Spark Program: AI for creative and intellectual autonomy

Today we are opening Spark Program, yeshcube’s research and development line for creating generative AI applications that help people think, practice and create without replacing their intellectual work.

The program starts from a simple idea: a tool can improve an immediate result without developing the ability needed to produce it independently. Spark focuses on that difference. We design systems in which AI asks questions, offers contrast and provides feedback, while the person retains authorship, decisions and responsibility for the outcome.

Why we are opening Spark Program

Generative AI can save time and make certain tasks more accessible. Delegation is not a problem when the sole objective is to complete them efficiently. But when we also want to learn, develop judgment or retain a voice of our own, it matters which parts of the process the person still performs.

Available evidence indicates that interaction design can change the outcome. In a field experiment with nearly one thousand mathematics students, access to a general generative assistant improved performance during practice, but the group performed worse after the tool was removed. A version that provided teacher-designed hints instead of directly supplying answers mitigated that effect (Bastani et al., 2025).

In another study, 319 professionals described 936 real-world uses of generative AI. Greater confidence in AI was associated with less self-reported critical-thinking effort, while cognitive work shifted toward verifying and integrating responses (Lee et al., 2025).

These findings do not show that all generative AI reduces human capability, nor can an effect observed in education be transferred to every context. They do support the question that organizes Spark Program: how should AI intervene to improve the process without taking ownership of it?

How we design AI that activates human capability

Spark Program applications are built around three principles:

Activation rather than substitution. AI may flag a gap, raise an objection or offer an alternative perspective. The final wording and decision remain with the person.

Productive friction. The system does not automatically remove every difficulty. It introduces calibrated questions and constraints when effort is part of the capability we want to practice.

Progression toward autonomy. Assistance must be adjustable and capable of being reduced. The aim is to determine whether the person retains what they practiced after they stop using the tool. Increasing dependence on it falls outside the purpose of the program.

This approach can be applied to writing, critical thinking, creativity, argumentation, decision-making and problem-solving. In every case, we will distinguish between a strong result while using the tool and a capability that remains without assistance.

The spark ignites. It does not execute.

First application in development

Spark Program’s first application focuses on writing with a distinctive voice for professional social networks. Instead of drafting a complete post, it analyzes the person’s own draft, compares it with their style and objectives, detects clichés and asks questions that help them review their own decisions.

This first line allows us to turn the program’s principles into a concrete experience and assess aspects such as intellectual participation, feedback quality, the amount of support needed and transfer to new tasks.

A program open to new projects

We are opening Spark Program to research teams, developers, startups and organizations that want to explore AI applications designed to strengthen human capability.

We are looking for projects that clearly define the capability they want to develop, preserve human authorship and are prepared to assess more than usage or satisfaction. yeshcube contributes methodological design, integration into the hub and a progressive validation framework.

Discover Spark Program →

Collaborate through Allies →

Contact yeshcube →