devotd.
Browse the Asia/Seoul (KST) issue calendar. Retrospective entries show their assigned coverage day; actual publication times remain separate.
CodeQL 2.27.2, released October 7, rewrites the Go control flow graph on the shared CFG library, which can change what custom Go queries assume. On October 9 GitHub added that autobuild and manual modes for compiled languages are not supported on macOS 27.
audio.cpp v0.9.1 replaces the PocketTTS CPU Mimi upsampler’s 32 MiB dense diagonal weights with a 64 KiB depthwise module. In the maintainer’s measurements, four-thread decoder time per frame fell from 30.50 to 10.11 ms, but GPU paths are unchanged and some release binaries may require AVX-512.
LightOn released LightOnOCR-3 on October 8. The grounding prompt adds region labels, 0–1000 normalized coordinates, image descriptions and chart tables, and it changes output-token and rendering-resolution budgets as well.
On October 8 MSI opened Hong Kong preorders for the Prestige N16 Flip AI+, an RTX Spark-based 2-in-1. Up to 128GB is shared by the CPU and GPU, and for local AI the question is whether a real software stack on Windows on Arm runs within that space.
TRL v1.15.0 scores tokens in SFT, DPO, KTO, GRPO, RLOO and Distillation without building the full logits tensor. In the project's measurements DPO's maximum trainable length grew 5.80×, but the effect depends on the previous loss path and vocabulary size.
On October 7 Microsoft made Microsoft Execution Containers (MXC) generally available to enforce access policy outside AI agents. Learning mode records while keeping denials, but permissive mode actually allows what it records, so the two modes produce different kinds of evidence.
VS Code 1.141 adds a cleanup view for seeing and removing the disk space used by inactive agent session worktrees. Restoring a session recreates its worktree, but the release notes do not say uncommitted local files come back.
WHIRL v0.1.4 narrows the CORS default of its native Windows LLM server to web pages served from loopback addresses. Inference code and speed are the same as v0.1.3, and a streaming client that stops reading no longer blocks other users.
On October 2 GitHub began expiring unvalidated npm trusted publishing configurations 48 hours after creation. Only a first successful publish exempts a configuration, and tokens from GitHub Actions issue_comment events are now rejected too.
On October 5 Reflection announced Beam, a 501B-parameter MoE model, and began opening its beta API gradually. For moving an existing agent, what matters is the service token limits, the supported endpoints and parameters that are accepted but have no effect.