Codex may have legitimately done something novel using "Idea Descent" today. Gonna write about this one soon.
Funny aside: the spike is me suggesting a stupid idea that it tried, and clearly was bad.
8:22 AM · Jun 7, 2026 · 40.8K Views
Microsoft Research's Dimitris Papailiopoulos used OpenAI Codex to shrink a novel solver from 117.2 kB to under 50 kB via 'Idea Descent
An ineffective manual suggestion temporarily spiked the code size
335761926058.7K
Sentiment
Positive users praise Codex's Idea Descent for novel code optimizations and scientific exploration, while negative users criticize Claude models for declining performance and unreliability.
Pos
59.1%
Neg
40.9%
21 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS12.1KBOOKMARKS26LIKES142RETWEETS5REPLIES15
Dimitris Papailiopoulos@DimitrisPapail
Claude Code with 4.8 was given the same prompt, and the same goal, but kept insisting I was wrong and the goal is not achievable.
My number 1 reason for preferring Codex the past 2 weeks. So frustrating.
Dimitris Papailiopoulos@DimitrisPapail
Codex may have legitimately done something novel using "Idea Descent" today. Gonna write about this one soon.
Funny aside: the spike is me suggesting a stupid idea that it tried, and clearly was bad.
9hViews 12.1KLikes 142Bookmarks 26
Roman Vorushin@vorushin
@DimitrisPapail user: "you did great job. next step - 25kB target. go ahead, you have 10 hours till I wake up"
9hViews 921Likes 1Bookmarks 1
Dimitris Papailiopoulos@DimitrisPapail
@xeophon
Florian Brand@xeophon
@DimitrisPapail The weekly Dimitris-tease-post that will result in a banger blog down the line
9hViews 2.4KLikes 7Bookmarks 0
Florian Brand@xeophon
@DimitrisPapail The weekly Dimitris-tease-post that will result in a banger blog down the line
9hViews 173Likes 2
Dimitris Papailiopoulos@DimitrisPapail
@vorushin i literally asked it to go to 21kb lol
8hViews 703Likes 3
Wilkins Micawber@Me5466255992308
@DimitrisPapail which Claude effort level did you use?
9hViews 45
Patty@Patty_H93
@DimitrisPapail Seeing a lot of people upset with 4.8..is this a sign of model collapse? Or is it impossible to tell because anthropic could be throttling compute? But why would they do that right after a release of best available model ??
9hViews 37
Dan McAteer@daniel_mac8
@DimitrisPapail Interested.
8hViews 347Likes 1
Noah@Noah64165746
@DimitrisPapail What happened with the small target? You wanted it to be as close as possible to 50kB, and not smaller?
9hViews 519
Matt Rickard@mattrickard
@DimitrisPapail i like the concept already
9hViews 518
Nicholas Bardy@NicholasBardy
@DimitrisPapail This is cool, would be happy to chat, I have done something similar sounding with 4D Gaussian splats.
8hViews 505
Alok Bishoyi@alokbishoyi97
@DimitrisPapail would love to lear more about your setup !
9hViews 494
Alpha Batcher@alphabatcher
@DimitrisPapail something new from Codex again?
we love it
7hViews 382
Garima@sincerelycheesy
@DimitrisPapail love when the model has the courtesy to show you exactly where you went wrong
6hViews 326
Daniel Auras@rasdani_
@DimitrisPapail i'm sorry Dave. i'm afraid i can't do that.
4hViews 80Likes 1
Maapu@Rajesh23MD
@DimitrisPapail What’s this graph about?
7hViews 172
Shuying Luo@shuying_luo
@DimitrisPapail I got similar experience! Or it claimed to solve something it never did.
7hViews 71
Aux@__12124__
@DimitrisPapail 2 weeks?? More like 6 months
8hViews 45
Lewis Tunstall@_lewtun
@DimitrisPapail Trying to win the Hutter Prize :P
4hViews 41
Alpha Batcher@alphabatcher
@DimitrisPapail Claude Opus 4.8 the great model ever yet
7hViews 36
Load more posts