Knowii Voice AI
You Think Faster Than You Type Your Keyboard Is Costing You Hours Every Week
News 📰
Short notes, quick updates, and things I find worth sharing. For the full graph of my thinking, browse my public notes. Subscribe via RSS.
Obsidian Aliases
Aliases give a note alternative names. They are defined in the note's frontmatter (see Obsidian Properties) and picked up everywhere Obsidian resolves names: wikilink autocomplete, the Obsidian Quick Switcher, and unlinked mentions.
Obsidian Embeds
Embeds display the content of another file inside the current note. The syntax is a wikilink prefixed with `!`. This is Obsidian's implementation of Transclusion: content lives in one place and gets reused by reference, never copied.
Obsidian Search
Obsidian search (backed by the Search core plugin for Obsidian) is full-text search across the whole vault, with operators that most users never discover. Learning a handful of them is the difference between scrolling through results and getting exactly the notes you meant.
My AI assistant is turning into an operating system
The way I use AI has gone through many iterations already. Writing them down in order makes the direction obvious, so here they are.
Model routing
Model routing is the practice of directing different tasks to different AI models based on the task's requirements. Instead of using one model for everything, a routing layer selects the most appropriate model considering capability, cost, speed, and context needs.
Microsoft eXecution Container (MXC)
Microsoft eXecution Container (MXC) is an open-source (MIT License) sandboxed code execution system from Microsoft for running untrusted code (model outputs, plugins, tools) safely on Windows, Linux, and macOS. Instead of shipping one isolation technology, it puts multiple containment backends, from
Zellij
Zellij is a modern terminal multiplexer written in Rust. It positions itself as a "terminal workspace" that doesn't sacrifice simplicity for power; discoverable UI with on-screen keybinding hints makes it approachable without limiting what power users can do.
GitHub Enterprise Cloud with Data Residency (GHE.com)
GHE.com is the version of GitHub Enterprise Cloud with data residency: a managed cloud GitHub where a company chooses the region its code and data are stored in. Each enterprise gets a dedicated subdomain (e.g. `octocorp.ghe.com`) with its own namespace, fully isolated from the open-source community
Awesome GitHub Copilot
Awesome GitHub Copilot (awesome-copilot.github.com) is GitHub's official, community-driven catalog of resources that extend GitHub Copilot. Maintained by GitHub with community contributions, backed by the `github/awesome-copilot` repository. It's the de facto marketplace for Copilot customizations a
GitHub Copilot Canvases
Canvases are durable, shared visual surfaces in the GitHub Copilot App where you and an AI agent operate on the same plan. Instead of workflow state living buried in chat history, a canvas makes it explicit and persistent: phases, decision points, validation gates, and drafts update live as the agen
GitHub Copilot MCP Registry
GitHub Copilot MCP Registry is the ability, in public preview, to point a GitHub organization or enterprise at a private MCP) registry so that GitHub Copilot clients only discover, and optionally only allow, the MCP servers your company approves. Admins configure it under AI Controls → MCP in the or
GitHub Copilot Sandboxing
GitHub Copilot Sandboxing is GitHub's answer to the "agent with full shell access" problem: two complementary isolation modes for running GitHub Copilot CLI agent sessions, both in public preview. Local sandboxing wraps the commands an agent runs on your machine in an OS-level sandbox powered by Mic
Terminal Multiplexers
Programs that let you run multiple terminal sessions inside a single terminal window or remote connection. Key capabilities: split panes, multiple windows/tabs, session persistence (detach/reattach), and remote session survival (SSH disconnects don't kill your processes).
Obsidian Callouts
Callouts are styled block elements in Obsidian that highlight important information using a special blockquote syntax. They replace the older Admonition plugin for Obsidian approach with a native, built-in feature.
Obsidian URI
Obsidian URI is the custom `obsidian://` protocol that Obsidian registers with the operating system. It lets any external application (browser, launcher, task manager, script) open vaults, jump to notes, run searches, and create content. No plugin needed; this is the official way to control Obsidian
Agent Plugins - one package format for skills and MCP servers
A new open specification landed: Agent Plugins 1.0, with an initial technical steering committee made of Amazon, Cursor, Microsoft, OpenAI and Vercel. It defines a portable package format for extending agents, so an extension is authored once instead of being repackaged per client.
Remotely Save plugin for Obsidian
Remotely Save, synchronizes an Obsidian vault with cloud storage you already have: Dropbox, OneDrive, Google Drive, any WebDAV server, or S3-compatible object storage (AWS S3, Cloudflare R2, MinIO, Backblaze B2, and similar). It is a lightweight alternative to Obsidian Sync for people who prefer to
OpenRouter adds unified audio transcription
OpenRouter extended its "one API for every model" promise to speech-to-text. A new endpoint (`POST /api/v1/audio/transcriptions`) accepts base64-encoded audio and returns the transcript as JSON, using the same API key as chat completions. No separate STT provider, no extra server.
OpenAI shrank the Codex context window on purpose
While every model announcement brags about bigger context windows, OpenAI went the other way: the Codex model's context size dropped from 372k to 272k tokens. No announcement; the change surfaced as a pull request in the open-source Codex CLI repo, and promptly hit the Hacker News front page because
Notebook Navigator is now Obsidian's most popular plugin
The Notebook Navigator plugin for Obsidian just hit a milestone: it's now the #1 most popular plugin in the Obsidian community directory, and should pass a million downloads within weeks. Not bad for a file explorer replacement.
Faster coding shifts the bottleneck to the outer loop
Luca Rossi makes a point about AI coding that matches my thinking: most AI coding workflows optimize the INNER loop, the rapid cycle of describing a feature and generating code. But that's a fraction of software delivery. The OUTER loop (deciding what to build, shaping the solution, releasing, learn
Use code for control flow, agents for reasoning
Give deterministic code everything it can handle (routing, scheduling, data transformation, file writes), and reserve the Large Language Models (LLMs)) for the parts that need judgment (interpreting ambiguous, unstructured input).
Orchestrator tax
According to Rahul Garg (Thoughtworks, on martinfowler.com), the real value of a subagent is what it keeps OUT of the orchestrator's context, not how fast it runs. Context isolation is more valuable than parallelism.
Google Agent Development Kit (ADK)
Google's framework for building AI agents, available in Python (since March 2026) and Go. Version 2.0 is a response to a production reality: agents that orchestrate everything through the LLM prototype well and then fall apart in production, with infinite loops, hallucinated routing, and runaway cos
Nano Banana 2 Lite
Google DeepMind's distilled version of Nano Banana 2, published as Gemini Image Flash Lite. The trade is explicit: give up some quality on highly nuanced prompts, get a massive latency drop; roughly 2-5 seconds per image where base NB2 takes ~30. Pricing lands at about half the full model ($30/M out
APM (AI)
APM (Agent Package Manager) is Microsoft's open-source dependency manager for AI agent configurations. You declare your agentic dependencies (skills, prompts, plugins, MCP servers, agent primitives) in a single `apm.yml` manifest, and `apm install` reproduces the exact same agent setup on any machin
Claude Desktop on Linux
Since late June 2026, the Claude desktop app runs on Linux in beta. All three tabs are there: Chat, Claude Cowork, and Claude Code with parallel sessions, visual diff review, integrated terminal and editor, and live app preview. Same experience as macOS and Windows, on all paid plans (claude.ai subs
Context size reduction
The context-window arms race runs in one direction in the marketing (bigger!), but vendors sometimes quietly go the other way. OpenAI reduced the Codex model's context size from 372k to 272k tokens, and because the Codex CLI is open source, the change surfaced as a GitHub pull request rather than an
Treat AI context as a budget to manage
The Context Window is a BUDGET, not a capacity to fill. Every token in context costs money (you pay for it again on every turn), latency, and attention (quality decays as the window fills; context rot sets in long before the hard limit). Filling the window because it's there is spending a budget bec
Ori Eval
Ori Eval is OpenRouter's tool for writing your first eval. Their pitch, and it's the right one:
HuggingNews
HuggingNews is a news wire for AI and open source, where the pipeline itself writes the stories. Its own summary: "Headlines and news articles, written from primary sources." The site framing is simpler still: what is happening in AI today?
OpenWorker
OpenWorker is an open-source "AI coworker" announced by Andrew Ng, built with Rohit Prasad. It's an agent that doesn't just chat with you, but delivers FINISHED work. Polished documents and reports, Slack messages, calendar updates, customer briefs, alert triage; tasks completed across your files an
Bento
Bento is a presentation tool that is a single HTML file. Not an app that exports HTML; the file is the software. Open it in a browser and you have the viewer, the presenter and the editor. Send it to someone and they need nothing at all.
qm (Y Combinator)
qm is an open source agent harness built by Y Combinator Software and released under the MIT License. Its own one-liner: "A multiplayer agent harness for work. In Slack and on the web."
varlock
varlock is an Open Source tool for managing environment variables, built by DMNO Inc. and released under the MIT License. Its tagline says the whole thing: "AI-safe .env files: Schemas for agents, secrets for humans."
Voice Clone Lab
Voice Clone Lab clones a voice from a few minutes of audio and generates speech locally on your own GPU. Give it 5 to 15 minutes of clean speech and it fine-tunes Qwen3-TTS into a personal voice model you drive from a CLI or a web UI. By tetsuo-ai, Apache-2.0, Python.
Buzz
Buzz is a self-hosted workspace where humans and AI agents are members of the same community, built by Block on top of Nostr. Open source, Apache 2.0, Rust backend, React desktop client, Flutter on mobile.
Microsoft Teams AI Agents
Microsoft's pitch for agents in Teams is that the agent stops being an app you open and becomes a participant in a conversation you're already having. The tooling is the Teams SDK (formerly Teams AI Library), now the recommended path for all new Teams development — agents, bots, tabs, message extens
Agent-Native Product Decomposition
Agent-native product decomposition is the practice of breaking a product or service into three primitive layers (sensors, actuators, logic) and assigning each layer to whichever computing paradigm fits it best: classical Software 1.0 code, Software 2.0 neural networks, or Software 3.0 LLM-based reas
Voice Clone Studio
Voice Clone Studio is a modular Gradio-based web UI for voice cloning, voice design, and multi-speaker conversation. It is powered by Qwen3-TTS, VibeVoice, and LuxTTS, and supports both Whisper and VibeVoice ASR for automatic transcription. Written in Python, licensed under Apache 2.0.
Vale CLI
Vale is a command-line linter for prose, written in Go by Joseph Kato (`@jdkato`), under the MIT License. Its tagline: "Vale: Your style, our editor."
Agent-Native (Builder.io)
Agent-Native is a framework for building applications that are an app and an agent, from Builder.io. MIT License, TypeScript, React.
Capture Process Use (CPU)
A PKM idea shared by Theo Stowell: strip away the apps, the methods, and the acronyms, and every note-taking system reduces to three stages. CAPTURE information, PROCESS it into knowledge, USE it to produce something (i.e., turn knowledge into wisdom). This loop is the processor of your second brain
Claudexor
Claudexor by Anton Razzhigaev is a local-first control plane that sits ABOVE the AI coding agents. One interface orchestrates Claude Code, Codex, Cursor, and OpenCode: coordinate them, race them against each other, review one model's work with another, and audit everything, without committing to a s
Problem statements aren't enough
Shine Garg (former Staff Engineer, now a career coach for senior+ engineers) makes an argument that generalizes well beyond software: a problem statement alone doesn't contain enough context to make a wise decision about it.
Hammock Driven Development
Hammock Driven Development is Rich Hickey's approach to problem solving, presented at Clojure Conj in 2010. The name is a joke that carries a serious argument: the most valuable work you do on a hard problem happens when you are lying in a hammock, not when you are at the keyboard.
Pattern for exposing Obsidian plugin features via the Obsidian CLI
Chris Gurney's Note Toolbar plugin for Obsidian has a clean answer to a question I care about for my own plugins: how do you expose plugin features through the Obsidian CLI ?
Agentic Context Management
Elvis Saravia (of DAIR.AI) surfaced a paper that puts a name on something worth keeping in mind: most agent failures in production come from context accumulation, NOT from reasoning defects. The model is fine. Its working memory is drowning.
Auto-link to subheadings anywhere in the Obsidian vault
Here's a tiny Obsidian tip. Everyone knows `text` searches note titles, and `[#text` searches headings in the CURRENT note. The lesser-known third form: `[[##text` searches headings across the ENTIRE vault and builds the full `[[Note#Heading` link for you.