View as Markdown (for AI/KB) Whole knowledge base (llms.txt)
Probe builds a knowledge tree over the board: it’s broken down into functional blocks — power → rails, analogue, digital, buses, interfaces, protection, subsystems — and each node carries a state, findings, open questions and two percentages. Understanding the two figures is the key to reading a review right.
The node’s state
Each node has a review state: understood, shallow (needs deepening), attention (needs attention) or contested (conflicting, needs review). Each node also points to which schematic sheets back it.
The two figures — deliberately
Probe reports two percentages per node, and the difference is the heart of why the system doesn’t lie to you:
- Headline % = pin-weighted evidence over the active parts. It measures verified pins ÷ total active pins, so a large device (an FPGA, a complex IC) dominates — exactly the risk-bearing minority. It’s the figure that matters.
- Breadth % = structural mean over all elements (secondary). The simple majority (passives, decoupling) inflates it, so it’s never the headline.
Why a node stays shallow
A node stays shallow as long as it holds any unproven active part. You can’t average your way to a green board by solving the easy 90 percent. The measure is built so the headline rises only when evidence is gathered over the parts that can actually do harm.
Design principle: a progress figure should measure evidence over the risk-bearing minority, never a coverage mean that the simple majority inflates.
Next: the driven review and observations, concerns and Green.
Probe is E-Sharp’s design-review engine. The headline can’t move until the evidence is there — an honest “don’t know yet” beats a confident guess.