I want "Neural Networks from Zero to Hero" Karpathy back
MACHINE LEARNING
-
Using AI to psychoanalyze yourself from years of chat logs
By
–
A fun exercise is exporting your 5+ year of chat logs and putting it in Claude Code to do psychoanalysis of yourself
-
AI agents memorize fixed benchmarks, not real conditions
By
–
AI agents trained on a fixed benchmark eventually just memorize it!
— Sumanth (@Sumanth_077) 25 juin 2026
Once an agent learns to pass a fixed set of test scenarios, the benchmark stops teaching it anything new. It's also nothing like the real, messy, unpredictable conditions agents actually operate in.
Patronus… https://t.co/7gl7lDSvJu pic.twitter.com/ZE5ADOAV7UAI agents trained on a fixed benchmark eventually just memorize it! Once an agent learns to pass a fixed set of test scenarios, the benchmark stops teaching it anything new. It's also nothing like the real, messy, unpredictable conditions agents actually operate in. Patronus
-
Loops rescue generative AI as neurosymbolic AI dominates
By
–
Loops, one of the most symbolic tools ever invented, are rescuing generative AI.
— Gary Marcus (@GaryMarcus) 25 juin 2026
Neurosymbolic AI is completely dominating. https://t.co/86zugWJlYRLoops, one of the most symbolic tools ever invented, are rescuing generative AI. Neurosymbolic AI is completely dominating.
-
MAI-image-2.5 ranks #2 text-to-image, #3 editing; Flash best quality/price
By
–
MAI-image-2.5 is new at #2 for text-to-image and #3 for image editing on @ArtificialAnlys
! We’re second now only to GPT models. Also, MAI-Image-2.5-Flash is world-beating for quality/price. Super proud of the team. Now available through the Foundry API and rolling out across -
Gemma 4 prioritizes local on-device intelligence across hardware classes
By
–
Gemma 4 is best in class at each hardware class, not designed to compete on server side frontier intelligence like GLM, it’s designed to enable local on device intelligence without needing advanced hardware
-
SambaNova’s fastest inference cloud and Ricoh’s custom Japanese AI models
By
–
Customers Scaling AI Faster General Compute @fastinference launched the world's fastest inference cloud for AI agents, powered by SambaNova. Meanwhile, @ricoh is using SambaCloud for custom Japanese AI models and agentic business workflows, moving from tens of tokens per
-
Build Faster Coding Agents with SambaNova Responses API and MiniMax M2.7
By
–
Build Faster Coding Agents SambaNova now supports the Responses API, giving AI engineers a cleaner way to connect modern coding agents to fast, production-ready models. Pair the Responses API with @MiniMax_AI M2.7 for repo edits, tests, patches, and high-volume coding
-
SambaNova demonstrates first disaggregated inference cloud for AI agents
By
–
Disaggregated Inference Is Live! At COMPUTEX, we demonstrated the world's first disaggregated inference cloud for AI agents.
GPUs for prefill. RDUs for decode. CPUs for orchestration. The result: faster agent workflows and better economics than homogeneous infrastructure. -
Revolut’s transaction foundation model accelerated by NVIDIA and Nebius
By
–
Revolut built a transaction foundation model – accelerated by the NVIDIA full-stack platform on Nebius – to improve fraud detection, product recommendations, and other use cases in financial services.
— NVIDIA (@nvidia) 25 juin 2026
The results:
📈 2.3x better credit risk accuracy
⚡ Up to 5x higher training… pic.twitter.com/vAtbVJXfPZRevolut built a transaction foundation model – accelerated by the NVIDIA full-stack platform on Nebius – to improve fraud detection, product recommendations, and other use cases in financial services. The results: 2.3x better credit risk accuracy Up to 5x higher training