today's signal · no scroll
the brief
Incremental but meaningful movement today: CodexBar expands multi‑provider workflows and Next.js canary pushes Turbopack internals, while Hugging Face spotlights a multi‑agent CNC pipeline on AMD MI300X. Practical posts anchor the day—from running LLMs on an M4 to a security warning in Obsidian—amid voices reframing agent-built frontends and insisting agents must lower maintenance costs.
A malicious Obsidian plugin campaign deployed the PhantomPulse remote access trojan—audit installed plugins, review permissions, and avoid unvetted community extensions to reduce exposure.
Obsidian plugin was abused to deploy a remote access trojan — Article URL: https://cyber.netsecops.io/articles/obsidian-plugin-abused-in-campaign-to-deploy-phantom-pulse-rat/ Comments URL: https://news.ycombinator.com/item?id=48088576 Points: 91 # Comments: 47
New menu‑bar AI switcher adds Manus, MiMo, Qwen, Doubao, Venice, quota warnings, stacked account switching, and faster cost history—useful for teams juggling multiple model backends.
🎚️ CodexBar 0.25 is live 🧩 New providers: Manus, MiMo, Qwen, Doubao, Venice + more 🔔 Quota warning notifications 👥 Stacked Codex account switchers 📊 Faster cost history via models.dev Big one. Menu bar still tiny. github.com/steipete/Codex…
v16.3.0‑canary.19 ships a turbo‑tasks backend proof‑of‑concept for task eviction after snapshot, signaling ongoing performance and reliability work in the build pipeline.
v16.3.0-canary.19 — Misc Changes Proof of concept: task eviction after snapshot for turbo-tasks-backend: #91790 Credits Huge thanks to @lukesandberg for helping!
A case study builds a multi‑agent pipeline on AMD MI300X to evaluate part manufacturability, showing real industrial LLM use with GPU‑efficient orchestration and reproducible components.
MachinaCheck: Building a Multi-Agent CNC Manufacturability System on AMD MI300X
A practical walkthrough runs LLMs on Apple’s M4 with 24GB RAM, covering model choices, tooling, and performance tradeoffs for on‑device development workflows.
Running local models on an M4 with 24GB memory — Article URL: https://jola.dev/posts/running-local-models-on-m4 Comments URL: https://news.ycombinator.com/item?id=48089091 Points: 115 # Comments: 50
Argues agent‑authored UIs should start from index.html, browser primitives, and Web Components with strict routing/state conventions—not traditional frameworks—reshaping agent toolchain design.
For 100% agent-written frontends, I keep coming back to this: Maybe we don't start with a frontend framework Maybe we start with an index.html, browser primitives, Web Components for reusable UI and a strict convention for how agents route, render, mutate state and handle data
James Shore contends coding agents only make sense if they measurably reduce maintenance, shifting evaluation from novelty to long‑term ownership and total cost of change.
An AI coding agent, used to write code, needs to reduce your maintenance costs — Article URL: https://www.jamesshore.com/v2/blog/2026/you-need-ai-that-reduces-your-maintenance-costs Comments URL: https://news.ycombinator.com/item?id=48089289 Points: 39 # Comments: 6
A reflective essay pushes back on over‑automation, favoring clarity and personal control—useful ballast when deciding where AI belongs in everyday development.
I'm going back to writing code by hand — Article URL: https://blog.k10s.dev/im-going-back-to-writing-code-by-hand/ Comments URL: https://news.ycombinator.com/item?id=48090029 Points: 71 # Comments: 30
