projects
- skills agent skills that run software work as a pipeline where no agent reviews its own work. done is a command that exits 0.
- openflow free, 100% local ai voice dictation. hold a key, talk, clean text lands in any app. open alternative to wispr flow.
- pi my harness. planner, coder, debugger, reviewer, each on a different model.
- t3-pstack pstack for t3 code, with review panels across models.
- 72 hours in my own scores for new models, three days after launch.
live
how i code
1 · design, until we agree
2 · build, one subagent per issue
3 · ship, one pr per issue
- bugbot + devin review
- babysit
- fix each finding
- merge
orange is my skills. the rest is pstack by lauren tan, run in t3 code through t3-pstack. press a skill to read it. verify-app is each repo's own check, made with create-verification-skill.
no model grades its own work. done = a command that exits 0.
stack · $130 a month
- harnesses
- t3 code · grok bot
- claude pro · $20
- opus 5.5 medium · haiku 5.5 xhigh
- devin pro · $20
- swe-2 max · fusion of opus 5.5 medium and swe-2 high · devin review
- x subscription premium · $30
- grok bot · cursor · x premium+
- opencode go · $10
- deepseek v4.1 flash · muse spark 1.3 contributor · 30 to 100 build subagents
- inco · $20 api
- deepseek v4.1 flash · glm 5.3 · glm 5.3 flash · 500+ tok/s for fast work
- openrouter · $30
- any model, to test it
- proxy
- cliproxyapi
- languages
- python, typescript, rust, go, swift
- rig
- alienware m15 r6 · rtx 3070 · i9-11900h
games
-
god of war series
- god of war (2005)
- god of war ii
- god of war iii
- god of war (2018)
- god of war ragnarök
watch dogs 2
assassin's creed shadows
black myth: wukong
clair obscur: expedition 33
gta v
hollow knight
nine sols
subnautica
forza horizon 6
terminal
help · whoami · ls projects · cat stack · games · models · clear