Last Friday, Cedra Talks brought together builders from across the ecosystem to discuss how AI is changing software development, learning, and product creation.
The conversation was moderated by Diana King, Core Marketing Contributor at Cedra, and featured:
- James (Nova Wallet and Nova Desk)
- Francisco (Nova Wallet and Nova Desk)
- Vsevolod (Mangusta)
From AI-assisted coding and learning new technologies to the future of development workflows, the speakers shared their firsthand experiences building products with AI tools.
Diana: Welcome everyone to another Cedra Talks. Today we're discussing one of the biggest shifts happening in technology right now: AI-powered development, vibe coding, and how AI tools are changing the way builders work.
Before we dive in, could each of you briefly introduce yourselves and tell us what you're building within the Cedra ecosystem?
James: We're building a native desktop experience for Cedra. Our goal is to make interacting with the ecosystem more accessible and efficient for users and developers.
Francisco: Nova's main products are Nova Wallet for mobile and Nova Desk for desktop. Beyond wallets, we're building infrastructure, tools, and applications that make it easier for developers to build within the ecosystem.
Vsevolod: Mangusta is building an NFT marketplace and social reputation layer on Cedra. We want NFTs, user profiles, and reputation to become part of a broader identity experience across the ecosystem.
Diana: How did AI become part of your development workflow, and what was the first task you trusted it with?
James: AI has been part of my journey from the beginning. I'm self-taught, and I started learning to code around the time ChatGPT became available. At first, I used it to explain code, debug problems, and help me understand concepts. It felt less like a coding tool and more like having a patient teacher available whenever I needed one.
Francisco: My first use cases were code completion and logic assistance. AI helped reduce manual errors and speed up repetitive work. Over time, it became a much bigger part of my workflow.
Vsevolod: I started programming before AI tools were widely available. Back then, learning often meant writing everything manually. Once AI became accessible, I began using it for syntax help, prototyping, and breaking complex problems into manageable pieces. Today, tools like Cursor are integrated into my daily workflow.
Diana: Has AI changed the way you learn new programming languages and technologies?
James: Completely. AI allows me to learn by doing. Instead of reading an entire textbook before starting, I can ask questions as I build. Whether it's Rust, TypeScript, Docker, or zero-knowledge proofs, AI makes learning interactive.
Francisco: Previously, learning a new framework meant building test projects and spending significant time understanding documentation. AI accelerates that process. You can quickly understand concepts, generate examples, and move faster.
Vsevolod: AI has effectively given every developer a personal tutor. You can ask follow-up questions endlessly until a concept finally clicks. The focus shifts from memorizing syntax to understanding systems, architecture, and design patterns.
Diana: Let's imagine a world where AI tools suddenly disappear. How much would your workflow change?
James: For me, it would be a massive change. AI isn't just an autocomplete tool anymore. It's part of how I think, research, and learn. Losing it would significantly slow down my ability to build and experiment.
Francisco: Development would become much more time-consuming. Tasks that now take an hour could easily return to taking an entire day. Planning, implementation, and testing would all become slower.
Vsevolod: We would essentially return to previous workflows. Development would continue, but we'd lose a huge amount of efficiency. AI allows builders to focus on solving problems instead of spending time on repetitive implementation details.
Diana: Have you ever shipped something entirely generated by AI?
James: Not really. AI can help create code quickly, but shipping production-ready software still requires understanding what you're building. Intent alone isn't enough. Developers need to review, validate, and take responsibility for the final product.
Francisco: Today's AI agents are capable of generating large portions of a product. In some cases, you can build almost everything through prompts. However, having development knowledge remains critical, especially when it comes to security, architecture, and quality control.
Vsevolod: AI is an incredible accelerator, but developers still need to think critically. The goal isn't replacing engineering judgment. The goal is removing repetitive work so builders can focus on more complex challenges.
Diana: Let's move from general experiences to day-to-day development. Which parts of your workflow benefit the most from AI?
James: For me, it's the entire process—from planning and brainstorming to implementation. AI helps me iterate quickly, validate ideas, and move from concept to execution much faster. It has become an essential part of how I build software.
Francisco: My workflow is more collaborative with AI. I describe what I want to build, AI asks clarifying questions, and together we refine the implementation plan. Once the coding starts, however, I personally review everything. Deployment also remains entirely manual. AI is a great assistant, but the final responsibility is still mine.
Vsevolod: I use AI across many different tasks, but I always think about efficiency. If writing something manually is faster, I'll do it myself. AI is especially useful for generating repetitive code, documentation, testing, and routine backend integrations. Afterwards, I often refactor the generated code to improve readability and maintainability.
Diana: What's one thing AI is surprisingly good at? And what's one area where it still struggles?
James: One thing that constantly surprises me is how well AI adapts to the way each developer thinks. It feels like it understands my workflow and explains things in a way that's natural for me.
Francisco: AI performs exceptionally well when building web applications because that's where it has the most training data. Desktop applications, however, are often more challenging. Building native desktop software still requires much more manual refinement.
Vsevolod: AI doesn't always follow best practices unless you explicitly ask it to. It can duplicate code, ignore reusable components, or overcomplicate simple problems. Prompting makes a huge difference. Sometimes the real issue is giving AI tasks that are too vague. The clearer your instructions, the better the results.
Diana: Security is always an important topic. How often do you review AI-generated code? Have you ever found serious issues during review?
James: Always. At Nova, every pull request is reviewed by another developer before it reaches production. AI has actually become useful for security reviews as well—the latest models can identify vulnerabilities, suggest improvements, and even look for ways to break your own code. Still, every suggestion needs to be verified by humans.
Francisco: Code review is non-negotiable. When we first started using AI, it regularly introduced security issues that had to be fixed manually. Today's models have improved significantly, but developers still need to understand exactly what the code is doing. Experience remains essential.
Vsevolod: My approach is simple: don't trust—verify. AI is an excellent assistant, but security, infrastructure, and production systems should always be reviewed by experienced developers before deployment.
Diana: You have years of development experience. Do you think AI is helping junior developers or making things more difficult for them?
Francisco: AI isn't the problem. The problem is relying on it without understanding what's happening behind the scenes. Junior developers still need mentors who can teach software architecture, security, clean code, and engineering principles. Otherwise, they risk building applications they don't truly understand.
James: I actually think the role of a junior developer is changing. AI will become as fundamental as a keyboard or IDE. Future developers will simply grow up working alongside AI from day one. The skill won't be avoiding AI—it will be knowing how to collaborate with it effectively.
Vsevolod: I think it's both helping and changing the learning process. Beginners should still write code themselves first to understand the fundamentals. Once you know how systems work, AI becomes an incredible mentor and productivity tool rather than a shortcut.
Diana: AI is improving every year. Which developer skills will remain valuable regardless of how capable these models become?
Francisco: Strong programming fundamentals will always matter. Developers need to understand architecture, security, and what's happening behind the code AI generates. Creativity will also remain one of the most valuable skills.
Vsevolod: Critical thinking, communication, and continuous learning will become even more important. AI is a tool, not a replacement for engineering knowledge. Developers who understand computer science and can work effectively with people will always have an advantage.
James: AI should amplify developers, not replace them. The better you understand systems and know how to guide AI, the more valuable you'll become.
Closing Remarks
Diana: AI is transforming the way software is built, but today's discussion made one thing clear: while tools continue to evolve, strong engineering fundamentals, critical thinking, and human creativity remain irreplaceable.
A huge thank you to James, Francisco, and Vsevolod for sharing their experiences and practical insights into AI-assisted development.
We'll continue these conversations in future editions of Cedra Talks, exploring the technologies, ideas, and builders shaping the Cedra ecosystem. Until next time!
Final Thoughts
One theme emerged throughout the conversation: AI is no longer just a productivity tool. It has become an active collaborator in the development process.
While the speakers had different backgrounds and experiences, they all agreed on one thing: AI has fundamentally changed how builders learn, prototype, and ship products.
The future may not belong to developers who write the most code, but to those who can best combine human judgment with increasingly capable AI systems.
Stay tuned for more Cedra Talks as we continue exploring the technologies, builders, and ideas shaping the Cedra ecosystem.