Topic

edge cases

3 signals

Infrastructure2 hours ago

Building Fixed Hardware Implementations of Neural Networks

Researchers from Yale, Cornell, Boston University, and NTT Research published "Physical Foundation Models: Fixed hardware implementations of large-scale neural networks," arguing that the ~1-year release cadence of trillion-parameter foundation models justifies building fixed, special-purpose hardware where the network is realized directly in physical structure (such as 3D nanostructured glass) rather than executed on general-purpose digital inference chips.

Why it matters

The economic argument here matters more than the optics demo. The authors are saying that once a single foundation model gets reused across thousands of downstream tasks, the amortization math flips: it becomes rational to tape out hardware whose weights are physically baked in, on the same yearly cadence as model releases. That is a direct challenge to the Nvidia general-purpose GPU thesis for inference, and it is the same logic Groq, Etched (Sohu), and MatX are already pursuing in digital silicon. PFMs push it one step further into analog physics.

For operators, the near-term read is not glass holograms running GPT-class models. It is edge inference: trillion-parameter capability in a power envelope a robot or drone can carry, which today caps out at far smaller models. If even a fraction of the claimed energy-efficiency gain materializes in nanoelectronic form, the on-device vs cloud calculus for humanoids and AVs changes. Watch NTT Research specifically; they have been quietly funding coherent optical compute for years and are the most likely party to move this from arXiv to a fab partnership.

LaunchesMay 21

Waymo halts freeway rides after robotaxis struggle in construction zones

Source: Transportation & Auto News | TechCrunch

The service suspension comes as Waymo has also paused operations in Atlanta and San Antonio after its robotaxis drove into flooded streets.

Why it matters

Waymo pulled freeway service in its four largest markets (SF, LA, Phoenix, Miami) less than six months after launching it, and the timing matters: freeway routing was the unlock for airport connections and the one million paid rides per week target by year-end 2026. Surface-street-only operation caps trip length and kills the peninsula commute use case that was driving Bay Area ride growth. Construction zones are the failure mode operators should have expected, since cones, lane shifts, and human flaggers are exactly the long-tail scenarios HD maps cannot pre-encode.

Stack this with the Atlanta and San Antonio flood pauses and the pattern is environmental edge cases compounding faster than Waymo can ship recalls. For competitors (Zoox, Tesla, and the Chinese players watching closely) the read is that L4 freeway deployment is gated by construction-zone perception, not policy. For Waymo's own roadmap, the Zeekr-built Ojai launch now arrives into a service map that is visibly contracting.

LaunchesMay 21

Waymo pauses Atlanta and San Antonio service after robotaxis repeatedly drive into floods

Source: TechCrunch

Waymo has suspended robotaxi operations in both Atlanta and San Antonio after vehicles repeatedly drove into flooded roads. The company is working on software updates to prevent the behavior before resuming service in the two cities.

← All signals