/uses · July 2026
The tools I use.
My setup changes all the time. I keep a tool when it helps me understand the work, make a better decision, or get something done without losing the evidence behind it.
01
Desk
I do not think much about the desk. It just needs to keep the work close and stay out of my way.
02
Builders
I use different models for different jobs. One can build while another is better for challenging the direction.
CodexWhere I do most of my investigation, implementation, and verification. It is my default for turning a decision into working code.Claude CodeI use Claude when taste, framing, careful review, or second-order risk matters.GeminiUseful when there is too much context and I need help making sense of it.GrokUseful for pressure-testing a direction, tightening scope, and finding where a plan is still soft.
03
Memory
Agent work gets better when the next session can pick up from the real state.
agent-session-resumeHelps me pick up from where the work actually stopped across Codex, Claude Code, Antigravity, and OpenCode.KontinuoA short checkpoint of what changed, what was verified, what is still open, and what should happen next.AgentCanvasA local view that makes a repo easier for an agent and a person to inspect together.SkillsThe rules I want agents to reuse when a job repeats: where to look, what proof counts, and when to stop.
04
Guardrails
I am comfortable moving quickly when I can still see the important control points.
Second-model reviewFor important decisions, I ask a different model to challenge the direction before I ship.Focused agentsGive research, implementation, and review clear owners, then bring the evidence back together.Runtime proofOpen the route, hit the endpoint, run the simulator, or inspect the workflow. A successful build is only one check.
I will keep changing this page as tools earn or lose their place.