Many users praised the T3 CEO's discovery of major GPU optimization gains that LLMs missed in the codebase, while some voiced disappointment at the models' limitations.
Based on 11 visible X reactions from 73 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@theo 5.6 has been nerfed to oblivion its producing UIs that wouldve looked dated in the 1980s
2:48 AM · Jul 15, 2026@theo Sometimes the biggest wins come from the simplest changes
2:04 AM · Jul 15, 2026Just landed some really, really dumb changes to T3 Code that reduce GPU utilization by ~85%. LLMs failed to find any of them. Both Fable and gpt-5.6 kept getting caught on things that didn't matter. Most weren't even in the render path! They had all the tools they needed, but they got too distracted by the code to solve the real issues. I did get them to be helpful. I had gpt-5.6 write me a super useful script I could paste into the console to enable/disable the CSS features I assumed were pegging the gpu. After a long process of elimination, I identified tailwind class `animate-pulse` + a blur + a grain layer in the bg as the root cause. Fable and Sol didn't do my job, but they did end up helping.
2:01 AM · Jul 15, 2026(to be fair - the bottleneck was mostly Chrome. Debugging tools for heavy gpu workloads is not very good)
2:03 AM · Jul 15, 2026Just landed some really, really dumb changes to T3 Code that reduce GPU utilization by ~85%. LLMs failed to find any of them. Both Fable and gpt-5.6 kept getting caught on things that didn't matter. Most weren't even in the render path! They had all the tools they needed, but they got too distracted by the code to solve the real issues. I did get them to be helpful. I had gpt-5.6 write me a super useful script I could paste into the console to enable/disable the CSS features I assumed were pegging the gpu. After a long process of elimination, I identified tailwind class `animate-pulse` + a blur + a grain layer in the bg as the root cause. Fable and Sol didn't do my job, but they did end up helping.
2:01 AM · Jul 15, 2026(to be fair - the bottleneck was mostly Chrome. Debugging tools for heavy gpu workloads is not very good)
2:03 AM · Jul 15, 2026Many users praised the T3 CEO's discovery of major GPU optimization gains that LLMs missed in the codebase, while some voiced disappointment at the models' limitations.
Based on 11 visible X reactions from 73 accounts; directional sample.
Ask a question below.
Published answers will appear here.
