Last updated Apr 26, 2026
Hogan, CEO of Inference.net, shared that seasoned coders prefer manual coding amid frustrations with AI tools like GitHub Copilot, which a 2025 METR study showed slowed experts by 19% on tasks despite their perceptions of speed gains. Developers cited AI's 'slop'—plausible but flawed code needing fixes—and a renewed joy in learning languages like Zig or Rust by hand, as shared by figures like Ben Dicken and Theo of t3.gg. While critics note AI helps juniors and prototypes, with adoption at 93% monthly, the discussion reveals top talent prioritizing precision and control over convenience.
This story is a summary of posts on X and may evolve over time. Grok can make mistakes, verify its outputs.