Mamdani’s campaign did not respond to a request for comment.

The NYPD’s turn toward mass surveillance was begun in earnest by Commissioner Raymond Kelly during the immediate aftermath of September 11, buoyed by hundreds of millions of dollars in federal anti-terrorism grants. However, Ferguson says Kelly’s rival, former commissioner William Bratton, was a key architect behind the NYPD’s reliance on “big data,” by implementing the CompStat data analysis system to map and electronically collate crime data during the mid-1990s and again during his return to New York City in 2014 under Mayor Bill de Blasio. Bratton was also a mentor to Jessica Tisch and has spoken admiringly of her since leaving the NYPD.

Tisch was a main architect of the NYPD’s Domain Awareness System, an enormous, $3 billion, Microsoft-based surveillance network of tens of thousands of private and public surveillance cameras, license plate readers, gunshot detectors, social media feeds, biometric data, cryptocurrency analysis, location data, bodyworn and dashcam livestreams, and other technology that blankets the five boroughs’ 468-square-mile territory. Patterned off London’s 1990s CCTV surveillance network, the “ring of steel” was initially developed under Kelly as an anti-terrorism surveillance system for Lower and Midtown Manhattan before being rebranded as the DAS and marketed to other police departments as a potential for-profit tool. Several dozen of the 17,000 cameras in New York City public housing developments were also linked through backdoor methods by the Eric Adams administration last summer with thousands more in the pipeline, according to NY Focus.

Though the DAS has been operational for more than a decade and survived prior challenges over data retention and privacy violations from civil society organizations like the New York Civil Liberties Union, it remains controversial. In late October, a Brooklyn couple filed a civil suit along with Surveillance Technology Oversight Project (STOP), a local privacy watchdog, against the DAS, alleging violations of New York State’s constitutional right to privacy by the NYPD’s persistent mass surveillance and data retention. NYPD officers, the suit claims, can “automatically track an individual across the city using computer vision software, which follows a person from one camera to the next based on descriptors as simple as the color of a piece of clothing.” The technology, they allege, “transforms every patrol officer into a mobile intelligence unit, capable of conducting warrantless surveillance at will.”

If you’re like me, Steve Kornacki is just as adored by your aunt as he is in your group chats. He’s become a staple of Election Day coverage, putting in long hours at the big board and copious amounts of prep beforehand.

His granular knowledge of key counties and voter turnout trends made him not just indispensable for many Americans on election night, but also a full-blown celebrity. I caught up with him bright and early this morning to talk about Tuesday night’s election results.

We broke down what the returns mean heading into the 2026 midterm elections, where Democrats currently hold an 8 percentage point advantage over Republicans in the latest NBC News poll, and what they say about President Donald Trump’s second-term agenda. We also spoke about what surprised him in the New Jersey governor’s race, whether Trump’s base is weakening, and, of course, New York mayor-elect Zohran Mamdani’s historic win. Heading into the midterms, Kornacki is taking on an expanded role at NBC News following parent company Comcast’s decision to spin off its cable TV properties, including a soon-to-be rebranded MSNBC.

Kornacki is not someone to put too much stock into an off-year election, but the breadth and depth of Democratic victories suggested a political environment that’s radically changed in the year since Trump’s election—and if anyone can find some important details to follow going forward, it’s Steve.

This interview has been edited for length and clarity.

WIRED: Steve, thanks for joining us after a long night. Before we get into the meat and potatoes here, let’s start with a quick lightning round: How many hours of sleep were you shooting for, how many did you get, and can you tell us if you have any election night superstitions?

Steve Kornacki: Well, I shoot for zero, so I’m not disappointed and therefore I’m pleasantly surprised with whatever I get, which I think was about two and a half last night.

There we go.

So that’s not too bad. Superstitions? I don’t know about that. My challenge is to just tune out all the anecdotal turnout data on Election Day. I just think it’s a ton of noise that starts messing with your head.

What surprised you from last night?

What surprised me was—it’s probably not the most original observation this morning—but New Jersey. [Representative Mikie Sherrill, the Democratic nominee, won with more than 56 percent of the vote.] The margin there for Sherrill, which is about 13 points, is much more than expected. I mean, I was talking to Democrats right up through Election Day who were telling me some version of: “She’s run a terrible campaign, she’s not been a good candidate. Maybe she’ll still win because of Trump, but this is going to be closer than it should be.” I mean, that was a widely shared view between the two parties, that Sherrill had run a bad campaign and was in danger of even losing, and that was not the case at all.

A robot searching for workers trapped in a partially collapsed mine shaft must rapidly generate a map of the scene and identify its location within that scene as it navigates the treacherous terrain.

Researchers have recently started building powerful machine-learning models to perform this complex task using only images from the robot’s onboard cameras, but even the best models can only process a few images at a time. In a real-world disaster where every second counts, a search-and-rescue robot would need to quickly traverse large areas and process thousands of images to complete its mission.

To overcome this problem, MIT researchers drew on ideas from both recent artificial intelligence vision models and classical computer vision to develop a new system that can process an arbitrary number of images. Their system accurately generates 3D maps of complicated scenes like a crowded office corridor in a matter of seconds. 

The AI-driven system incrementally creates and aligns smaller submaps of the scene, which it stitches together to reconstruct a full 3D map while estimating the robot’s position in real-time.

Unlike many other approaches, their technique does not require calibrated cameras or an expert to tune a complex system implementation. The simpler nature of their approach, coupled with the speed and quality of the 3D reconstructions, would make it easier to scale up for real-world applications.

Beyond helping search-and-rescue robots navigate, this method could be used to make extended reality applications for wearable devices like VR headsets or enable industrial robots to quickly find and move goods inside a warehouse.

“For robots to accomplish increasingly complex tasks, they need much more complex map representations of the world around them. But at the same time, we don’t want to make it harder to implement these maps in practice. We’ve shown that it is possible to generate an accurate 3D reconstruction in a matter of seconds with a tool that works out of the box,” says Dominic Maggio, an MIT graduate student and lead author of a paper on this method.

Maggio is joined on the paper by postdoc Hyungtae Lim and senior author Luca Carlone, associate professor in MIT’s Department of Aeronautics and Astronautics (AeroAstro), principal investigator in the Laboratory for Information and Decision Systems (LIDS), and director of the MIT SPARK Laboratory. The research will be presented at the Conference on Neural Information Processing Systems.

Mapping out a solution

For years, researchers have been grappling with an essential element of robotic navigation called simultaneous localization and mapping (SLAM). In SLAM, a robot recreates a map of its environment while orienting itself within the space.

Traditional optimization methods for this task tend to fail in challenging scenes, or they require the robot’s onboard cameras to be calibrated beforehand. To avoid these pitfalls, researchers train machine-learning models to learn this task from data.

While they are simpler to implement, even the best models can only process about 60 camera images at a time, making them infeasible for applications where a robot needs to move quickly through a varied environment while processing thousands of images.

To solve this problem, the MIT researchers designed a system that generates smaller submaps of the scene instead of the entire map. Their method “glues” these submaps together into one overall 3D reconstruction. The model is still only processing a few images at a time, but the system can recreate larger scenes much faster by stitching smaller submaps together.

“This seemed like a very simple solution, but when I first tried it, I was surprised that it didn’t work that well,” Maggio says.

Searching for an explanation, he dug into computer vision research papers from the 1980s and 1990s. Through this analysis, Maggio realized that errors in the way the machine-learning models process images made aligning submaps a more complex problem.

Traditional methods align submaps by applying rotations and translations until they line up. But these new models can introduce some ambiguity into the submaps, which makes them harder to align. For instance, a 3D submap of a one side of a room might have walls that are slightly bent or stretched. Simply rotating and translating these deformed submaps to align them doesn’t work.

“We need to make sure all the submaps are deformed in a consistent way so we can align them well with each other,” Carlone explains.

A more flexible approach

Borrowing ideas from classical computer vision, the researchers developed a more flexible, mathematical technique that can represent all the deformations in these submaps. By applying mathematical transformations to each submap, this more flexible method can align them in a way that addresses the ambiguity.

Based on input images, the system outputs a 3D reconstruction of the scene and estimates of the camera locations, which the robot would use to localize itself in the space.

“Once Dominic had the intuition to bridge these two worlds — learning-based approaches and traditional optimization methods — the implementation was fairly straightforward,” Carlone says. “Coming up with something this effective and simple has potential for a lot of applications.

Their system performed faster with less reconstruction error than other methods, without requiring special cameras or additional tools to process data. The researchers generated close-to-real-time 3D reconstructions of complex scenes like the inside of the MIT Chapel using only short videos captured on a cell phone.

The average error in these 3D reconstructions was less than 5 centimeters.

In the future, the researchers want to make their method more reliable for especially complicated scenes and work toward implementing it on real robots in challenging settings.

“Knowing about traditional geometry pays off. If you understand deeply what is going on in the model, you can get much better results and make things much more scalable,” Carlone says.

This work is supported, in part, by the U.S. National Science Foundation, U.S. Office of Naval Research, and the National Research Foundation of Korea. Carlone, currently on sabbatical as an Amazon Scholar, completed this work before he joined Amazon.