Robostral Navigate, an 8B model, enables robots to autonomously navigate using only a single RGB camera, achieving a 76.6% success rate on unseen R2R-CE benchmarks, outperforming multi-sensor systems by significant margins.
Meta reuses old RAM in new servers with custom bridge chip
Meta has developed a custom CXL chip named Vistara to efficiently reuse older RAM in new servers, addressing the 40% performance limitation in its server fleet due to memory shortages.
Show HN: Getting GLM 5.2 running on my slow computer
colibrì enables running the GLM-5.2 (744B-parameter MoE) model on consumer hardware with ~25 GB of RAM, utilizing a unique architecture that streams experts from disk, allowing for efficient memory usage and processing.
SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence
SWE-1.7 achieves frontier-level intelligence at a significantly reduced cost, driven by enhancements in reinforcement learning (RL) infrastructure, data quality, and training techniques, challenging previous assumptions about the limits of post-training improvements.
Muse Spark 1.1
Muse Spark 1.1 is a multimodal reasoning model that significantly enhances performance in coding, tool use, and agentic tasks, boasting a context window of 1 million tokens for improved memory and task management.
Meerkat is a new distributed consensus service developed by Cloudflare, utilizing the QuePaxa algorithm, which allows all replicas to write simultaneously, enhancing availability and consistency across its 330+ global data centers.
Benchmarking coding agents on Databricks' multi-million line codebase
Databricks' internal benchmark reveals that a mix of coding agents, including models from OpenAI and GLM 5.2, provides optimal performance for real-world coding tasks across a multi-million line codebase, emphasizing the need for diverse tools to achieve the best results.
Meta reuses old RAM in new servers with custom bridge chip
Meta's innovative approach involves repurposing DDR4 memory from old servers into new machines using a custom CXL ASIC called "Vistara," achieving a 25% reduction in server count for certain workloads.
Why we're moving off Cloudflare Durable Objects
Wire has transitioned from Cloudflare Durable Objects to a self-built container runtime, enhancing retrieval efficiency by embedding the vector index directly within the container, thus eliminating network latency and state drift issues.
How Version Control Will Evolve for the Agent Boom
Git's evolution is essential as AI agents become primary code producers, necessitating the integration of session logs to capture the context behind code changes, enhancing understanding and collaboration among developers and agents alike.
Why the Next Era of AI Is About Infrastructure, Not Just Models
The next era of AI emphasizes infrastructure over models, as organizations face challenges like fragmentation, cost opacity, and governance gaps in deploying AI at scale.
LingBot-Video: sparse-MoE video diffusion transformer (13B total, 1.4B active) post-trained as an action-conditioned world model
LingBot-Video employs a sparse-MoE architecture with 128 experts and 1.4B active parameters out of 13B total, enhancing its action-conditioned world modeling capabilities.
Introducing Muse Spark 1.1
Muse Spark 1.1 is a multimodal reasoning model from Meta Superintelligence Labs, enhancing performance in coding, tool use, and multimodal understanding, thus pushing the boundaries of AI capabilities.
NVIDIA Nemotron Achieves Benchmark-Leading Performance With LangChain Deep Agents Harness
NVIDIA Nemotron 3 Ultra achieves benchmark-leading performance by tuning the LangChain Deep Agents harness, resulting in 10x lower inference costs compared to top closed models while maintaining high accuracy and throughput.