MLOps news and analysis on Web Pulse: 739 curated articles tracking the latest MLOps developments, tools, and research, updated continuously from vetted sources.
A developer found that `nvidia-smi` reported 97% GPU utilization on an H100 cluster while actual training throughput was less than half of expected benchmarks. Tracing via eBPF revealed the GPU was idle 51.7% of the time…
A developer built TokenWatch, an open-source tool for tracking LLM costs, after Helicone entered maintenance mode following its acquisition by Mintlify in March. The lightweight alternative uses SQLite and requires no pr…
Eighty engineers, researchers, and community builders gathered for the inaugural PyTorch Meetup Singapore, hosted at the Red Hat Asia Pacific office. The event featured technical talks on inference, distributed training,…
A developer has identified three distinct causes of the "double-reply" problem in email agents—webhook redelivery, concurrent workers, and shared inboxes—and has published a set of deduplication and locking patterns to p…
MLJAR AutoML and Mercury now enable developers to build web applications for machine learning models using only Python, eliminating the need for HTML, CSS, or JavaScript. The new tool generates two types of web apps — on…
A SQL mapper like MyBatis removes the safety net of identity maps and dirty checking that makes rich entity graphs workable in a full ORM. Agent-focused platforms like Stack Overflow for Agents use Cloudflare WAF in "Und…
An engineer at Fortem.dev has published a four-step guide to controlling runaway CloudWatch Logs costs on Amazon ECS, where default settings create log groups with "Never Expire" retention and INFO-level verbosity. The f…
A centralized workflow engine that manages the entire software development lifecycle through a single state machine inevitably fails as teams accumulate exception branches and workarounds, turning the engine into an obst…
A developer created a CLI tool called "no-mistakes" that intercepts every `git push` and routes code through an automated validation pipeline before it reaches GitHub. The tool initiates an agentic review loop that tests…
The HelperX team built a circuit breaker system to prevent cascading failures when external API dependencies go down. The system monitors failures for each dependency per slot, automatically opening circuits after three …
A developer spent three hours debugging a production LLM memory system bug where a `rollback` method wiped out entire conversation histories instead of just undoing erroneous operations. The root cause was a code refacto…
A developer has proposed replacing equality-based assertions with invariant checks for testing AI agents, arguing that non-deterministic outputs make traditional tests flaky or fake. The approach uses structural and fact…
Anthropic's 2026 Claude model lineup includes four tiers — Fable 5, Opus, Sonnet, and Haiku — each designed for specific task complexities and cost profiles. Teams that implement intelligent model routing by matching tas…
Upriver, a San Francisco-based AI-native data engineering platform, raised $14 million in seed funding on June 8, 2026, led by Valley Capital Partners and Hetz Ventures. The company's platform autonomously resolves data …
A blog post titled "Failure numbers every programmer should know" provides reliability napkin math for hardware, cloud services, and software defects. The post is part of a collection that also covers topics including AI…
Redb.Route.Llm 3.1.1 adds seven nullable audit fields to every persisted message, capturing effective sampling parameters, a SHA-256 hash of the tool set, and the provider's system fingerprint for compliance and replay p…
A developer resolved two common deployment errors—GHCR “Unauthorized” and Docker “Cannot perform interactive login from non-TTY”—by configuring GitHub Actions to use a personal access token (PAT) with the `write:packages…
ORP, an open experience layer for AI agents built on OpenTelemetry, converts agent failures into regression tests and retrievable lessons. The tool captures agent traces, challenges unproven claims, and compiles executab…
GKE Labs has launched OpenRL, an open-source, self-hosted training API for fine-tuning large language models on Kubernetes. The API decouples post-training infrastructure from reinforcement learning research workflows, u…
Bifrost, an open-source AI gateway written in Go, can handle thousands of concurrent LLM requests on Kubernetes with only 11 microseconds of overhead per request at 5,000 requests per second. The gateway demonstrates 54 …