01
Observe
See the correction
A camera watches the product and the machine where production drifts. People stay out of frame.
vision-first automation
Factories should be able to see the work they are still doing by hand.
sightcrafts is building the layer between what a factory can observe and what it can safely automate. We start with the corrections people already make, then close the smallest loop first.
LOOP0.1
product + machine
correction
closed-loop data
the thesisClose the smallest loop first.Camera → evidence → machine control → physical action
The approach
From seeing to doing, one reliable step at a time.
Most factories already have automation. The gap is what happens when the process leaves its normal path. That is where a person notices, decides, and intervenes.
02
Measure
Make the loss legible
Every intervention becomes a record: what happened, what changed, how often, and what it cost.
03
Assist
Help before the miss spreads
The system surfaces the right signal early. Operators keep the decision while the pattern proves itself.
04
Control
Use the machine already there
When the loop is trusted, the next step is a setpoint, a hold, or a safe stop through existing controls.
05
Automate
Add physical action last
Robots come after the repetition is understood. The correction log is the starting point, not a side effect.
What we are exploring
The product sees the object, the machine, and the consequence.
01↗
Quality
Catch drift before a batch becomes waste.
02↗
Capacity
Turn visible downtime into recoverable output.
03↗
Labor
Move repetitive correction out of the critical path.
Designed to run close to the line. Product and process first; people stay out of frame.
A quiet beginning
We are still in the field, testing whether the same loop can travel.
Different lines. Different materials. The same question: can a vision system learn a recurring correction well enough to help the machine take it over?