Build specialized multimodal models with LFM3
Build specialized multimodal models with LFM3
AI NEWS JOURNAL
Search AI news
Search headlines, summaries and articles to connect the dots.
What Okta learned governing context for AI agents at scale
Okta rebuilt how it governs context for AI agents, using dbt as the single source of truth.
Inside dbt Wizard, the AI agent built to show its work
dbt Wizard's designer and engineering lead show how the agent validates its own work before code ever reaches production.
dbt v2 lands: faster parsing, one shared adapter layer, and a docs experience that skips the giant manifest.json
dbt v2 bundles adapters, speeds up parsing, and replaces manifest.json with smaller, queryable files for people and AI.
b11909
hexagon: MUL_MAT_ID row skipping optimizations and improvements (#31169) * hex-mmid: vectorize skipped row zeroing and added zeroing to fused MM_ID_NX * hex-mmid: remove scalar div **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11909/llama-b11909-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11909/llama-b11909-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11909/llama-b11909-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11909/
b11905
vulkan: rms norm optimization using subgroup reductions (#29882) * RMS Norm Subgroup reductions optimizaton * Fix build error * Improved optimization * Make usage of BLOCK_SIZE consistent * Add fast path to skip 2nd subgroupAdd * Add AMD Vendor to subgroup reduction opt * Add comment clarifying rms_dict and merge_maps behavior * Add unroll * Rebase * Fixed syntax issue * Add 128/256/512 block sizes to non-subgroup path * Remove trailing whitespace **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11905/llama-b11905-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.c
Release: v5.21.0
# Release v5.21.0 ## New Model additions ### MiniCPM-OCR MiniCPM-OCR is a document-understanding model from OpenBMB's MiniCPM-V family. It keeps the MiniCPM-V 4.6 architecture, which pairs a SigLIP vision encoder and a window-attention merger with a Qwen3.5 language model. It is specialized for reading text-dense images such as scanned pages, screenshots, receipts and tables. **Links:** [Documentation](https://huggingface.co/docs/transformers/main/en/model_doc/minicpm_ocr) | [Paper](https://huggingface.co/papers/2509.18154) * Feat test (#49920) by @guarin in [#49920](https://github.com/huggingface/transformers/pull/49920) ### Flint Flint is Recursive's mixture-of-experts language model. Compared to a standard SwiGLU MoE transformer such as Qwen3-MoE, it uses hybrid attention (sliding-windo
Marvell’s early chipmaking gamble pays off amid AI demand
Eight years ago, chipmaker Marvell Technology Inc. decided to play leapfrog with nanometers, and its trajectory has been steadily upward since. In semiconductor manufacturing, nanometer designations identify generations of chipmaking technology, known as process nodes, rather than the exact physical dimensions of transistors. Advanced process nodes generally enable greater transistor density and improvements in performance […] The post Marvell’s early chipmaking gamble pays off amid AI demand appeared first on SiliconANGLE.
How DocuSign cut analytics engineering time in half by pairing Claude Code with dbt v2, dbt MCP, and agent skills
DocuSign wired Claude Code, dbt Agent Skills, dbt MCP, and dbt v2 into one workflow and measured a 50% productivity gain.
b11900
common: pin model resolution on real HF repo layouts (#30548) * tests: pin model resolution on real repo listings Extend test-model-resolution with 63 repo listings captured verbatim from the HF API, chosen to cover every file naming pattern found across 1237 GGUF repos of the major publishers and pruned to whole quant groups. Each repo pins its resolved plans and its cache listing. The expectations follow what each file really is per its GGUF header, not what the current resolver picks, so a mismatch is a resolution bug. Mismatches are collected and reported together before the test fails. * tests: leave the listing unchecked on paths the filesystem refuses Some real split names (DevQuasar's Model.Q2_K.gguf-00001-of-00012.gguf under a quant folder) exceed MAX_PATH once materialized in the
Import AI 478: Math for alignment; ProgramBench; and Mirendil shows what RSI means
What is “night” to a machine?
LazyFlash: Forward-Dense, Backward-Sparse Attention for Accelerated Long-Context Fine-Tuning
arXiv:2610.31911v1 Announce Type: new Abstract: As large language models (LLMs) process increasingly long sequences, attention becomes a major computational bottleneck. Sparse attention mostly targets inference, yet many methods still train densely or cannot support sparse training, and their added complexity makes sparsity hard to turn into wall-clock speedups, leaving dense attention widely used in practice. This paper introduces LazyFlash, a forward-dense, backward-sparse attention method built on the observation that attention training is dominated by the backward pass. LazyFlash keeps the forward pass exactly dense, so the trained model remains a standard dense model that existing systems such as vLLM can serve without any change. For the backward pass, LazyFlash proposes a gradient-a
How Shipsidekick replaced Neon with ClickHouse Managed Postgres for 10x faster queries and better reliability
Shipsidekick moved from Neon to ClickHouse Managed Postgres in a single weekend, cutting CPU usage from 100% to 5-10% and making queries about 10x faster.
Introducing the (blazing fast!) native TCP client for .NET
ClickHouse’s new native TCP client for .NET brings faster reads and inserts with fewer allocations, columnar APIs, POCO support, and a path to migrate from HTTP.
Scaling ClickStack alerting with built-in anomaly detection
Managed ClickStack now flags unusual values on the charts you already use with built-in z-score anomaly detection, and a faster, sharded scheduler lets teams bring tens of thousands of alerts over from tools like Grafana.
Why Redshift Serverless can’t match ClickHouse Cloud for real-time analytics
ClickHouse Cloud delivered 1,996× better performance per dollar than Redshift Serverless in CostBench. We trace the gap from fresh data arriving to fast answers coming back.
Why BigQuery can’t match ClickHouse Cloud for real-time analytics
ClickHouse Cloud delivered 438× better performance per dollar than BigQuery in CostBench. We trace the gap from fresh data arriving to fast answers coming back.
What is direct I/O, and why does ClickHouse Managed Postgres use it for backups?
ClickHouse Managed Postgres uses direct I/O and stripe-sized reads to keep backups fast while protecting the page cache and query latency.
Memory safety for Postgres extensions in C/C++
How ClickHouse’s Postgres extensions handle the memory safety challenges of combining C and C++, from clean language boundaries to isolated helper processes.
ClickHouse release 26.9
ClickHouse 26.9 introduces conditional LIMIT boundaries, incremental refreshes for append-only materialized views, disk spilling for DISTINCT, time-limited tokens, and faster min/max/count queries.