MLOps news and analysis on Web Pulse: 3397 curated articles tracking the latest MLOps developments, tools, and research, updated continuously from vetted sources.
An engineer at Weio, an AI-run company, shared findings from instrumenting production traffic that revealed a single frontier model, gpt-4o, handled 77 percent of calls but accounted for 97 percent of model spend. The te…
Software Sausage released six open-source AI workflow kits, each containing a README, an editable evidence ledger, and a dependency-free shell verifier. All 17 verifiers pass at release v0.17.0, and the kits are designed…
Docker Captain and AI Engineer Yhary from Colombia built a Docker Sandbox Kit mixin that automatically starts an MLflow tracking server for Claude Code, enabling out-of-the-box experiment logging and model versioning. Th…
A developer benchmarked cold start latencies and costs across serverless GPU platforms, finding Modal fastest at 1.8 seconds on an A100, followed by RunPod at 4.2 seconds and Replicate at 6.5 seconds. The results highlig…
Roboflow published a guide detailing how manufacturers should run computer vision inference at the edge rather than in the cloud, citing a standard cloud round trip of 100-300 ms, roughly 1.49 Gbit/s of raw video per cam…
A GPU lock bug that had been quietly degrading an internal RAG sweep was exposed during a pipeline validation run on 2026-06-19, according to a technical post from the engineering team. The team fixed the issue by treati…
A developer proposes a characterization-testing approach for refactoring untested 'glue' code that writes files, reads environment variables, or changes directories. The method records a frozen ledger of process behavior…
Google separated pre-integration verification from merge coordination more than 15 years ago, with its monorepo handling 40,000 commits per day by early 2015, including 24,000 from automated systems, according to Paul Ha…
A developer reported that their local LLM agent fleet experienced silent failures during a 58-day unattended production run, including process errors and timeouts that went unnoticed for days. The root causes included mo…
A technical guide details the production deployment of on-premises large language model inference using vLLM, covering an 8-GPU node running GLM-5.2/5.3 (NVFP4 MoE) and a single L40S running gemma-4-26B FP8, with specifi…
An engineer at Chase built Agent Review Studio, an open-source, local-first workbench for evaluating and refining AI agents and agent harnesses. The tool lets engineers inspect execution traces, classify evidence, and tu…
Masterpoint engineer Veronika Gnilitska outlines a workflow that uses AI agents and MCP servers to convert AWS click-ops resources into Terraform/OpenTofu code, emphasizing human review gates and verification scripts to …
Maxim's open-source AI gateway Bifrost, developed in Go, provides an OpenAI-compatible interface that equips n8n workflows with automatic failover, semantic caching, and granular governance. Routing n8n AI workflows thro…
A developer argues that teams routing LLM traffic to cheaper models must measure whether the routing actually worked, not just celebrate cost savings. The post warns that cheap-model calls can succeed technically while s…
Vamgan released review-to-rule, an open-source plugin and skill that converts accepted code-review feedback into durable, scoped repository rules for coding agents like Claude Code and Codex. The tool supports GitHub, Gi…
A developer detailed the architecture of production AI systems on AWS, emphasizing that the core challenge is building reliable systems around LLM API calls rather than making the calls themselves. The post maps common p…
A new open-source tool called trimtab lets operators change SGLang and vLLM scheduler settings live, cutting configuration changes from a 1-7 minute redeploy to about 15 ms, with zero dropped requests on H100, RTX PRO 60…
A developer spent two days debugging an LLM endpoint that returned answers for the wrong prompts, initially blaming model drift and sampling noise. The root cause was a cache key that hashed only the user message, ignori…
A developer's analysis of production AI agent failures reveals that silent failures—where systems complete execution without errors but produce incorrect or harmful outputs—are the dominant failure mode, often slipping t…
Gartner expects roughly 40 percent of enterprises to demote or decommission autonomous agents by 2027, driven by governance gaps that only surfaced after production incidents, while DORA's State of DevOps research coveri…