Use AI Like It's Already Superhuman | Felix Rieseberg

Summary · 3 min read

Rieseberg is an engineering manager and member of technical staff at Anthropic, where he works on Claude products, including Claude Cowork and the Claude desktop app. He published How to Be More Ambitious, a tutorial for humans working with models.

Rieseberg argues that AI's larger promise is to expand what people can attempt by extending their knowledge, creativity and agency, rather than only automating existing work. He says products should help users discover ambitious uses and maintain enough interaction for people to feel authorship. He traces Claude Code to an early internal prototype in 2024 and says builders should design for where models may go, not only their current capabilities. The discussion also considers transferable engineering skills, artists' use of coding agents, and how AI could loosen job boundaries.

Minute by minute

1. 0:19

   AI can expand human ambition. Felix Rieseberg says AI is most exciting when it lets people do things they could not do alone, not just automate existing tasks. He cites product manager Pavel's attempt to find a planet in a large dataset, saying the user's ambition impressed him more than the model's ability.

2. 1:59

   Products should expose possibilities. Rieseberg says building a button for every impressive use case would be easy, but the range of possible tasks is too broad. He sees a product challenge in helping people move beyond basic questions and discover what they can attempt with AI.

3. 3:56

   Interaction can strengthen authorship. Rieseberg draws on the IKEA effect and the resistance of artistic materials to argue that taking part in a process can make its result more meaningful. He says people tend to feel more authorship when they work back and forth with Claude than when they leave it to make something on its own.

4. 7:07

   Learning differs from delegation. Rieseberg distinguishes asking Claude to make something with Blender from learning to use Blender. He says Claude can write Python scripts that use Blender as a library, while someone learning the tool would open the app and work through the task with computer use and its Teach me feature.

5. 9:01

   Claude Code anticipated stronger models. Rieseberg says Boris began showing Claude Code internally around September 2024, when Sonnet 3.5 scored in the 30s on SWE-bench and products focused more on human-in-the-loop autocomplete. He says Boris took an optimistic view of future model capability, and that builders should consider what they would make if models became superhuman.

6. 11:10

   Empathy matters more than coding rituals. Rieseberg says clear prompts share skills with people management, and that translating what someone wants into technical requirements is highly transferable. He argues managers should spend less time directing how reports do their jobs and more time on communication and opportunities, and questions whether debates like spaces versus tabs ever improved products.

7. 15:17

   Creative tools can support self-discovery. A host describes using AI and mood boards to explore visual identity; Rieseberg says AI use does not determine whether art reflects its creator. He compares AI with recorded music and photography, arguing that new methods need not diminish the value of older ones. A host says ongoing use of Claude has helped with self-discovery, and Rieseberg agrees that personalization can do that.

8. 19:39

   Guidance can unlock creative output. Rieseberg says choosing between options is easier than starting from a blank canvas, and that guiding a creative client can help them express what they want. He hopes AI will increase creative output, comparing it with 3D printers making it easier for people to try unusual ideas.

9. 21:31

   Coding agents widen access beyond chat. Rieseberg says people are often most excited when AI helps them do something they could not do before; he points to Claude Design users outside full-time design and Claude Code users outside top-tier engineering. He suggests describing coding agents as a way to accomplish new tasks, rather than treating them as just another chat box.

10. 24:42

    AI may loosen job boundaries. Rieseberg says AI makes it easier to have an impact across different fields, and that roles may depend less on credentials such as a computer science degree. He still sees genuine interest as a guide to people's work, while noting that product managers and engineers may differ in whether they prefer prototyping or discussing prototypes.

Published Oct 19, 2026.