Writing C++ with Jev by picking valid next tokens
The author experimented with TypeSafe's Jev decision/classifier model for C++: at each step it was given only grammatically valid next tokens to pick from, avoiding free-form generation.
The post argues that autonomous UI testing agents need to spot visual bugs and contrast issues in sub-second turns, notes that the Jev model cannot accept images, and cites Cloudflare as proof that high-speed decision models are not necessarily blind.
If your autonomous UI testing agent cannot see visual bugs and inspect contrast in sub-second turns, you are shipping broken software to production Cloudflare proved that high-speed decision models do not have to be blind text classifiers While Jev cannot accept images at all,