very surprising that fifteen years of hardcore computer vision research contributed ~nothing toward AGI except better optimizers we still don't have models that get smarter when we give them eyes
RESEARCH
-

Summit and Lobster Outperform Qwen3-Coder in Latest Tests
By
–
This is interesting, in my tests, Summit and Lobster (never got Zenith) were way better than Qwen3-Coder every single time. Expect that whatever @OpenAI model version makes it to the leaderboard will be miles above everything else. Nectarine and Starfish around Kiki K2 level
-

Diffusion Models Are Evolutionary Algorithms Study
By
–
Diffusion Models are Evolutionary Algorithms Zhang et al.: https://
arxiv.org/abs/2410.02543 #ArtificialIntelligence #DeepLearning #MachineLearning -

Grace Hopper: Computing Career Spanning Four Decades
By
–
To give context on age, Grace Hopper began computing at 38, completed the first compiler at 46, helped shape COBOL at 53, kept developing COBOL for the Navy in her 70s, retired from the Navy at 80 & then became a consultant for the Digital Equipment Corporation. v/
@cooperx86 & -
Running Convolutions in Frequency Space Like a Federation Science Officer
By
–
It is a trope that TV technobabble usually involves “reversing the frequencies” or “modulating the frequencies”, so I feel like a federation science officer when I declare “I am running the convolutions in frequency space!”
-
High Schoolers Beat World’s Smartest AI Models at Math Olympiad
By
–
https://
wsj.com/tech/ai/imo-go
ld-math-olympiad-google-deepmind-openai-2450095e?st=eQPApK
… via @WSJ The High-Schoolers Who Just Beat the World’s Smartest #AI Models — maybe for the last time I think it was very low class for OpenAI to "compete" in the event without actually entering it and cooperating with officials, and then to announce the -

Open Source AI Models Show Inconsistent Performance vs DeepSeek
By
–
These new open source models (GLM, Kimi) continue to be odd. Great stats, some solid performances, but also fail tests that DeepSeek & smaller closed models have beaten for months.
-

Learning to Plan from Raw Pixels for Autonomous Driving
By
–
PRIX Learning to Plan from Raw Pixels for End-to-End Autonomous Driving
-
Graphene Promises Versus Reality: Manufacturing Challenges Persist
By
–
El grafeno sigue siendo una promesa más que una realidad. Los métodos de fabricación son caros, lentos o poco sostenibles.
— Juan Merodio (@juanmerodio) 28 juillet 2025
Resultado: muchas expectativas… pero pocas aplicaciones reales y nosotros con el teléfono de siempre 🥺 pic.twitter.com/8HjKphSQMWEl grafeno sigue siendo una promesa más que una realidad. Los métodos de fabricación son caros, lentos o poco sostenibles. Resultado: muchas expectativas… pero pocas aplicaciones reales y nosotros con el teléfono de siempre
-
AI Early Warning System Prevents Student Dropouts in Gujarat
By
–
Instead of reacting to student dropouts, we're preventing them. In Gujarat, our Early Warning System with @EduMinOfGujarat & @UNICEFIndia proactively identifies and supports students at risk. Discover the story behind the system