MLOps news and analysis on Web Pulse: 3354 curated articles tracking the latest MLOps developments, tools, and research, updated continuously from vetted sources.
A team of developers built a 9-node LangGraph agent that investigates fraud cases end-to-end, using TigerGraph Community Edition as its knowledge graph substrate, TigerGraph MCP for tool access, and NVIDIA NIM for LLM re…
A developer proposes treating LLM-based code reviewers as production systems that need their own test suites, using labeled PR fixtures and pytest to assert on reviewer behavior rather than exact wording. The approach pa…
A developer got Qwen3.8-Flash-Next running on a 128 GB Mac Studio M4 Max by memory-mapping the model's 32 GB n-gram embedding table from disk on demand, avoiding both the quality loss of the expert-pruned REAP-288 build …
A model upgrade to GPT-5.6 (Sol, Terra, Luna) dropped one team's prompt cache hit rate from roughly 91% to near zero because the new caching model only stops at the end of an eligible message, not at fixed 128-token inte…
Modal detailed how it scaled inference for Moonshot AI's Kimi K2.6 coding-agent model, achieving 2.8x per-user and 5.6x across-user performance gains per replica, with one service processing hundreds of billions of token…
A developer documented a failure mode in a local-AI news pipeline running on a Mac where scheduled jobs reported no consecutive failures while the last successful news generation was overdue, because every job — includin…
Cloudgeni announced it is now a fully autonomous cloud platform, with an agent that scans AWS, Azure, GCP and OCI, selects the next improvement to make, writes the fix, validates it with `terraform plan` and opens a pull…
A developer published a shell-script benchmark harness that compares Qdrant, Milvus and Elasticsearch on a FineWeb-10B slice, scoring all three engines against a single shared exact brute-force ground truth and sweeping …
A developer built SentinelGraph, an AI agent that runs end-to-end fraud alert investigations on TigerGraph, combining graph queries, a deterministic evidence layer, a policy engine, and an LLM that only probes open quest…
VLLM is an open-source large language model serving engine built around PagedAttention, which manages the KV cache the way an operating system manages virtual memory, and continuous batching, which keeps GPU utilization …
Cast AI's 2026 Kubernetes Optimization Report, analyzing tens of thousands of production clusters from January 2025 through April 2026, recorded average GPU compute utilization of 5% before any optimization was applied, …
A developer outlined a production architecture for enterprise AI chatbots that treats the system as a distributed application rather than a simple LLM wrapper, with the LLM as only one component. The writeup argues that …
A comparative review evaluates seven auto routing platforms for LLM model selection in 2026, including Bifrost, an open-source AI gateway written in Go that provides multi-provider routing, load balancing, and failover b…
A developer's two-stage evaluation of MCP tool selection found that a small model (gpt-5.4-mini) picks the correct tool from a forced five-tool shortlist 90–97% of the time regardless of description verbosity or paraphra…
Maxim AI has released Bifrost, an open-source AI gateway written in Go that decouples application requests from upstream model providers while enforcing centralized security policies. The project targets enterprise gover…
A developer who failed a system-design interview for a sales-agent memory layer built CogLayer, a working implementation of the Mem0-style hybrid retrieval plus LLM-judgment pipeline described in arXiv paper 2504.19413. …
A developer who has been building in the AI governance space outlines five structural blind spots that current tooling fails to instrument: shadow AI running outside sanctioned policy, prompt-level secret leaks that occu…
TypeSafe has released Jev, a small, fast model designed to answer structured decision questions about a given state rather than generate text, priced at $0.042 per million input tokens with no output-token cost and 70–50…
Internet measurement has identified 39,155 assets presenting the LiteLLM AI gateway fingerprint, with deployments concentrated in the United States (9,792), China (2,629), Japan (2,030), India (1,855) and Australia (1,83…
A developer building Infrawise, an MCP server that gives coding agents context about live infrastructure, disclosed that a name-matching linker shipped with a green test suite for 82 days while carrying a structural blin…