TLDR Hardware 2026-10-30

Apple's new 14- and 16-inch MacBook Pro uses a tandem OLED panel rated at 2,000 nits peak HDR and SDR brightness with a 10,000,000:1 contrast ratio and 10 to 120Hz refresh, with the touch sensor built into the panel rather than layered on top. It drops to 3 and 4 pounds, adds a Dynamic Island, and runs M5 Pro or M5 Max with up to 128GB of memory. Pricing starts at $2,899, shipping November 6.

PsiQuantum unveiled an 8×8 optical circuit switch built on the silicon photonics platform it developed for quantum computing, measuring typical insertion loss of 1.2 dB, 40 dB crosstalk, and under 3 ms reconfiguration without amplification. Samples are going to select customers, with commercial products promised in the coming months. A 64×64 barium titanate switch used in its quantum systems isn't needed for this first version but offers a path to faster future parts.

Altera announced Agilex 2X, four FPGA families spanning high-end to I/O-centric parts, led by Agilex 27 with a third-generation HyperFlex fabric claiming 1.8x performance per watt over Agilex 7. It adds a chip-wide HyperFlow data network, transceivers up to 256 Gbps, hardened 800GbE and 1.6TbE, PCIe 6.0 and 7.0, LPDDR6, UCIe, and post-quantum-ready secure boot. Quartus support starts in mid-November, with engineering samples in the second half of 2027.

Eclypsium demonstrated changing a model's weights in memory after integrity checks pass, so the model file, container, hash, and signatures on disk all stay valid. Attack paths run through host firmware, management controllers, and DPUs, and the report says Nvidia, AMD, Intel, and Arm systems are all exposed. Its fix is DPUs that enforce security policy independently of the AI system, plus full component inventories.

MIT built a tethered soft enteroscope to cover the small intestine, a 20- to 25-foot tube that double balloon enteroscopy currently covers slowly by hand. Four-chamber inflatable actuators bend and extend the body, anchoring balloons at the head and tail take turns gripping, and kirigami sleeves pop out needles for traction. In animal tests, it advanced about 6.5 feet and used its camera, drug sprayer, and biopsy clips.

Seoul Biosys has started producing rare earth-free white micro LEDs using WICOP technology, which eliminates the need for traditional packaging. The new LEDs measure 25µm x 10µm and achieve 0.015 lumens, with plans to double light efficiency to 200 lumens per watt. This innovation aims to reduce manufacturing costs and environmental impact, targeting applications in smart glasses, AR displays, and AI data transmission.

Marvell is in advanced talks to buy Lumentum, which makes the lasers and optical modules that move data between AI servers, with an announcement possible as soon as next week. Lumentum rose about 11% after hours to a roughly $91 billion valuation, while Marvell fell more than 8%. Marvell held only about $3.9 billion in cash last quarter. Neither company has confirmed the talks.

SpaceX no longer depends on NASA contracts to survive, while NASA depends heavily on SpaceX's rockets and capsules, which cuts the agency's negotiating leverage. SpaceX told NASA it will stop flying Crew Dragon by or before 2030 and has stopped selling commercial Falcon 9 launches for missions after 2028. That would leave NASA's ISS crew transport and station deorbit plans exposed.

Intel, in partnership with SAP, tested an open robotics architecture at its Chengdu site, integrating SAP's Warehouse Management with AI and autonomous mobile robots.

Inside-out tracking took years to survive real rooms, where glass, reflections, moving people, curtains, and ceiling fans break the scene model.

Google and XREAL launched AURA, transparent-lens glasses that run Android XR and project apps across a 70-degree view.

FuriosaAI raised $620 million at a $2.9 billion valuation, about four times its prior round, to scale its RNGD inference chip, a 180 W part it claims matches GPU throughput at half the power or less.

Reuters reports the DOJ is questioning whether owning the main distribution hub for open models would let Nvidia disadvantage rival chip and AI companies or gain leverage in negotiations.