recruiter?

(follow the red line) →

open loops

  • agents that remember
  • code that explains itself
  • trajectories + evals
  • when helpful becomes agreeable

software engineer

  1. fuwafull-stack + ai product engineer
  2. octalfull stack engineer ii
  3. devops.phcontractor → intern

typescript, python, rust, go, infrastructure, agentic systems.

current role / shipping
shipped 114 merged production prs in ~5 weeks (~22/week), improving reliability, usability, and workflow coverage across a complex saas platform for high-stakes meeting and document operations.
résumé · may 2026
previous role / infrastructure
provisioned and automated azure infrastructure using terraform, ensuring scalable and reliable cloud environments.
résumé · may 2026
taylor ai interface
meandry interface for organizing an adhd-shaped workload
selected proof

114 production prs

in roughly five weeks

plus 2,000+ daily requests at 99.3% uptime in an earlier role.

current build, open in this window.

no final form. no portfolio template. this board changes when attention changes.

genuinely alive

this websiteatlasweavemayspaceaiproxyorchestratorweppi

this website + ai stuff. open this week.

i hate spinners.

same circle. over and over.

atlas

structural code intelligence for developers and agents, with answers grounded in the repository graph.

current benchtypescriptsqlitemcp

weave

an experimental memory layer for long-running agents, state tracking, and better recall.

current benchtypescriptagentsmemory

writings for discussion

how WFDs work, when to write one, the states a document can be in, frontmatter format, and the markdown features available.

wfd · 2026-07-22T00:00:00+08:00 · 9 min

weppi

a browser and electron interface for pi with a real local extension bridge.

current benchreactelectrontypescript

Atlas: structural code-intelligence for LLM agents (an empirical evaluation)

2,239-trial benchmark across 8 OSS repos: Atlas beats a text-search baseline by +0.223 deterministic, +0.127 LLM-judge, at 42% fewer tokens.

wfd · 2026-04-19T18:00:00 · 17 min

mayspace v2

markdown-native personal knowledgebase with an agent that reads and mutates notes through explicit tools.

current benchnextmarkdownagents

aiproxy

personal codex account router with usage tracking, refresh, and fallback routing.

current benchtypescriptoauthproxy

smollama

small terminal interface for chatting with ollama cloud models.

current benchterminalollama

orchestrator

local llm wrapper that reads and controls a live chromium page through the devtools protocol.

current benchtypescriptcdpagents

giving an ai agent a hippocampus

i watched a kdrama about a girl who forgets everything every morning and realized my AI agent has the same condition. so i built it a brain.

wfd · 2026-02-28T02:01:00 · 14 min

weppi blueprint
weppi blueprint
twinkle dashboard
twinkle dashboard
clerk is very slow
clerk is very slow
trusted with working remotely
website under construction
thinking about it
shipped. probably.
too many open loops
experiment escaped
open question
busy being employed
research branch escaped