Obsidian Starter Kit - Foundation
🏠 User Guide / Foundation
Sebastien Dubois
I help you organize your thoughts, life & work⚡
Gemma 4 Gets Multi-Token Prediction Drafters: 3x Faster Inference Without Quality Loss
On 2026-05-05, Google released a companion line of small autoregressive drafter models for the Gemma 4 family, plus a Multi-Token Prediction (MTP) head. Available on Hugging Face and Kaggle. Supported in Transformers, MLX, vLLM, SGLang, and Ollama from day one. Apache 2.0.
Build in public: keeping the OSK v4 AI Assistant robust at 250+ skills
Build-in-public update on the AI Assistant Architecture inside OSK v4.
AI SDKs
An AI SDK is a programmatic library you embed in your application to talk to AI models, orchestrate tool use, and ; in the agent-runtime variants ; run a full plan-act-observe loop. Where a CLI like Claude Code or gh copilot is a user-facing product, an SDK is the same machinery exposed for another
Git 2.54: Config-Based Hooks Replace Husky and Pre-Commit
Git 2.54 ships config-based hooks — the single most practical improvement in years for teams that care about code quality gates. If you've been managing Husky config files, `pre-commit` YAML, or manually copying hook scripts across repos, this replaces all of it.
Claude Code via GitHub Copilot
How to route Claude Code through your GitHub Copilot subscription using a local LiteLLM proxy. This lets you use Claude Code without a direct Anthropic subscription, authenticated via GitHub OAuth.
Markdown-based Installation (MD Scripts)
Markdown-based installation replaces shell scripts (`install.sh`) with prose installation procedures (`install.md`) that an LLM) reads and executes against the local environment. Instead of writing brittle Bash that branches on OS, package manager, shell version, and existing state, you write the in
Kimi CLI
Kimi CLI (official name: Kimi Code CLI) is Moonshot AI's open-source terminal-native AI coding agent, the CLI surface of the Kimi Code product. It runs an agentic loop in the terminal, operating on code, files, and shell commands via natural language, backed by the Kimi K2 model line.
Browser-Provided Language Models
An emerging pattern where web browsers ship with or download trained language models, exposing them to web applications through standardized APIs. Enables developers to access LLM capabilities without managing model weights, infrastructure, or API keys.
On-Device Machine Learning
Running machine learning models locally on the user's device rather than sending data to a remote server for inference. Also called edge ML or client-side ML. Core value proposition of the WebMachineLearning initiative and the Prompt API.
Warden
Warden is an AI-powered code review tool from Sentry. It "watches over your code by running skills against your changes" — analyzing diffs, files, or whole branches for security issues, API problems, performance regressions, and any custom patterns a team cares about. Warden runs both locally (pre-p
Vercel Sandboxes
Vercel Sandboxes are isolated, ephemeral microVMs that Vercel provides for running untrusted or AI-agent-driven workloads. They give a coding agent a real machine (file system, shell, network) without exposing the host or production infrastructure to the agent.