By 2028, will any coding copilot look more like "pairing with an LLM" than "smart autocomplete"?
Mini
8
321
2028
89%
chance

Resolves 2028-02-01, based on my vibes from the coding I've done in the last month-ish. Resolves YES if any copilot exists that "feels like pairing" and costs under $50 for me to try for an hour.

Things that would make a copilot seem more "pair"-y to me:

  • voice I/O rather than text

  • asking intelligent clarifying questions

  • pointing out potential mistakes I've made, or better ways to organize the code

  • displaying some knowledge of the broad invariants of the system not obvious from local code analysis

Things that seem like "smart autocomplete":

  • its main contribution being suggestions for what to type next

  • being able to extrapolate long function bodies by gluing together code that already exists scattered around the code base

  • being able to do clever type-system constraint-solving to figure stuff out

Get Ṁ1,000 play money