kevin@site:~$ whoami
I write about building software, system design, and whatever I'm learning. These are notes for myself that I'm putting out into the world.
contributions
●merged○open✕closed
NousResearchhermes-agentThe agent that grows with you
★ 244kpython4 PRs ↻
hermes-agent
paperclipaipaperclipThe open-source app everyone uses to manage agents at work
★ 80.4ktypescript3 PRs ↻
blockbuzzA hive mind communication platform
★ 32.5krust1 PR ↻
harbor-frameworkterminal-bench-1A benchmark for LLMs on complicated tasks in the terminal
★ 2.6kpython2 PRs ↻
10xappcore-ossOpen-source, all-in-one productivity platform. Email, calendar, chat, files, projects, in one app.
★ 434python1 PR ↻
xgi-orgxgiCompleX Group Interactions (XGI) is a Python package for higher-order networks.
★ 254jupyter notebook1 PR ↻
pymovementspymovementsA python package for processing eye movement data
★ 147python2 PRs ↻
pangeo-datapangeo-docker-imagesDocker Images For Pangeo Jupyter Environment
★ 139dockerfile1 PR ↻
pangeo-docker-imagesrecent -- 20 entries
- agent wire + herdr customizations: one orchestrator, many agents
- i bought a calendar. got a side project.
- The skills economy
- loops, graphs, and state machines
- agents guiding agents
- prompts are programs
- proof gates, not feature increments
- generation is cheap. clarity wins.
- The Internet After the Browser
- stack choices in the agentic era
- i ran my own llm benchmarks (52 cycles, 5 models)
- root cause vs symptom
- handoff: the deliberate context reset
- vibecheck: auditing ai code against the plan
- context-behavior-constraint: acceptance criteria that work
- enforcing test-first development with claude code hooks
- building a multi-llm plan critique system for claude code
- the system
- how i got here
- hello world