AI Morning Brief — 1 July 2026
Overnight in AI: US export controls on Claude Fable 5 lifted, Claude Sonnet 5 and Claude Science ship, and Meituan trains frontier models without Nvidia.
Full feed
Feed 27,755 posts
Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
CUP (Common Useful Python): Building Reliable Python Workflows with Baidu’s Utility Toolkit
In this tutorial, we explore CUP, Baidu's Common Useful Python library, as a practical utility toolkit for stronger Python workflows. We install it in a Colab-friendly…
Multi-Block Diffusion Language Models
Block Diffusion Language Models (BD-LMs) improve diffusion-based text generation with KV caching and flexible-length generation. A natural next step is to extend them from Single-Block Diffusion…
Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs
Metacognition is a critical component of intelligence that describes the ability to monitor and regulate one's own cognitive processes. Yet LLMs exhibit systemic deficiencies in key…
TerraDiT-Ω: Unified Spatial Control for Satellite Image Synthesis with Any Geospatial Primitive
Generative models have achieved remarkable progress, yet applying them to satellite imagery remains challenging. Unlike natural imagery, satellite scenes are structured by spatially complex and semantically…
Nemotron 3 Ultra is taking off on Together AI 🚀 In just days, it's climbed to 35B tokens/day on @OpenRouter. That's the community voting with their…
Nemotron 3 Ultra is taking off on Together AI 🚀In just days, it's climbed to 35B tokens/day on @OpenRouter.That's the community voting with their tokens for…
not much happened today
**Anthropic** re-enabled **Claude Fable 5** with updated cybersecurity safeguards routing some requests to **Opus 4.8**. The relaunch influenced tooling adoption by **Cursor**, **Devin**, and **Perplexity**. Builders…
Meta is adding rate limits and soft paywall to smart glasses
Article URL: https://www.theverge.com/gadgets/959899/meta-ai-glasses-paywall-rate-limit Comments URL: https://news.ycombinator.com/item?id=48742717 Points: 15 # Comments: 1
Multi-GPU kernels are the real test for coding models. Today at @aiDotEngineer, @simran_s_arora shared ParallelKernelBench, an open-source benchmark f…
Multi-GPU kernels are the real test for coding models. Today at @aiDotEngineer, @simran_s_arora shared ParallelKernelBench, an open-source benchmark for evaluating whether LLMs can write fast CUDA…
Re Read more on ParallelKernelBench: https://www.together.ai/blog/parallelkernelbench
Re Read more on ParallelKernelBench: https://www.together.ai/blog/parallelkernelbench
Introducing GeneBench-Pro — testing whether models can handle the kind of judgment-heavy analysis that real-world computational biology requires. Pro…
Introducing GeneBench-Pro — testing whether models can handle the kind of judgment-heavy analysis that real-world computational biology requires.Problems would take a human expert around 20-40 hours…
b9853
ui: Remove PWA navigate fallback to prevent caching API endpoint requ…
Codex has gotten very good
Codex has gotten very goodCorey Quinn: Okay I owe my @OpenAI friends an apology for sleeping on Codex. I was not aware how strong your game…
b9852
opencl: initial q1_0 support (#25160) opencl: general q1_0 support opencl: add Adreno GEMM/GEMV for q1_0 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled)…
Matrix Orthogonalization Improves Memory in Recurrent Models
Article URL: https://ayushtambde.com/blog/matrix-orthogonalization-improves-memory-in-recurrent-models/ Comments URL: https://news.ycombinator.com/item?id=48742514 Points: 9 # Comments: 0
The first early human eggs from stem cells
Article URL: https://www.conception.bio/science-and-updates/the-first-early-human-eggs-from-stem-cells Comments URL: https://news.ycombinator.com/item?id=48742483 Points: 14 # Comments: 2
For all that's said about risk aversion of Chinese capital, it's absolutely *frothing* with regard to AI, if we take into account actual revenues. P/E…
For all that's said about risk aversion of Chinese capital, it's absolutely *frothing* with regard to AI, if we take into account actual revenues. P/E of…
PhotoQuilt: Training-Free Arbitrary-Resolution Photomosaics via Bootstrapped Tiled Denoising
Photomosaics are large images whose local regions are seen as independent tiles while their overall arrangement forms a coherent scene. Generating them at high resolution, with…
I wonder why that didn't work out
I wonder why that didn't work outTeortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞): this teortaxes guy can't keep getting away with it!
AIEWF Daily Dispatch: Loops, Software Factories & Forward Deployed Engineers
On Tuesday at the AI Engineer World's Fair, there was a lot of talk about loops, agent engineering, and the emergence of software factories. Also a…
Re 9/ ParallelKernelBench: Benchmarking LLMs on Multi-GPU Kernel Generation Paper: https://www.alphaxiv.org/abs/2606.parallel-kernel-bench
Re 9/ ParallelKernelBench: Benchmarking LLMs on Multi-GPU Kernel GenerationPaper: https://www.alphaxiv.org/abs/2606.parallel-kernel-bench
Re 8/ Opportunistic Expert Activation: Batch-Aware Expert Routing for Faster Decode Without Retraining (OEA) Paper: https://arxiv.org/abs/2511.02237
Re 8/ Opportunistic Expert Activation: Batch-Aware Expert Routing for Faster Decode Without Retraining (OEA)Paper: https://arxiv.org/abs/2511.02237
Re 7/ Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking Paper: https://arxiv.org/abs/2602.21196
Re 7/ Untied Ulysses: Memory-Efficient Context Parallelism via Headwise ChunkingPaper: https://arxiv.org/abs/2602.21196
Re 6/ When RL Meets Adaptive Speculative Training: A Unified Training-Serving System (Aurora) Paper: https://arxiv.org/abs/2602.06932
Re 6/ When RL Meets Adaptive Speculative Training: A Unified Training-Serving System (Aurora)Paper: https://arxiv.org/abs/2602.06932