Keep what works
If an existing platform is valuable, make it cleaner, faster, and easier to operate. Use what you already purchased well.
For operators, founders, and technical teams
Move past bloated platforms. Build tools around the work.
I help teams get more from the systems they own and retire the parts that slow work down. Existing platforms can be cleaned up when they still earn their keep. When they become drag, the better move is usually one small purpose-built tool. AI belongs where it makes the job clearer, faster, or easier to trust.
The Point of View
Sunset what doesn’t
When a system is expensive, slow to change, and no longer matched to the team, replace the smallest useful part first.
Build for the actual workflow
Purpose-built tools can match the way people work instead of forcing people to work the way a platform wants.
Getting started
A practical path out of platform drag.
I use a simple process: map the workflow, separate what matters from what adds noise, build one useful slice, and iterate from real use with operators, founders, and technically forward teams.
Talk through the first sliceHow I Use AI
Claude for the plan
I use Claude while the problem is still fuzzy: ideating, pressure-testing assumptions, mapping options, and working through decisions before code exists. A plan only matters if it turns into something real.
Codex for the grind
I use Codex once the direction is clear: implementing, refactoring, writing tests, wiring commands, reviewing diffs, and pushing through details. It works best with a tight scope and a way to prove the change.
OpenClaw and local AI for control
I use local tools for private exploration, sensitive workflows, offline review, and helping people learn what is possible without handing every task to a cloud service. Production paths still need reliability and traceability.
What This Looks Like
Legacy platform audit
Separate useful infrastructure from inherited complexity, unused licenses, duplicate workflows, and avoidable integration cost.
Purpose-built replacement
Design and ship a smaller tool that solves the real job: intake, review, tracking, approvals, reporting, handoffs, or customer operations.
Fast iteration loop
Use Claude for planning, Codex for implementation, OpenClaw and local AI for private exploration, and tight feedback cycles to improve the tool while the team is already using it.