[ SLANG ][ AI ][ NOUN ]
The practice of letting a model spend extra computation while actually answering a query — longer reasoning chains, multiple sampled attempts, self-checking — rather than only investing compute during training.
REAL-WORLD EXAMPLE
"The new model is slower because it's burning inference-time compute on a longer reasoning chain before it answers."
LORE & ORIGIN
Named directly by AI researchers to distinguish this newer lever (more thinking time per query) from the older assumption that a model's capability was fixed once training finished.
FIRST SEEN
2024
PEAK POPULARITY
2025-2026 (Reasoning Model Era)
CURRENT STATUS
Mainstream Slang
Related To:P(doom)Vibes-Based Eval
EQUIVALENT CONCEPTS IN OTHER LANGUAGES
AURA IMPACT ⓘ
+15 AURA
TREND STATUS
↑ HIGH RISING FAST · MODERATE
CULTURE CATEGORY
[ SLANG ][ AI ][ NOUN ]
RELATED TERMS
MENTIONS OVER TIME
2023202420252026
RELATED SEARCHES
TOP PLATFORMS
TikTok
52%
YouTube
20%
Twitter / X
14%
Reddit
9%
Others
5%
WHEN DID YOU FIRST HEAR THIS?
CLICK AN OPTION BELOW TO CAST YOUR VOTE.
[ VOTE TO REVEAL COMMUNITY RESULTS ]