes. Super Agent must guide users toward clearer task definitions and offer validation.
While the product includes a free tier (e.g. 200 credits/day in some community posts), sustained heavy use could incur significant costs—both for users and the platform.
Genspark plans to roll out further capabilities, such as an AI browser that acts on web content and AI-powered documents (e.g. formatted PDF reports or contracts). These would extend agentic abilities across browsing, document authoring, and interactive workflows.
Future updates may allow users to fine-tune agents with custom knowledge sources, personality traits, or domain templates—turning Super Agent into a personalized assistant over time.
Genspark could adapt Super Agent for enterprise settings—e.g., internal helpdesk agents, onboarding flows, sales outreach, and document automation. Enterprise customization, integration, and privacy controls will be essential.
Though agents are no-code by design, a developer ecosystem of plug-ins or templates could emerge—allowing technical users to build reusable workflows or extend agent functionality while still leveraging GPT‑4.1.
Super Agent exemplifies the shift from conversational assistants that produce content to agents that drive real-world outcomes. Users expect not just answered questions but completed tasks and deliverables. Genspark’s success is part of a broader move toward agent-first products.
OpenAI’s Multimodal APIs and Realtime platform not only power agents but also provide core infrastructure for startups like Genspark to build and scale rapidly. Their support has lowered development friction and accelerated product-market fit.
Achieving $36 million ARR with no marketing and a 20-person team underscores strong product-market alignment based on genuine value—suggestive of a product resonating deeply with a real-world need.
Genspark’s launch of Super Agent marks an inflection point in AI-driven productivity: a leap from search-based interfaces to fully agentic, no-code automation powered by GPT‑4.1 and the OpenAI Realtime API. By enabling users to turn plain-language prompts into complete workflows—calls, presentations, videos, summaries—Genspark is democratizing access to agentic AI at unprecedented speed.
With rapid organic growth, multimodal integration, structured output formats, and a lean team, Super Agent sets a new benchmark for what next-gen personal AI assistants can achieve. While challenges remain—around trust, data governance, and task ambiguity—the platform’s early success showcases the transformative potential of combining OpenAI’s large-model capabilities with user-friendly, outcome-first design.
As agent-driven applications continue to proliferate, Genspark may become a model for how AI startups can build polished automation experiences, powered by OpenAI, that meaningfully amplify human capabilities—without requiring coding skills.
Generative AI has the potential to revolutionize actuarial science by transforming how actuaries eng...
This comparative study highlights that no single LLM dominates across all NLP tasks. ChatGPT shines...
