What You'll Learn
Fireside Chat Recap: Jonas Templestein at Create With 2025
In this 25 minute fireside chat at Create With 2025, Jonas Templestein shares practical perspectives on building products in the age of AI. The conversation focuses on how teams and founders can combine rapid experimentation, community feedback, and thoughtful design to ship useful AI powered experiences. Below we break down the key themes, actionable takeaways, and how to apply these ideas to your own projects.
Why this conversation matters
Fireside style talks are valuable because they trade polished slide decks for candid experience. Jonas frames product building as a loop of learning: prototype, ship, measure, and iterate. For teams building with AI agents and vibe coding workflows, that loop is essential to reduce risk and get to product market fit faster.
Core themes from the chat
1. Start with a clear customer outcome
Jonas emphasizes starting from the user need rather than the model. AI is a capability, not the product. Define the clear outcome you want for users and design the agent or integration to deliver that outcome reliably.
2. Iterate with low friction prototypes
Use quick prototypes to test hypotheses. Keep early experiments cheaply reversible so you can validate whether AI actually improves the workflow. The faster you can run a real-world test, the clearer the signal from users.
3. Leverage community as a development force
Treat active users and early adopters as co‑designers. Community feedback surfaces edge cases, informs prioritization, and helps you build for real workflows rather than idealized ones.
4. Balance innovation with guardrails
Shipping AI features requires both creativity and responsibility. Design guardrails into early releases and monitor outcomes to prevent negative experiences. Observability and clear failure modes are critical.
5. Design for composability and iteration
Build components that can be swapped as new models or tools emerge. A modular approach lets you iterate on parts of the stack without rewriting the whole system.
Actionable advice you can apply today
- Define the one user outcome your AI feature must reliably deliver in the first release. If the feature cannot do that well, delay additional complexity.
- Prototype with the simplest possible pipeline and run a small closed beta. Use real tasks and workflows so you get useful signals.
- Create a feedback loop with early users. Track qualitative feedback alongside quantitative metrics to decide what to change next.
- Instrument for failures. Log when the agent cannot fulfill a task and capture user recovery paths to improve robustness.
- Keep components modular. Separate prompts, orchestration, and UI so you can replace or upgrade layers independently.
Common pitfalls Jonas warns teams to avoid
- Over engineering the first release. Avoid adding too many bells and whistles until core value is proven.
- Treating the AI as a silver bullet. AI should reduce friction or add capabilities in service of a clear user need, not become product theatre.
- Ignoring deployment observability. Without metrics and logs, it is impossible to know how an AI feature performs at scale.
Example roadmap for an early AI feature
1. Problem definition: Define the user outcome and acceptance criteria.
2. Quick prototype: Build a minimal end to end flow using cheap tooling.
3. Private beta: Ship to a small set of users and collect structured feedback.
4. Measure and iterate: Use both qualitative and quantitative signals to improve.
5. Harden and expand: Add safety nets, monitoring, and scale to more users.
Final thoughts
This fireside chat is a reminder that building with AI is still fundamentally about product thinking. The right experiments, community oriented development, and careful observability will get teams much further than chasing novelty. For creators focused on vibe coding and rapid prototyping, the emphasis on low friction testing and modular design is especially actionable.
If you build with AI, treat your early releases as learning mechanisms and prioritize delivering clear user outcomes. The combination of iterative product development and community engagement is a durable approach for shipping useful AI experiences.
