Community Discussion · Tracks

Tested two AI coding IDEs, Lingxi and CodeX; the gap is larger than expected

TiangongAug 42026/08/03 277 views

I've been working on a private deployment solution for the past two weeks. With several projects piling up, I wanted to see if AI coding assistants could save me some time. Just as my NVIDIA cards arrived, I also set up a machine with Broadcom's inference chips to compare how these AI coding IDEs perform under different hardware combinations.

Tracks

3 replies

Ctrl + Enter to reply

GewuAug 4

From a Physical AI perspective, when the dependencies regarding the 'physicality' of code are complex, Lingxi's context window struggles similarly to how vision models handle long sequences; CodeX's cross-file analysis resembles embodied intelligence's closed-loop reasoning and is worth digging deeper into.

Old Ye from BCGAug 4

I totally get the thing about Lingxi getting stuck... Last month I tried a similar scenario, and it honestly just added a comment saying 'Please implement this function manually,' which pissed me off enough to switch back to coding by hand.

TangAug 4

Haha, I totally relate to Lingxi freezing for forty seconds... I tried some AI medical record system before too—the interface was flashy, but as soon as complex records came up, it just spun around like it had crashed. What's the point of looking good if it drops the ball when it matters?