@rezoundous Try, Run no tests until you have completed the task fully at the root. Upon completion run 1 rebuild and 1 round of tests. If errors are found fix all exposed at the root stage before again running a new rebuild or test suite.
@thsottiaux Last night I fell asleep with codex ultra 5.6 running in vscode. I awoke to the window grayed out I closed vs code reopened it and the last 3 message I sent and any work codex did after was gone. My usage depleted but the work done so I'm not to mad but still a bug.
Are rhe claims of reduced usage being investigated or swept under the rug?
A we are looking into it would go along way more so then heavy moderation on reddit curtailing the reports.
@thsottiaux
@thsottiaux Now that we reached 10m. Is that the end of the random reset era. Or should we stay tuned?
I know bugs a plenty are causing high resource uses. Especially when having it follow phased documents. 75% burn of 100 dollar plan for 2 phases and about 4k loc painful.
Hmm, I asked sol high and glm 5.2 to do the same task. A 11 phased plan. 18 hours later Sol high was usage stopped at phase 5.
GLM 5.2 completed it all in under 4 hrs. With minor correction.
when my usage returns I plan to ask codex to review glms work.
@thsottiaux
@thsottiaux Replied 44minutes after your post with only 2200 replies and didnt get it @thsottiaux did the ai perhaps fail here in deciding who got it? Ps. Bet you wont do a surpise sunday reset =p
@thsottiaux I abadoned a complicated dsp app after months if effort and 100s on other AI failed to get it working. A few hrs on gpt 5.6 sol medium and it finally achieved the quality I was aiming for!