MLOps

A developer built Soma, an open-source governance framework for AI coding agents that makes rule violations structurally unprofitable by attaching expiries to every rule, verifying agent claims against transcript evidenc…

A paper posted September 2, 2026, "EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction" (Shi, Sun, Dong, Wan, Lo, Gu), proposes halting agent benchmark runs whose outcome is already predictable, using two gr…

A developer published a practical guide and open-source GitHub repository (WhiteCollarAgent) showing how to turn white-collar knowledge-work tasks into agent environments, expose tools via MCP servers, sandbox the agent …

A volunteer-run project called Coop has moved to Stage 2, pretraining a ~145M-parameter decoder-only transformer from scratch on FineWeb-Edu using DiLoCo-style low-communication data parallelism, with weights and optimiz…

Halv cut recorded model cost by 57.1% versus vanilla Codex across 14 repository tasks, spending $144.67 against $337.50 while both arms passed 25 of 42 runs, according to Halv's own interim benchmark checkpoint. The Halv…

Graph-based contextual retrieval, which uses GraphRAG to traverse a context graph rather than relying on vector similarity alone, improves AI agent accuracy by following explicit relationships between data entities, acco…

SmartBear's State of Software Quality and Testing 2026 report found that 81% of organizational leaders believe AI can reliably catch its own errors despite having no evidence, while only 64% of software practitioners tru…

A community tester published deployment notes for the Halogen Flash Server, an optimized server for running Qwen3.8-Flash-Next on AMD Strix Halo (gfx1151) hardware, reporting a roughly 10% performance uplift from the amd…

A developer argues that AI model providers should not be the sole authority on their own billing, proposing an independent "second ledger" that preserves the provider's usage report and checks it against delivery, comple…

A developer benchmarked the same Gemma 4 E2B build served from an Amazon SageMaker endpoint and from a plain EC2 instance with an identical NVIDIA T4 or L4 GPU, finding decode speed and answers match within 2% while EC2 …

CoreWeave Inc. (Nasdaq: CRWV) announced CoreWeave Forge, a development layer that unifies Weights & Biases Models, post-training expertise from OpenPipe, and the open-source marimo notebook project with CoreWeave's own s…

A developer built local-agent-sandbox, an evaluation harness using pure QEMU with KVM and QCOW2 copy-on-write overlays plus llama.cpp to test whether Google's quantized 4B Gemma model can autonomously repair a broken Lin…

A developer working on Renovate performance found that ~30MB OpenTelemetry traces exceeded HTTP size limits when ingested into a local Jaeger instance, and resolved it by streaming each line of /tmp/traces.jsonl into cur…

A developer deployed Gemma 4 (E2B, E4B, 12B and 26B A4B) to Amazon SageMaker on a single NVIDIA T4 (ml.g4dn.xlarge), measuring decode throughput at 0.77x to 0.82x of an L4 while producing identical answers, with 12B the …

A developer outlines a server-rendered feature flag admin design that keeps a single authoritative backend, uses compare-and-swap writes on a versioned flag document, and rejects unknown flag keys to prevent stale-tab ov…

A developer built agentic-arena, a benchmark that holds the model, tools, datasets and iteration budget fixed to measure how seven agent frameworks differ on the wire. Six of seven frameworks stayed within a 1.15x band o…

A developer outlines a set of AI-native Service Level Indicators (SLIs) designed to catch semantic failures in large language model applications that traditional infrastructure monitoring misses. The piece argues that co…

Researchers at KAIST and Microsoft published "Follow the Entities: A Corpus Map for Agentic Search" (arXiv:2609.37226), quantifying how much compute LLM search agents waste navigating raw document corpora. On EnterpriseR…

A new paper, REAL-Q, proposes a dynamic gradient descent approach to post-training quantization that the authors say fixes structural flaws in GPTQ, including upstream and downstream misalignments and a frozen Hessian. T…

A developer built Paveo, a policy checkpoint that intercepts and blocks unsafe tool calls before an AI agent executes them, and replayed two months of their own Claude Code session logs to test it. Across 59 days, 106 se…