The Problem Solvers: Kay Zhu at Genspark

1. Following Your Heart in the Age of AI

Kay Zhu opens with a story that reveals everything about how he thinks. His son is turning 18 and chose to study commercial dance instead of computer science. Three years ago, Zhu says, he might have pushed the other direction — might have steered his son toward a more “practical” technical path. Now he supports the decision fully. The reason: “AI can help people to do so much of things. You basically should pursue your heart.”

This is not a throwaway anecdote. It captures Zhu’s fundamental thesis about what AI unlocks for human beings. When machines can handle large categories of technical and analytical work, the scarcity that previously made some skills more economically valuable than others starts to shift. The premium on being able to produce certain outputs — code, spreadsheets, written analyses — decreases. What remains scarce, and therefore valuable, is what humans bring that AI cannot: passion, judgment, relationships, creativity, embodied experience.

For a father supporting a son who wants to dance professionally, this is a lived expression of a worldview, not just a product pitch. Zhu’s conviction that AI should free people to follow their hearts is personal before it is professional — and that sequence matters. It suggests his product decisions at Genspark are grounded in a human value he actually believes in, not just a market thesis he’s pitching.

2. What Genspark Actually Builds

Genspark is an all-in-one workspace for white-collar workers. Where most AI tools pick a lane — an AI presentation builder here, an AI spreadsheet tool there — Genspark integrates AI spreadsheets, AI slides, AI documents, and various other work tasks into a single coherent product. The user persona is the “normal white collar worker”: the person spending their days in Microsoft Office or Google Workspace, generating documents and analyses and presentations in service of business decisions.

The bet here is that the fragmentation of AI tools is itself a problem. If you need to switch between five different AI tools to complete a typical workday, the cognitive overhead starts to exceed the productivity gains. An integrated workspace could change that equation — reducing the context-switching cost and letting AI assistance flow more naturally across the different artifact types a knowledge worker produces.

Zhu is explicit that Genspark cannot do everything itself, particularly as a small startup. This is where the partnership with Anthropic becomes strategic rather than just technical. “Working with kind of trusted partner is really, really important,” he says. For a small team trying to build across a broad product surface, relying on a world-class foundation model rather than building model capabilities in-house is a rational allocation of effort. The team’s value-add is the product integration, the user experience, and the domain understanding — not training models.

3. Team Culture as the Only Real Moat

Asked what differentiates Genspark, Zhu gives a counterintuitive answer. “Everybody is talking about moat these days. I don’t really think nowadays anybody have any moat because everything is happening so fast. The only moat is the team’s culture.”

This is a strong claim that deserves unpacking. In a conventional startup analysis, a moat is some structural advantage — proprietary data, network effects, switching costs, a patented technology. Zhu is suggesting that in the current AI environment, all of these can be overtaken by rapid capability improvements, new entrants, or platform shifts. What cannot be copied quickly is the cultural ability to continuously learn, adapt, and execute.

He elaborates: “We always exploring and trying the newest technology to execute faster.” The culture he describes is one of relentless curiosity and speed. When a new model capability arrives, Genspark’s engineers treat it like “an invitation to the party” — they rush to understand what it can do and integrate it into the product. This stance — treating every new development as an opportunity rather than a disruption to manage — is itself a competitive capability.

4. Openness, Trust, and the Tight Feedback Loop

Zhu’s philosophy on partnerships is captured in a striking line: “The sum of the secret you kept today will be worthless, you know, tomorrow.” In a world where information has a short half-life — where the capability you’re protecting today will be surpassed or replicated by next quarter — hoarding information becomes a bad strategy. The value of a secret decays so quickly that the cost of keeping it (reduced collaboration, slower iteration, weaker partnerships) often exceeds the benefit.

Instead, Zhu advocates for radical openness as the mechanism for success. “The success of the partnership, it comes to deep neutral trust. When it’s really working, it’s very tight feedback loop.” The best collaborations — between Genspark and Anthropic, between the product team and customers, between leadership and engineers — are ones where information moves quickly in both directions without friction.

The phrase “tight feedback loop” is crucial. In a fast-moving environment, the team that learns fastest wins. Tight feedback loops accelerate learning. Openness enables tight feedback loops. Therefore, openness accelerates learning. This chain of reasoning underlies Zhu’s entire operating philosophy. It also explains why he is comfortable saying he has no idea what Genspark will look like in two years: “I don’t know. A lot of new things will happen and it will even seem like magic.” He has chosen adaptability over predictability.

🧠 Check your understanding

4 questions · your answers are saved in this browser only

1. 1. Why did Kay Zhu change his mind about his son studying commercial dance instead of computer science?

   He realized computer science jobs are being automated away entirely He believes AI can handle so much technical work that people should now pursue their passions His son showed exceptional talent that made the career risk worthwhile He lost confidence in the value of technical education generally

   Zhu's reasoning is that AI can handle large categories of technical work, which shifts what's scarce and valuable toward human passion, creativity, and judgment — freeing people to follow their hearts.

2. 2. What is Genspark's core product offering?

   An AI coding assistant for software engineers A model training platform for enterprise AI teams An all-in-one AI workspace with spreadsheets, slides, and documents for knowledge workers A customer service automation tool

   Genspark targets normal white-collar workers with an integrated AI workspace covering the major artifact types they produce: spreadsheets, slides, documents, and related work.

3. 3. What does Zhu identify as the only real moat in the current AI landscape?

   Proprietary training data collected from enterprise customers A first-mover advantage in the AI workspace category Exclusive model access through partnerships with foundation model providers Team culture — specifically the cultural ability to explore new technology and execute fast

   Zhu explicitly dismisses conventional moat thinking and says 'the only moat is the team's culture' — meaning the organizational ability to continuously learn and adapt faster than competitors.

4. 4. What does Zhu mean when he says 'the sum of the secret you kept today will be worthless tomorrow'?

   AI models can extract any secret from company databases In fast-moving fields, information has a short half-life, making secrecy less valuable than open collaboration Legal protections for intellectual property are weakening Competitors always discover secrets through reverse engineering

   Zhu is arguing that the value of protected information decays rapidly in a fast-moving AI environment, making radical openness and tight collaborative feedback loops more strategically valuable than secrecy.

Reset quiz

🛠️ Build it yourself

Follow these exact steps to reproduce it yourself · estimated time: ~30 min

Prerequisites

• Access to an AI workspace tool (Genspark, Notion AI, or similar)
• A real work task involving multiple document types

Step 1 — Pick a multi-artifact work task

Choose something you actually need to do that spans document types — for example, researching a topic and then creating a presentation and a one-page summary memo about it.

Step 2 — Use a single AI workspace environment

Complete the entire task without leaving the AI workspace tool. Notice where the integration helps (no re-explaining context) and where it falls short.

Step 3 — Map the feedback loop

After completing the task, write down: what information did the AI need from you, what did it produce, and what did you need to correct? This is your feedback loop map.

Step 4 — Evaluate the “follow your heart” thesis

Reflect: did using AI on this task free up time for higher-value human contribution, or did it create new overhead managing the AI? What would need to change to genuinely free you to do more of what you care about?

Step 5 — Test the culture moat hypothesis

If you were building a competing product, what would be the fastest path to matching this tool’s capabilities? Does “team culture and speed” feel like a durable advantage to you after this exercise?