Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Qdrant, the vector database company, published a guide on what to check before tuning a Qdrant collection, emphasizing that retrieval goals must be defined first and that certain settings verify correctness rather than p…
Pinterest is hiring a Principal Engineer for its Core Pinner Journeys team in San Francisco or remote US, offering a salary of $243k–450k per year. The role will define the technical vision for Homefeed, Search, and the …
Amplia (ASX:ATX), a pancreatic cancer drug developer, has diversified its clinical program by partnering with Eli Lilly to combine its lead compound narmafotinib (AMP-945) with an Eli Lilly KRAS inhibitor, while also adv…
A developer built an interactive browser-based tool that visualizes the minimax algorithm and alpha-beta pruning on a Tic-Tac-Toe board, allowing users to step through the search and see how pruning skips branches. The t…
A developer argues that scaling post-training data may no longer be effective, advocating for smaller sets of difficult examples with clear ground truths. The developer highlights Parsewave, a company that creates post-t…
Wayve, the Embodied AI company founded in 2017, is hiring a Senior Machine Learning Engineer for Vision Models in Sunnyvale, offering $312k–370k/yr. The role involves building computer vision and scene understanding mode…
Johns Hopkins University researchers developed MOSAIC, a system using two wearable sensors on the chest and wrist to continuously monitor blood pressure in ICU patients without arterial lines, addressing the limitation o…
Bridgewater Associates' AIA Labs and Thinking Machines Lab reported in June 2026 that fine-tuning an open-weight model, Qwen3-235B, on expert-labeled data achieved 84.7% accuracy on six financial information-triage tasks…
Unsloth released Dynamic 3.0, an update to its post-training quantization method for Qwen3.8-27B, claiming more than 10% better top-1 accuracy at the same disk size than competing GGUF files, based on its own benchmarks.…
Generalist AI, a Cambridge, Massachusetts-based startup founded by former Google DeepMind and Boston Dynamics engineers, demonstrated robots that learn new tasks from short instructional videos without task-specific trai…
Google DeepMind launched SL2T, a sign-language-to-text model that adds American Sign Language (ASL) dictation to Pixel 11's Gboard and Live Transcribe, while OpenAI introduced GPT-Live, a voice model replacing Advanced V…
Dots.mocr, a multimodal AI gateway, now offers an OpenAI-compatible API for running open-weight vision models including GLM-OCR, DeepSeek-OCR-2, and Dots.mocr, with pricing starting at $0.01 per 1M input tokens for Flore…
Tempus AI's $1.5 billion acquisition of Personalis, announced July 20, 2026, looks prescient after Moderna and Merck reported positive Phase 3 results for mRNA-4157, a personalized cancer vaccine that relies on Personali…
A developer created a series to demystify reinforcement learning by tracing its evolution chronologically, from early psychology and mechanical machines to modern mathematical breakthroughs. The series covers foundationa…
A developer detailed a CUDA-based approach to finding the closest island to any ocean point using spherical geometry and GPU acceleration. The method leverages the Haversine formula and spatial indexing to handle million…
A technical guide details a 46-layer hybrid model architecture alternating between Linear Attention (LA) and Full Attention (FA) layers, with LA layers using qkv, z, and out projections plus A_log, dt_bias, conv1d, a/b p…
Unsloth released Dynamic v3.0 GGUFs for Qwen3.8-27B, claiming more than 10% better top-1% accuracy at the same size compared to every other provider. The new quants, which work with llama.cpp and Unsloth Desktop, preserv…
DeepReinforce released Ornith-1.5, an open-source AI model that closes the self-improvement loop by proposing its own training tasks, writing its own scaffolds, and generating its own RL rollouts without human curation. …
NVIDIA FLARE, an open-source federated learning framework, enables federated training of vision-language models (VLMs) across distributed sites by coordinating model updates through externalization, tensor streaming, and…
A developer-built hybrid retrieval-augmented generation (RAG) platform, hRAG, ranks #9 on the EnterpriseRAG-Bench leaderboard with an overall score of 44.74, running entirely on a €116-per-month Hetzner cluster. The syst…