Meta offers a glimpse through its supposed iPhone killer: Orion

Image Credits: David Paul Morris/Bloomberg via Getty Images / Getty Images

For years, Silicon Valley and Wall Street have questioned Mark Zuckerberg’s decision to invest tens of billions of dollars into Reality Labs. This week, Meta’s wearables division unveiled a prototype of its Orion smart glasses, a form factor the company believes one day could replace the iPhone. That idea sounds crazy… but maybe a little less crazy than it did a week ago.

Orion is a prototype headset that combines augmented reality, eye and hand tracking, generative AI, and a gesture-detecting wristband. Through micro LED projectors and silicon carbide lenses (which are quite expensive), Meta seems to have cracked a longstanding AR display challenge. The idea is that you can look through Orion — you know, like a pair of glasses — but also see application windows projected on the lenses that appear as if they’re embedded in the world around you. Ideally, you can use your hands, eyes, and voice to navigate the environment.

Meta Orion
The Orion smart glasses need a wristband and wireless compute puck to work. (Meta)
Image Credits: Meta

Though to be clear, Meta’s Orion smart glasses are chunkier than your average readers, reportedly cost $10,000 a pop, and won’t be available for sale anytime soon. We’re talking years from now. All the technology in Orion is relatively young, and all of it needs to get cheaper, better, and smaller to work its way into a pair of smart glasses you can buy at the mall. Zuckerberg says the company has already been working on Orion for 10 years, but there’s still no path to a sellable product.

However, Meta is hardly the only company trying to put a smartphone replacement on your face.

This month, Snap unveiled its latest generation of Spectacles smart glasses, which are larger than Orion and have a more limited field of view. One former Snap engineer called the latest Spectacles “obviously bad” — though you can actually order them. Google hinted during its I/O conference in May that it, too, is working on a pair of smart glasses, perhaps a revamp of its failed Google Glass experiment from last decade. Apple is reportedly working on AR glasses that sound a lot like Orion. And we can’t rule out Jony Ive’s new startup, LoveFrom, which he recently confirmed is working on an AI wearable with OpenAI (though we don’t know if they’re glasses, a pin, or something else entirely).

What’s brewing is a race among Big Tech’s richest companies to create a sleek pair of smart glasses that can do everything your smartphone can — and hopefully something more. Meta’s prototype made two things clear: there is something there, but we’re not “there” yet.

These devices are a notable departure from the Quest virtual reality headsets Meta has been pushing for years now, and Apple’s Vision Pro. There’s a lot of similar technology involved, like eye-tracking and hand tracking, but they feel completely different to use. VR headsets are bulky, uncomfortable to wear, and make people nauseous from staring at the displays. Sunglasses and eyeglasses, on the other hand, are relatively pleasant to wear and millions of Americans use them everyday.

To Zuckerberg’s credit, he’s been pushing the eyewear form factor for quite a long time, when it certainly was not popular to do so. It’s long been reported that Meta’s CEO hates that his popular social media apps have to be accessed through Apple’s phones (perhaps leading to the ill-fated Facebook Phone). Now, Meta’s competitors are also dipping their toes into eyewear computing.

Andrew Bosworth, CTO of Meta and head or Reality Labs, wearing a clear pair of Orion smart glasses. (David Paul Morris/Bloomberg via Getty Images)

Meta’s early investment here seems to be paying off. Zuckerberg gave a keynote presentation of Orion on Wednesday that we won’t be forgetting anytime soon, filling a room full of skeptical journalists with electricity and excitement. TechCrunch has not demoed Orion yet, but initial reviews have been very positive.

What Meta offers today is the Ray-Ban Meta: a pair of glasses with cameras, microphones, speakers, sensors, an on-device LLM, and the ability to connect to your phone and the cloud. The Ray-Ban Meta is far simpler than Orion, but relatively affordable at $299 — actually not much more than a regular pair of Ray-Bans. They’re kind of like the Spectacles 3 that Snap released a few years ago, though the Ray-Ban Meta glasses appear more popular.

Despite the vast differences in price and capabilities, Orion and Ray-Ban Meta are more related than you might think.

“Orion is really the future, and we ultimately want to go for the full holographic experience. You can think about Ray-Ban Meta as our first step there,” said Li-Chen Miller, a VP of product at Meta who leads its wearables team, in an interview with TechCrunch. “We really need to nail the basic things, like making sure it’s comfortable, people want to wear it, and that people find value in it every day.”

One of the things Meta is trying to nail with Ray-Ban Meta is AI. Currently, the smart glasses use Meta’s Llama models to answer questions about what you see in front of you, by taking pictures and running them through the AI system alongside a user’s verbal requests. The Ray-Ban Meta’s AI features today are far from perfect: The latency is worse than OpenAI’s natural-feeling Advanced Voice Mode; Meta AI requires very specific prompts to work right; it hallucinates; and it doesn’t have a tight integration with many apps, making it less useful than just picking up my iPhone (perhaps by Apple’s deisgn). But Meta’s updates coming later this year try to address these issues.

Li-Chen Miller, VP of product during Meta Connect in 2023. (David Paul Morris/Bloomberg via Getty Images)

Meta announced it will soon release live AI video processing for their Ray-Bans, meaning the smart glasses will stream live video and verbal requests into one of Llama’s multimodal AI models and will produce real-time, verbal answers based on that input. It’s also getting basic features, like reminders, as well as more app integrations. That should make the whole experience a lot smoother, if it works. Miller says these improvements will filter up to Orion, which runs on the same generative AI systems.

“Some things make more sense for one form factor than the other, but we’re certainly cross-pollinating,” said Miller.

Likewise, she says some of Orion’s features may filter down as her team focuses on making the AR glasses more affordable. Orion’s various sensors and eye trackers are not cheap technologies. The problem is that Orion has to get both better and more economical.

Another challenge is typing. Your smartphone has a keyboard, but your smart glasses won’t. Miller worked on keyboards at Microsoft for nearly 20 years before joining Meta, but she says Orion’s lack of keyboard is “freeing.” She argues that using smart glasses will be a more natural experience than using a phone. You can simply talk, gesture with your hands, and look at things to navigate Orion; all things that come naturally to most people.

Another device that was criticized for lacking a keyboard was, ironically, the iPhone. Former Microsoft CEO Steve Ballmer infamously laughed at the iPhone in 2007, saying it wouldn’t appeal to business customers because it didn’t have a physical keyboard. People adapted though, and his comments sound naive more than 15 years later.

I think making Orion feel natural is definitely more of a goal than a reality at this point. The Verge notes in its hands-on review that windows occasionally filled the entire glasses lens, completely obstructing the user’s view of the world around them. That’s far from natural. To get there, Meta will have to improve its AI, typing, AR, and a long list of other features.

“For Ray-Ban Meta, we kept it very scoped to a few things, and then it does them really well,” said Miller. “Whereas, when you want to build a new, futuristic computing platform [with Orion], we have to do a lot of things, and do them all very well.”

Flying through Seattle's hacked airport

Image Credits: Devin Coldewey/TechCrunch

Several days after the Port of Seattle announced a “possible” cyberattack on its systems, Seattle-Tacoma Airport is still largely offline, causing chaos among travelers and acting as a standing warning against taking cybersecurity lightly. Ask me how I know.

The outage resulting from the recent hack has not, fortunately, caused planes to fall out of the sky or Air Traffic Control to double-book a runway. Those resources, run by the feds, are considerably more locked down.

Rather than catastrophe, what we have now — and for the foreseeable future, since authorities have offered no timeline for restoration — is an object lesson in why we have rules about where we put our eggs.

For my part, I found out on Sunday when — and I hesitate even to mention it, because no one seems to know about this miraculous service — I went to reserve my place in the security line via the SEA Spot Saver. The service was offline, and throwing the kind of error that you don’t have to be a sysadmin to know means deeper problems.

If I had been a good reporter and read my own publication over the weekend, I would have known this was the result of, among other things, the entire user-facing DNS configuration of the Port’s web architecture being totally cooked. (The Spot Saver site is still offline, but the function has been resuscitated by Clear for now.)

Luckily I was not checking a bag and security was light, possibly due to a jackknifed semi blocking all southbound traffic on I-5.

At the airport, the large screens one would ordinarily loiter under to find one’s flight were ominously dark. But considering the endless construction at Sea-Tac, I chalked this up to electrical work.

It was only at the “S” gates that the extent of the problem became clear. Every screen in the area was dark; the TVs above the waiting areas, the multi-display arrays directing travelers to gates, the monitors of the gate agents and the gate info displays themselves.

Though my boarding pass had directed me to a gate, there was no way to be sure that was the correct one, so I checked with the agents there. They confirmed it, and I asked about the hack.

“It definitely is a bit of a… show,” the airline agents agreed, politely eliding the same part of the word I had. All airport systems shared by multiple airlines were down. Baggage handling, they said, was getting the worst of it. The agents were (tell no one!) ignoring their own baggage size rules and didn’t bother collecting “volunteers” to gate-check bags and speed up boarding. Inter-airline communications were labored.

The gate desk was mostly offline, I was told, as it’s a shared system between Alaska, Delta and anyone else who comes to the “S” gates. The gate was unable to display the flight number, boarding groups or any delays — a half-hour for my flight — except over the public address system — which was extremely competitive due to the need to constantly repeat current gate numbers. Nearby, one gate had paper signs announcing the flight that had last departed, though that was obviously hours earlier. (Sea-Tac airport spokesperson Perry Cooper told me in an email that my experience was “not typical of the rest of the airport.”)

a photo of the S4 gate at Seattle-Tacoma airport with switched-off displays with a piece of paper taped to the screen, saying ICELANDAIR FI680.
Gate S-4 at Seattle-Tacoma airport, with no gate information due to the cyberattack.
Image Credits: Devin Coldewey/TechCrunch

The tablets for checking people in were working, “but limited,” the agents said. Changing flights or seats was not happening. (“I think maybe I got upgraded to first,” I ventured hopefully, but they just shooed me away.)

In situations where the digital infrastructure crashes, it can happen that those who cling to analog resources look smart rather than quaint. Not so today. As I waited, every few minutes someone would walk up to the gate with a paper ticket telling them this was where they departed. Some were lucky enough to be told it was just a few steps away, while one unfortunate soul was redirected all the way to the “N” gates — the polar opposite, as you may imagine, of the “S” gates.

The solution, as proffered by gate agents and paper signs taped to blank displays alike, was to use the app. But it’s precisely because of problems like this week’s that no one can ever really trust “the app,” because “the app” is as likely to get the hacker treatment as the rest of the Port.

What was extraordinary was that a suspected malicious hacker was able to tank so many systems in one go. We don’t have to expect that the baggage direction, gate guidance and security handling can’t be completely siloed and separate. This is an airport, not a nuclear power plant.

Yet at the same time it seems wrong that the resilience of the system is so lacking. Sure, the airport intranet might go down — but the full-on public-facing website? Baggage routing and gate updates, too? All on the same network? We’ve understood the necessity of breaking apart critical systems for centuries, and have built it into our power and network infrastructure so that when one person runs two hairdryers at the same time, it doesn’t knock out the whole neighborhood.

I’m not complaining because I was inconvenienced. To be honest, this airport trip was no better or worse for me personally than any other. But I saw countless people being put out due to what amounts to badly secured, probably woefully understaffed government IT infrastructure.

When the feds talk about refurbishing critical infrastructure, this is what they’re talking about. Yes, it’s also the ’80s-era computer running on COBOL that controls the traffic lights, dams or missile silos. But it’s events like this — not so much the recent CrowdStrike outage debacle, actually — that really show the soft, vulnerable underbelly of local and national systems. Critical infrastructure, like airports, have a disturbingly large attack surface that have comparatively few resources dedicated to their upkeep.

It’s not that an airport isn’t as valuable of a target as, say, a financial institution or a data broker, but that’s changing. Ransomware, for instance, has proven highly profitable and easy to automate, and AI (you knew it had to figure somewhere) is supercharging credential theft via spear-phishing operations. All this to say that the trend of unlikely targets — schools, libraries and hospitals — being held to ransom is only going to intensify — but these attacks can be prevented, just as they can in private industry where they have expected them for decades.

Anyone traveling through Sea-Tac should definitely budget a bit more time to get through the airport and install the relevant apps. State and city authorities are doing their best to keep everyone informed on this crisis page.

Flint Capital raises a $160M through an unusual fund-raising strategy

Image Credits: Yuri Samoilov (opens in a new window) / Flickr (opens in a new window) (Image has been modified)

Boston-based Flint Capital just closed its third fund at $160 million, four times the amount of its initial 2013 fund. The capital will be split evenly between early and late stage investments, with the firm doubling down on IT, cybersecurity, fintech and digital health startups. Its success is in large part because of a unique strategy over who its courts as limited partner investors. 

The firm is over a decade old, founded by partner Dmitry Smirnov, who was previously CEO of Russia-based investment firm FINAM Global. He immediately made an unorthodox decision: instead of pursuing traditional LPs like pension funds or endowments, he sought out IT entrepreneurs, believing they would want a front row seat to the next generation of technology. 

Sergey Gribov, one of Flint’s three partners, said the firm also has a global mandate and invests strongly in Europe and Israel — as long as the startup has its eyes on expanding into the US. “We don’t really care where physically the team is located, as long as we go off to the US market,” he said. 

That’s been a good strategy for Flint: the firm has backed identity verification startup Socure, last valued at $4.5 billion, adoption platform WalkMe, which was acquired by SAP for $1.5 billion, and Flo, the women’s health app recently valued at over $1 billion. 

For this latest fund, partner Andrew Gershfeld highlighted that several investors were actually founders that Flint backed years ago. He gave the example of Nir Giller and Omer Schneider, the founders of CyberX, a cybersecurity company that Microsoft acquired in 2020. For Gershfeld, founders like these reinvesting their profits into Flint was a sign “that we were doing something right.” 

Flint’s successful fundraise is a vote of confidence amongst a dire fundraising atmosphere for smaller, or younger emerging funds. This year, funding for venture firms is the lowest it’s been since 2019, and, of the few that have secured capital, established firms are taking a bigger and bigger cut of the pie, according to the Q2 2024 Pitchbook-NVCA Venture Monitor.  

It took the Flint partners 18 months to fundraise and, while the fund was anchored by previous investors, they felt the sluggishness of the current market. “The conversion from that first conversation into becoming a limited partner dropped during this year,” Gershfeld said. “It’s a fact – we can’t say that it is not the case.”

The fundraise is particularly impressive as the partners have spent the last year helping their Israeli startups, like Cynomi and Sensi.AI, fundraise throughout the war in Gaza. Gribov, who regularly travels to Israel, recalled video-chatting with founders decked out in combat gear, or coaching companies who had portions of their workforce pulled into the military. His efforts paid off: Sensi.AI, a digital health startup, closed its $31 million Series B in late June. 

Gribov said that seeing these companies thrive despite the global conflict made him more confident than ever in Flint’s global mandate. “A lot of companies continued to deliver and even outperform,” he said.

Flint Capital raises a $160M through an unusual fund-raising strategy

Image Credits: Yuri Samoilov (opens in a new window) / Flickr (opens in a new window) (Image has been modified)

Boston-based Flint Capital just closed its third fund at $160 million, four times the amount of its initial 2013 fund. The capital will be split evenly between early and late stage investments, with the firm doubling down on IT, cybersecurity, fintech and digital health startups. Its success is in large part because of a unique strategy over who its courts as limited partner investors. 

The firm is over a decade old, founded by partner Dmitry Smirnov, who was previously CEO of Russia-based investment firm FINAM Global. He immediately made an unorthodox decision: instead of pursuing traditional LPs like pension funds or endowments, he sought out IT entrepreneurs, believing they would want a front row seat to the next generation of technology. 

Sergey Gribov, one of Flint’s three partners, said the firm also has a global mandate and invests strongly in Europe and Israel — as long as the startup has its eyes on expanding into the US. “We don’t really care where physically the team is located, as long as we go off to the US market,” he said. 

That’s been a good strategy for Flint: the firm has backed identity verification startup Socure, last valued at $4.5 billion, adoption platform WalkMe, which was acquired by SAP for $1.5 billion, and Flo, the women’s health app recently valued at over $1 billion. 

For this latest fund, partner Andrew Gershfeld highlighted that several investors were actually founders that Flint backed years ago. He gave the example of Nir Giller and Omer Schneider, the founders of CyberX, a cybersecurity company that Microsoft acquired in 2020. For Gershfeld, founders like these reinvesting their profits into Flint was a sign “that we were doing something right.” 

Flint’s successful fundraise is a vote of confidence amongst a dire fundraising atmosphere for smaller, or younger emerging funds. This year, funding for venture firms is the lowest it’s been since 2019, and, of the few that have secured capital, established firms are taking a bigger and bigger cut of the pie, according to the Q2 2024 Pitchbook-NVCA Venture Monitor.  

It took the Flint partners 18 months to fundraise and, while the fund was anchored by previous investors, they felt the sluggishness of the current market. “The conversion from that first conversation into becoming a limited partner dropped during this year,” Gershfeld said. “It’s a fact – we can’t say that it is not the case.”

The fundraise is particularly impressive as the partners have spent the last year helping their Israeli startups, like Cynomi and Sensi.AI, fundraise throughout the war in Gaza. Gribov, who regularly travels to Israel, recalled video-chatting with founders decked out in combat gear, or coaching companies who had portions of their workforce pulled into the military. His efforts paid off: Sensi.AI, a digital health startup, closed its $31 million Series B in late June. 

Gribov said that seeing these companies thrive despite the global conflict made him more confident than ever in Flint’s global mandate. “A lot of companies continued to deliver and even outperform,” he said.

ServiceNow office building in Silicon Valley

ServiceNow is developing AI through mix of building, buying and partnering

ServiceNow office building in Silicon Valley

Image Credits: Andrei Stanescu / Getty Images

Every enterprise software company out there is working to bring more workflow automation and AI to the platform. ServiceNow has been on this journey for some time now, and given the kind of data it collects via interactions on its platform, it’s building more refined models.

Part of the shift to AI comes internally by building, some via acquisitions and some from partnering widely, says SVP of corporate business development Philip Kirk. But whatever the source, it’s all in service building a stronger platform, he says.

“It is kind of three dimensional chess right now to figure out whether to build, buy or partner. I think the biggest thing that we try to prioritize is how we can make decisions that are in the long-term best interest of our customers, and that differentiate us from what we know we’re world-class at, which is enterprise automation in our platform,” Kirk told TechCrunch.

Lara Greden, an analyst at IDC who covers ServiceNow, says going beyond building is a big part of every company’s strategy when it comes to AI. “Acquisition and strategic partnerships are an essential element of corporate strategy in the AI era,” Greden told TechCrunch.

“Like other major waves of technology innovation, breakthrough capabilities in generative AI are coming through entities that have laser-focused on the tech itself, in other words: startups. ServiceNow’s acquisition strategy has been in line with our expectations for being a leader in accelerating value from generative AI.”

On the heels of IBM partnership, Celonis announces one with ServiceNow

ServiceNow has built on that in its latest releases, dubbed Washington DC. The company has embraced generative AI in a way that makes sense in the context of the information that the platform monitors and collects, says Jeremy Barnes, VP of AI products at ServiceNow, who came to the company when it acquired his previous one, Element AI, at the end of 2020. That means providing the kind of features for customers looking to take advantage of generative AI in a customer service context without having to build it themselves.

“And so if you look at what’s coming up in the Washington release, we provide all kinds of features that companies would not have really gone out and built, or people who want to develop AI projects would not have been able to pull together proof of concept, and us building it in the platform for them just makes total sense,” he said.

Keith Kirkpatrick, an analyst at the Futurum Group, says ServiceNow is making it easier for people who have some domain knowledge to build things like intelligent workflows without bringing in a developer or workflow expert.

“ServiceNow’s focus has been on integrating generative AI to improve entire workflows, not just single processes or tasks. This is a critical point of differentiation for them, as it allows for intelligent automation of multi-step processes that once required a significant amount of effort and switching between applications to complete,” Kirkpatrick said.

Virtual agents or AI agents are also taking center stage in this release, which would make sense given ServiceNow’s capabilities helping answer customer questions and complete tasks. And just as ServiceNow is building AI capabilities for its customers, it’s working with partners where it makes sense and it’s outside of their areas of expertise, Greden said

“The company’s focus on the conversational capabilities in Virtual Agent is important because getting conversational generative AI right offers significant potential for ROI for end customers. However, getting it right is not easy,” she said. “That’s one reason platform vendors are strategic partners of choice to organizations in the AI era, because they can take on the data science work of getting the technology to work and work well. They also play an important role in guiding customers with best practices around data governance and control.”

The Washington release is available starting on Wednesday for all ServiceNow customers.

Salesforce is betting that its own content can bring more trust to generative AI