Fable 5 vs. GPT-5.6 Sol on an NP-Hard Problem: Does /goal help?
SMRTR summary
Claude Fable 5 and GPT-5.6 Sol were tested on a fiber-network optimization problem with search spaces exceeding 10^1223 possible solutions. Fable 5 dominated, with tighter, more consistent results. The /goal mode won most individual runs but worsened average performance, since it amplifies whatever strategy the model commits to, good or bad.
SMRTR provides this summary for quick context. The original article belongs to Hacker News.
Read the original article