MLOps news and analysis on Web Pulse: 3336 curated articles tracking the latest MLOps developments, tools, and research, updated continuously from vetted sources.
A Vienna-based founder running Postservice.at through Claude Code built a "How I fool myself" section in the project's CLAUDE.md that logs every agent error with its date and root cause, and each new session reads it bef…
A developer built six retrieval pipelines over 2,951 Wikipedia articles in TigerGraph Savanna and benchmarked them on 100 questions using gemini-3.1-flash-lite, finding that an agentic architecture alone added only 3 poi…
A developer built an open-source tool called receipts that reverts each fix in a change and re-runs its tests to determine whether the tests actually fail on the old code, then applied it to 81 maintainer fix commits and…
A developer known as @danisherror has published a self-directed M.Tech program in Networked Systems & Security for AI Infrastructure, a two-year, four-semester self-study curriculum of roughly 15 hours per week (~1,500 h…
Neon adopted Databricks' postmortem standard and rolled out reliability improvements across its Lakebase Postgres control plane after two incidents in one week earlier this year, both triggered by an external dependency …
Google published two reference architectures for networking AI inference model serving, one built on Google Kubernetes Engine (GKE) and one for all other backend types, both anchored by a Private Service Connect inferenc…
A developer detailed the ChimerAI stack's approach to treating prompts as database-backed data rather than code, using a PromptTemplate model with Mustache-style variables, per-category defaults, and permission-gated edi…
A developer at LogicLoop described a production hybrid-search RAG pipeline that runs BM25 keyword retrieval in Elasticsearch alongside dense vector retrieval in pgvector and fuses normalized scores, reporting a 22% incre…
An AI agent working for EU hardware retailer Mineshop.eu published a 15-line JavaScript function that estimates the VRAM needed to run a local LLM by summing quantized weight size, KV cache and a runtime reserve. The mod…
A practitioner at the consultancy iaastha reports that roughly nine in ten promising machine learning projects never reach durable production use, arguing the failure lies in the pipeline rather than the model. The firm …
A developer built a headless agent that automatically fixes broken frontend smoke tests by updating selectors when UI components change. The agent is triggered via a GitHub repository_dispatch event from the frontend CI …
AI-native systems need observability metrics beyond latency, error rate, throughput and saturation because an assistant can return a response in under a second at 99.9% availability and still deliver a fabricated answer,…
Pydantic released a hosted Logfire MCP server that exposes OpenTelemetry traces, logs, and metrics to any Model Context Protocol client, running at https://logfire-us.pydantic.dev/mcp for US projects and https://logfire-…
A developer running a four-node home LLM inference cluster replaced Caddy's lightweight liveness probe with an active health check that forced each node to generate a character, then reverted it the same day after a 10-s…
A developer building a multiplayer particle-in-cell game engine has created a workflow in which an AI agent performs exact 1:1 ports of monomorphic TypeScript simulation functions into Zig, allowing individual functions …
A developer detailed a production pattern for running a quantized Stable Diffusion model on-device in iOS via CoreML, using INT4 weight palettization with FP16 attention projections to cut U-Net size to roughly 750 MB an…
A developer built edit-guard, a stdlib-only Python script that hooks into Claude Code's PreToolUse event to prevent concurrent agent sessions from silently overwriting each other's work. The tool records a file hash and …
A guide from Wattle recommends five trace-aware metrics for production AI agents: task success rate, pass^k repeated-run reliability (with k of at least 3 to 5 runs), trajectory quality, tool-use correctness, and cost pe…
A developer outlines an event-driven reference architecture for physical AI and AIoT systems that chains identification, sensing, transport, understanding, decision, action and verification stages from sensors to physica…
A developer outlined a five-box RAG architecture diagram — index, retriever, reranker, context builder, and loop controller — arguing that each connecting arrow carries a payload paid for twice: once when moved and once …