MLOps news and analysis on Web Pulse: 3346 curated articles tracking the latest MLOps developments, tools, and research, updated continuously from vetted sources.
Amazon Web Services launched Amazon CloudWatch Omni, an observability service for generative AI and agentic workloads, according to an AWS Insights post. AWS argues that agent failures often produce no measurable error s…
A developer argues that evaluation is the oldest discipline in software applied to a component that broke its central assumption, citing OpenAI's April 2025 GPT-4o sycophancy update and its four-day rollback, whose postm…
An AI agent's context ballooned to 261,834 tokens and failed against a 128,000-token model limit, according to incident report INC-002. The growth came not from a single prompt but from replayed tool results accumulating…
A developer building the Wolf.IA platform benchmarked six local models via Ollama and OpenCode on an RTX 4060 Ti with 8GB of VRAM, finding response times for generating a simple Python file-editing script ranged from 31 …
Next.js 16 has marked the `instrumentation.ts` file stable, giving developers a single-execution lifecycle hook that runs before middleware, route handlers, or React components load, so observability SDKs like OpenTeleme…
A developer documented INC-001, an AI agent incident in which a simple planning request triggered 105 calls to the same goal tool because goal tools were exposed globally instead of being scoped to the node that owned go…
Privatemode researchers Johannes Hötter and Marko Rosenmüller demonstrated that the off-the-shelf LLM GLM-5.3-Flash can make typed decisions with a probability for every option in a single forward pass, matching TypeSafe…
A developer has built a command-center system that dispatches work to a fleet of autonomous coding agents across a portfolio of websites, keeping every dispatch operator-gated so agents act autonomously within a task but…
A developer detailed the design of an MCP gateway layer that sits between an agent host and a raw MCP server, adding per-key and per-team rate limits, budget caps, and one audit row per tool call. The writeup walks throu…
A developer published a 20-line Python Stop hook for Claude Code that blocks the agent from ending its turn until a configured verification command (such as `npm test && npx tsc --noEmit`) passes, feeding the failure out…
Okta added four new services to its Okta for AI Agents tool, giving organizations centralized governance over AI agents with end-to-end lifecycle management, runtime policy enforcement, and an instant kill switch to revo…
Masterpoint's Matt Gowie published a five-part CI checklist for validating AI-generated Terraform and OpenTofu infrastructure code, covering formatting and validation, linting, tests, security scanning, and documentation…
A developer argues that manual code review cannot keep pace with AI-assisted development and proposes using an AI code-review skill, such as Superpowers, paired with a pre-indexed local code knowledge graph called CodeGr…
Paperclip, an MIT-licensed open-source management layer for AI agent teams, has reached 86,000 GitHub stars by treating agents like employees with roles, reporting lines, and monthly token budgets, according to its GitHu…
A developer behind TaskFlow, a project management app, traced a 26% failure rate in its AI Quick Add feature to a 300-token max_tokens cap that truncated the output of a newly swapped-in reasoning model (openai/gpt-oss-1…
Sference implemented prefill concurrency in SGLang to improve time to first token (TTFT) in multi-tenant deployments, where stock SGLang allows only one request to be mid-chunk at a time and that request consumes the ent…
A developer has released SpendGraph, an SDK and stage-based framework that prices every retry attempt of an AI agent call rather than only the successful one, arguing that success-only cost tracking understates real spen…
A developer-built task runner called run ships with a built-in MCP server that cuts Claude Code's test-discovery overhead from roughly 6 tool calls and 2,500 tokens to 1 tool call and about 100 tokens, according to the t…
An autonomous AI agent operating an MCP server called contract-powers fingerprinted callers by matching nginx response sizes to its own tool-invocation log, confirming two genuine external tool calls from distinct automa…
Autonomous published a three-layer defense for unattended AI coding agents that combines Secure MCP Cloud, Runtime Protection, and session-state egress restrictions to break the "lethal trifecta" of private data access, …