×
HeartwoodA loop, not a layerPrepared by Phil Morehead
How developers get more done. Less guessing. More doing.
×
HeartwoodA loop, not a layerPrepared by Phil Morehead
Home Depot's codebase is a weathered deck. Good grain, grayed over by years of noise. Heartwood does not rebuild it and does not coat it. It uncovers the grain, the structure that was always there underneath.
It was always there. Now you can build on it.
Heartwood reads Home Depot's fifty thousand codebases and hands every developer a cheat-sheet of what their change will really touch, before they touch it, including the connections no map shows. Then it proves, in Home Depot's own numbers, that it lowers the cost of the work, from a quick fix to a whole new feature.
Fifty thousand repositories, decades of history, tangled with dependencies nobody wrote down. The map their developers are handed is missing more than half the roads. Heartwood reads the grain, and the roots underneath it, the part nobody wrote down, and hands it to the developer as a dossier. Blueprint.IQ, the engine underneath, uncovers the grain their code always had, without reworking a board.
Heartwood runs as a loop. Build the dossier, put it to work, check the result, sharpen it, every pass, and each lap leaves a ring: the dossier gets richer, the work gets faster, and the cost per closed issue drops.
A new loop returns a little. A mature one compounds. That is the number falling every pass, and it is why this is a loop worth owning, not a layer you install once.
Give the developer the specific knowledge the change needs and the right work follows. Give them junk and they ship junk.
Blueprint.IQ reconstructs the map nobody wrote down, the roots that have always been tangled, from the repo's own history.
The developer makes the real change, now knowing what it actually touches instead of guessing.
Heartwood does not write the code. Blueprint.IQ hands over the two lists at the moment of the change, the dependencies any tool shows and the ones nobody wrote down, so the developer moves with fewer surprises and less rework, and so does the coding agent they may be driving. A person stays in the loop, nothing acts on its own.
Separate what is sound from what is not. Keep what holds, drop what does not.
The output is checked against real architecture, real dependencies, and the real security posture, so what ships is what held up, not just what compiled. And each pass teaches the next, what held up and what did not becomes the next pass's context.
Tally the pass honestly, and keep the ring that makes the next one start thicker.
Flip a coin per session on the same team so the Claude rollout cancels out and what is left is Sparq's effect, in cost per closed issue, in your telemetry. Grading is co-signed, published either way, break-even about 1.3%. The learning compounds into the next pass.
Nobody is being careless. Underground, the roots braid together where nothing on the surface shows it, so a shovel near one trunk kills a tree three rows over. Their codebase is that tangle: the files that fall together were never drawn on any map. Heartwood reads the roots, and keeps reading them every pass.
The dossier and the loop, reached mid-task without leaving the workflow.
The prototyping platform, safe because Blueprint.IQ checks the work before it ships.
It was always there. Now you can build on it.
How developers get more done. · Less guessing. More doing.