MLOps news and analysis on Web Pulse: 3309 curated articles tracking the latest MLOps developments, tools, and research, updated continuously from vetted sources.
A 27B NVFP4-quantized model served by vLLM on a Windows 11 desktop with an RTX 5090 (32 GB VRAM) dropped from roughly 77 tokens per second to about 7 tokens per second during Microsoft Teams video calls, with no out-of-m…
A Jev-driven diagnosis pipeline with no LLM agent passed 80 of 105 diagnoses (76.2%) across 21 SREGym-Lite faults, with a median diagnosis time of 14.6 seconds, according to the SREGym team's study using jev-1.13.0. The …
A team evaluating the Cline VS Code agent extension found that a BYO-key setup with Kimi K3 on Terminal-Bench 2.1 raised its pass rate from 77.5% (69/89) to 88.8% (79/89) while cutting run cost from $79 to $49.80, attrib…
A developer published a step-by-step guide for self-hosting OpenMuse, an AI chat assistant that can also manage a CasaOS home server, using SSH and Docker Compose. The guide deploys the casaos-manager-v3 branch across th…
OpenAI and Ironclad published a case study describing how Ironclad's production contract lifecycle workflows are being converted into reproducible evaluation benchmarks for computer-use agents. The approach sanitizes rea…
A developer has published a deep dive on building a cost-aware LLM router that classifies task intent and complexity to route requests to the cheapest viable model, aiming to cut the total cost of ownership of high-volum…
Massdriver's October 2026 platform update changes how SCIM users are assigned seats, making the `active` flag and an optional seat rule — not group membership — determine who holds a seat, with the SCIM integration accep…
TJ Bai and Jason Eisner published "Prompt Choreography" in Transactions of the Association for Computational Linguistics Volume 14 (pages 253–270), a framework that maintains a dynamic global KV cache so each LLM call ca…
AegisGate, an open-source AI security platform, has shipped a federated threat intelligence system that lets organizations share indicators of compromise without exposing raw attack payloads. The system fingerprints dete…
VIDRAFT released POCKET-Darwin-180B-GGUF, a 4-bit quantized GGUF build of its 180B-parameter Mixture-of-Experts Darwin-180B-RSI model that runs on consumer hardware, including a gaming laptop with 8 GB VRAM and 32 GB RAM…
A study of 39 open-weight language models driving the agent mode of an open-source SQL client over eleven days found that 1,590 of 2,100 model-attributed agent-mode losses, or 75.7%, came from runs that had invoked at le…
A developer who builds client voice agents and subscription products argues that AI features should treat model calls as a metered resource from the first sprint, since hosted models charge per request by tokens in and o…
A consultant working with an eight-engineer ML startup cut its $22,000 monthly AWS GPU inference bill by $4,800 by aligning cluster capacity with actual traffic patterns. The startup's inference requests were concentrate…
A developer outlined a practical logging pattern for Node.js AI agent workflows that uses Pino or Winston to emit structured JSON completion events, each tagged with request_id, user_id, trace_id, environment, latency_ms…
A developer fine-tuning the Exolio AI machine-text classifier on rented Vast.ai GPU instances lost roughly £1,000 across dozens of training runs between January and April 2026 to silent failures that produced no error me…
A developer published a GitHub repository documenting 109 real-world security incidents involving autonomous AI agents deployed in 2026, alongside a falsification matrix and a multi-agent privilege attenuation harness th…
A developer detailed how quantizing the KV cache to FP8 or INT8 roughly halves per-token memory in LLM serving, from about 128 KB per token in FP16 to 64 KB, enabling longer contexts or larger batches on the same GPU. Th…
A developer demonstrated that 4-bit GGUF quantization of Mixture-of-Experts models can match full-precision quality when sensitive tensors are kept at higher precision, reporting MMLU-Pro 87.65 for a mixed-precision UD-Q…
ScienceLogic announced Skylar AI 2.7 at Nexus Live 2026, adding security, governance, and agentic controls to its Skylar Advisor platform, including an agent-routing trace and tool catalog, an upload gatekeeper that insp…
Only 50 percent of enterprises successfully move more than half of their AI projects from pilot to production, according to Outsystems' State of AI Development 2026 report cited by CIO. The report attributes the gap to s…