3 months ago
Thurs Jan 29, 2026 5:11pm PST
In vision, a lot of the ideas seem plausible until you sink weeks into them. How do you evaluate whether a research idea has real value versus being an incremental tweak? Are there heuristics you use before committing compute and time?