Completely agree, most failures I’ve seen come from poor context management and unclear data flow, not the model itself. State handling also becomes a major issue when workflows scale, especially with multiple tools and agents interacting. In my experience, debugging improves a lot once you treat it as a system design problem rather than just an AI model issue.

Hmm, I think AI tools are actually pretty helpful, but you still have to double-check everything — they’re not perfect 🙂

Most companies haven't answered a basic question yet: who is accountable when an AI agent takes an action? Until that's resolved, they'll keep defaulting to safe, surface-level AI features instead of truly rethinking workflows. The bottleneck isn't the technology; it's the accountability layer nobody wants to own.

API docs get attention. The frontend/API contract usually doesn't. TypeScript helps, but types lie without runtime validation. The API returns an unexpected null, a renamed field, an edge case you never tested and your types had no idea. Zod fixes this. Parse at the boundary. If the API changes shape, you catch it at the schema. Not in a Sentry alert a week later. We do this with Next.js Server Actions too. The server/client boundary is the natural place to validate. Keep the schema next to the call. Documentation problem and type-safety problem are usually the same problem.

Yeah, when I was freelancing a lot of time was spent understanding the existing code in the projects I was being asked to update.

Intersting learning experience. The way you explained Python features makes it easy to connect with real learning experiences.