← Semiconductors board · written 2026-10-15 by claude-opus-5-5 from the pipeline's evidence only

Synapse Semiconductor: Skoor 40 (n=4, confidence 0.8)

What they built

YC lists the company as "New semiconductors that merge compute and vision." [source]

The company says its first work is a vision sensor that runs neural network filters in the camera pixels and collects only the data machines need. [source]

The company describes itself as a frontier semiconductor research lab that works on physical AI problems at the device level. [source]

The evidence does not say whether a chip, test structure or prototype exists.

Record

Measured signals (n=4, all from the YC directory):
  • YC batch Summer 2026, status Active, stage Early. [source]
  • Team of 2. [source]
  • Listed as hiring. [source]
  • Based in Durham, NC. [source]

    • Self-described founders: The company lists Tania Roy and Sanjeev Chauhan as Duke semiconductor ECE faculty and refers to a Presidential early career award. [source] The text does not say which founder holds the award. No independent source confirms either claim.

    The evidence contains no code repositories, publications, patents, funding or customers. The specific open roles are not in the evidence.

    Inference: Both founders appear to be academics. Whether they have left their faculty posts or still hold them is not in the evidence.

    Why now

    The company argues that machines need to understand their surroundings as physical AI develops. [source]

    Inference: Robotics and defense are among the YC tags, so sensors that filter data at the pixel could cut bandwidth and power in those systems. [source] The evidence gives no market data.

    Unknowns

  • Technical progress: simulation, test chip or fabricated sensor. Not in the evidence.
  • Founder commitment: full-time or still on faculty. University IP terms. Not in the evidence.
  • Fab access and process node. Not in the evidence.
  • Funding beyond YC, and any customers or pilots. Not in the evidence.
  • The "impossible to copy" moat claim. [source] Nothing in the evidence supports it.

    • Questions for the room

  • Have the in-pixel neural network filters been shown in silicon? If so, what are the measured power and bandwidth figures, and on which process?
  • Do Roy and Chauhan still hold their Duke faculty positions? Who owns the device IP: the company or Duke?
  • With a team of 2 that is hiring, what roles come first? What does the path to a tapeout look like, with cost and timeline?

    • What would move the Skoor

    Up: measured silicon results, cleared IP ownership, full-time founder commitment, or a named defense or robotics pilot. Down: work that exists only in simulation, IP held by the university, or part-time founders.