I constantly experiment with new technology releases — apps, models, frameworks, design tools, media generation, and more. I keep my findings here.
Every entry gets a verdict. using is what I run day to day. tried means I ran it and it didn’t become a daily tool. watching means I’ve read about it and haven’t run it.
My setup is one Obsidian vault wired to Claude Code.
Scrapers for social sites and pages that block normal fetches
Gives an agent a live Chrome with DevTools traces and network logs
Anthropic's terminal coding agent, and the harness I run code and notes from
macOS terminal built for running coding agents side by side
OpenAI's coding agent, which I run as the second model that reviews Claude's work
One tool and auth layer shared across my coding agents
Personal finance tracker I can query from Claude
AI code editor
Calendar read and write for Claude across Google accounts
AI meeting notes without a bot in the call
Garry Tan's skill pack that runs Claude Code as a virtual eng team
Pipes Apple Health data to Claude
AI image and video generation
Design skill that gives the agent a designer's vocabulary and named commands
Open-source framework for composing LLM apps and agents
Searches 20+ platforms over the last 30 days, ranked by real engagement
Issue tracker and project management
Whiteboard for mapping ideas and processes, with an MCP server
GPU cloud for training and inference
Google's source-grounded research notebook
An expert's YouTube corpus in NotebookLM, queried from Claude Code into Obsidian
All-in-one workspace for docs, wikis and databases
Local markdown vault that everything I run sits on top of
Design canvas that Claude can read from and write HTML into
Lets an agent drive a real browser through Playwright
Open-source Postgres backend with auth and storage
Jesse Vincent's TDD and subagent development methodology for Claude Code
Form builder that agents can build and read through MCP
Frontend cloud that hosts this site
Anthropic's open-source reference stack for agent work in finance
Data enrichment for go-to-market work
Editorial HTML and SVG diagrams, brand-matched from a website
Website builder for landing pages
Garry Tan's searchable agent memory, Postgres and pgvector over markdown
Nous Research's self-hosted personal assistant agent
NVIDIA's extension to OpenClaw with GPU acceleration and enterprise features
Always-on personal agent on your own machine, reached over Telegram or iMessage
Open-source terminal coding agent with a read-only plan mode
Org charts, budgets and a task board for a company of AI agents
Agentic design canvas that coding agents can edit over MCP
AI search for quick answers with citations
Pushes agents toward the smallest change that works, used as a diff review
Browser IDE with an agent that builds and deploys apps
Writes and rewrites drafts in your own voice
Turns a book or folder of documents into a lazily loaded agent skill
Anthropic's tool for making designs and prototypes by describing them
Anthropic's tuned agent harness plus production infrastructure, sold as a product
Open-source 3D code city, the runnable answer to Makepad Scope's demo
Animation and interaction review skill built from Emil Kowalski's writing
Multi-agent coding SDK, the productized rebuild of Steve Yegge's Gas Town
Karpathy's setup for an LLM-maintained markdown wiki, viewed in Obsidian
Writing agent loops that decide what to build, and a test for when they're worth it
Self-hosted LLM gateway with failover across ~350 providers and token compression
Self-hostable SEO suite against Semrush and Ahrefs, with an MCP server and skills
Open-source Mac video editor that Claude can drive over MCP
Autonomous pentest agent that proves each finding with a working exploit
Skill collection meant to make AI coding tools produce premium-looking UI
Nothing matches both filters.