The intense, 24/7 engagement with AI tools is unsustainable and leading to burnout. A collective "come down" from the hype is imminent. In the aftermath, the most valuable professional skill will be the ability to disconnect from the noise and engage in deep, focused work for sustained periods.
To avoid burnout on side projects, adopt a paced workflow. Use the work week for asynchronous, back-and-forth planning conversations with an AI, submitting prompts and reviewing plans overnight. This results in a well-defined execution plan ready for focused, human-led implementation on the weekend.
In the age of AI, distinct roles like designer, PM, and engineer are converging. Long-term career success hinges on the ability to fluidly move between these disciplines and focus on shipping good software, rather than being confined by a rigid job title. Obsession with titles is a liability.
Contrary to traditional digital design, the modern AI-assisted workflow involves broad, conceptual exploration on canvas-like tools (e.g., Paper) and sweating the final visual details directly in code. Pixel-nudging in design software like Figma is becoming obsolete for last-mile fit and finish.
The underlying infrastructure for AI agents ('harnesses') becomes obsolete roughly every six months due to rapid advances in AI models. At Notion, this means completely rewriting the harness multiple times a year, demanding a culture comfortable with constantly rebuilding core systems and discarding previous assumptions.
Designing for AI is less about crafting pixel-perfect UIs in Figma and more about creating the underlying system or "harness." This involves enabling the agent to perform long-running tasks, verify its own work, and operate effectively within technical constraints, which is where the real design work lies.
To automate bug fixing, connect an AI agent to your error reporting (Sentry), database (Supabase), and log drains (Acxiom). When a bug is reported, the agent can autonomously replay events from logs, diagnose the root cause of the failure, and eventually fix it, creating a powerful self-healing loop for your application.
Seemingly non-technical prompts like "let's step back and think really hard" or "make it simpler and dumber" are highly effective. They work by adding key concepts to the AI's input context, which forces the model to change its mindset and extrapolate from that new framing, leading to better outputs.
AI models are poor at "last-mile" visual design. However, if a human designer invests heavily in creating a perfect set of primitives (e.g., buttons, cards), AI becomes incredibly effective at reusing and intelligently extrapolating from that foundation for new contexts. Human effort on the core system pays off exponentially.
When building at the frontier of AI, it's a valid strategy to ship imperfect, "vibe-coded" features. This approach assumes that rapid, near-future model improvements will clean up imperfections, making it better to launch an imperfect product now rather than wait for perfect model performance that is just around the corner.