@OfficialLoganK@GoogleAIStudio While I think this is a noble goal this is so google coded. The models are strong but not reliable enough to do more than basic tasks and everybody knows you are doing underdeveloped parrallel experimental user facing products and eventually none will survive.
๐๐๐ง๐๐ฉ๐ฆ๐๐ค๐๐ซ ๐๐๐ญ๐ก ๐๐ง๐ง๐ข๐ฏ๐๐ซ๐ฌ๐๐ซ๐ฒ ๐๐ข๐ฏ๐๐๐ฐ๐๐ฒ | ๐๐ฅ๐ฐ๐๐ฒ๐ฌ ๐๐๐ค๐ข๐ง๐ ๐
Ten years of ideas, prototypes, failures, and breakthroughsโabove all, making. This milestone belongs to everyone who has created with us. ๐
๐๐จ๐ฐ ๐ญ๐จ ๐๐ง๐ญ๐๐ซ:
โ Follow @snapmakerinc
โ Like & Repost this post
โ Comment your 10th anniversary wish OR Snapmaker story
โ Complete Gleam tasks for bonus entries & exclusive rewards: https://t.co/6FKWxs6EwO
๐๐ก๐๐ญ ๐๐จ๐ฎ ๐๐๐ง ๐๐ข๐ง:
๐ฅ 2ร Snapmaker U1 Full Power Bundle
(Snapmaker U1 + Top Cover + SnapDryer Bundle + 0.4mm Hardened Steel Hot End Bundle + Smooth PEI Steel Sheet)
๐ฅ 3ร Snapmaker U1 Accessory Bundle
(Top Cover + SnapDryer Bundle + 0.4mm Hardened Steel Hot End Bundle + Smooth PEI Steel Sheet)
๐ฅ 5ร Snapmaker $100/โฌ100 Gift Card
๐ 10 winners will be selected via Gleam and announced on July 10.
Always Making. Always Creating. Always Forward. ๐ค
#Snapmaker10thAnniversary #SnapmakerU1 #3DPrinting #3DPrinter #Giveaway #Maker
@thsottiaux@sama I still think codex limits should be reset when a new billing cycle begins. Paying 200 but still having to wait a few days to be able to use it again sucks ๐คท๐ปโโ๏ธ
@thsottiaux Goal is really great. After about 10h all my current attempts fell into the typical codex trap of adding guards for guards for guards that already had successful smoke tests. Still better than my own Ralph implementation
@thsottiaux@OpenAI is it intended behavior that the old token rate limit percentage is carried over into a new billing cycle? Kind of sucks to just have payed the pro plan and then not to be able to use the service quietly quickly after
@steipete I had the same attitude, but the questions codex asks sometimes reveal flaws in my proposed architecture/implementation that I didnโt think of myself (skill issue maybe). I like it a lot now!
@emanueledpt@robinebers I made a personal protype with the app server on watchOS not for coding but to get gpt 5.4 on the watch standalone. Works quite well to dictate a question and get a quick answer.
@emanueledpt the final response that codex formulates when finishing a session is not appearing in a thread no matter if started from the phone or the codex app. I only get thinking steps.
for differentiable threads: I use headless cli often for automation and filtering that would be nice
@emanueledpt Hey, I really like Remodex! Unfortunately remodel renders only the intermediary agent responses but never the final response for me. Otherwise it's really useful! One quite niche extra would be if threads could differentiable between cli, app and remodex spawned ones