TechCrunch+

To get a roundup of TechCrunch’s biggest and most important crypto stories delivered to your inbox every Thursday at 12 p.m. PT,subscribe here.

Welcome back to Chain Reaction.

For the first time in a long time, there was less talk circulating about a certain three-letter crypto exchange, ahem, FTX. But of course, that didn’t last long as the former FTX CEO (who also goes by three letters — SBF) created a Substack and shared his perspective on Thursday — in a 2,300-word overview — detailing what FTX and Alameda were “pre-mortem.”

Over the past seven days, there have also been a handful of cuts announced across the crypto space, from the second-largest exchange Coinbase cutting 20% of its workforce to NFT marketplace SuperRare axing 30% of its staff.

This is Coinbase’s second round of major layoffs, after eliminating 18% of its employees, or about 1,100 jobs last June, but there was “no way to reduce our expenses significantly enough, without considering changes to headcount,” Coinbase co-founder and chief executive Brian Armstrong wrote in a post Tuesday.

Amid the doomy and gloomy sentiment from these layoffs, a number of market players were trying to help those affected by the cuts, including Solana co-founder Anatoly Yakovenko, who tweeted, “If you are a departing Coinbase employee, reach out! I would love to find you a home in the ecosystem.”

See, not all hope in humanity has to be lost.

It was also a big week for partnerships as more crypto companies joined forces with mainstream Web 2.0 and financial companies like AWS and Mastercard. But instead of saying more here, I’ll let the stories speak for themselves. Read more about them below.

This week in web3

AWS partners with Avalanche to scale blockchain solutions for enterprises, governments

Amazon Web Services (AWS) has partnered with Ava Labs, the company building out layer-1 blockchain Avalanche, to help scale blockchain adoption across enterprises, institutions and governments, the two firms exclusively told TechCrunch. “Looking forward, web3 and blockchain is inevitable,” Howard Wright, VP and global head of startups at AWS, said to TechCrunch. “No one can call the time or date or quarter that it’s going to happen and it’ll be mainstream, but we’ve seen the cycles of growth before. The velocity of this one seems like it’s accelerating and we’re just excited to be a part of this.”

6 crypto investors talk about DeFi and the road ahead for adoption in 2023 (TC+)

The crypto venture capital industry has become more selective thanks to the general market downturn and wavering trust caused by a slew of scandals and market disruptions, but investors at major firms are still writing checks in the space. As the market looks toward the future, some venture capitalists are revamping their investing strategies, while others are holding to their current plans, with perhaps a small tweak or two. Read on to find out how active investors are thinking about DeFi, how they’re advising their portfolio companies amid the lack of funding, the best way to approach them and more.

Mastercard launches web3-focused artist incubator with Polygon

Mastercard, one of the biggest financial payments providers in the world, is launching a web3-focused incubator to help artists connect with fans through a new medium, the company shared. Mastercard partnered with Polygon, a scaling blockchain built on top of Ethereum, which has been making huge strides in the Web 2.0 ecosystem lately. After joining the incubator, participating artists should know how to mint NFTs, represent themselves in virtual worlds and establish a community, Raja Rajamannar, chief marketing and communications officer at Mastercard, said to TechCrunch.

Web3-focused Beacon launches flagship demo day with 13 crypto startups

We’re only in the second week of 2023, but demo days have already begun as founders try to keep momentum alive in the ever-changing crypto market. Beacon, a web3-focused early-stage accelerator program, launched last year, and its flagship cohort just graduated. The teams in the first cohort, known as Cohort 0, presented their ideas on Tuesday during a demo day, exclusively covered by TechCrunch.

When it comes to web3, investors say they’re in it for the long haul (TC+)

Hyped or not, web3 companies seem like they’re here to stay, and investors seem more than willing to keep backing them. To get a better idea of how the people writing the checks are thinking about web3, TechCrunch surveyed more than 35 investors, and it turns out the majority are not only actively investing in the category, they also harbor hopes of a shining future for what they feel is a potentially transformative technology.

The latest pod

Chain Reaction is back in action with the launch of Season 2!

For this week’s episode, I talked to Ryan Wyatt, president of Polygon Labs, one of the biggest market shakers and layer-2 blockchains in the crypto space that’s building on top of the Ethereum ecosystem.

The past year has been huge for Polygon as it partnered with big-brand names like Starbucks, Disney and Mastercard to launch loyalty rewards and accelerator programs. Now, Polygon is looking to 2023 and new opportunities, and Wyatt shares what’s in store for it and how the space still has room to grow.

We also discussed:

Polygon’s big themes and product vision for 2023
Mass adoption of crypto and what it takes to get there
Wyatt’s outlook for the gaming and NFT market

Subscribe to Chain Reaction on Apple Podcasts, Spotify or your favorite pod platform to keep up with the latest episodes, and please leave us a review if you like what you hear!

Follow the money

Venom Ventures Fund announced a $1 billion venture fund and made its first investment in Nümi Metaverse’s $20 million funding round
The Easy Company, a “social” crypto wallet, raised $14.2 million in a seed round
Cosmo blockchain-based DeFi protocol Quasar raised $5.4 million at a $70 million valuation
Open Forest Protocol raised $4.1 million to scale nature-based solutions
C14 raised $2.5 million to help build crypto payment flows

This list was compiled with information from Messari as well as TechCrunch’s own reporting.

Can we get more crypto partnerships and less crypto layoffs, please? by Jacquelyn Melinek originally published on TechCrunch

Are you thinking about getting a mechanical keyboard and asked a friend which one to get? Chances are, they said: get a Keychron. It’s basically a meme at this point. After making its name with affordable mechanical keyboards (with a bit of a focus for Mac users) — the company now has something on offer for every budget and in virtually every layout, ranging from the 40% Q9 to the 100% Q6 — and everything in-between. And while the Q-line gets a lot of press, the more affordable pre-built K and V lines go after the non-enthusiast market.

At this point, Keychron has so much of the potential market covered, it’s now starting to offer some more exotic layouts. The latest of these is the Q10, a 75% Alice board with a few twists. It’s probably no surprise that it, too, is an easy recommendation for anybody who is looking for this kind of board.

As a 75% board, the hefty Q10 (it comes in at just under five pounds), with its full aluminum body, offers function keys, number row and dedicated insert, delete, page up/down and home keys. As you would expect from a modern keyboard, there are hot-swap sockets so you can easily try new switches and there is a knob on the left side. And because there’s a bit of extra room here, Keychron added five macro buttons under the knob as well that you can map to anything you’d like, using what has now become the industry-standard VIA app. Like most modern keyboards, it’ll happily work with your Mac or Windows machine, but there’s no wireless option here (that’s coming to Keychron’s Q line soon, though, starting with the upcoming Q1 Pro). Oh, and there’s per-key RGB, too, if that’s your thing.

This is, of course, a board with an Alice layout — that is, the keys are not in a straight line but the left and right half are slightly angled, which some would claim makes for a more ergonomic typing experience. I don’t know about the ergonomics, especially since the keyboard doesn’t tent like something like an ErgoDox EZ would, but at its core, it’s not all that different from some of the more popular ergonomic keyboards from the likes of Microsoft. It’s quite easy to get used to, even for touch-typists, and quite comfortable to type on.

Otherwise, Keychron is going with its standard setup here: gasket mount for a bit of flex and silicon gaskets between the top and bottom cases to reduce ping and other noise. I never quite find that I notice the gasket mounting in daily use, because you really have to hammer on your keys to get any of that bounce you may have seen in YouTube videos, but then I switch to an older keyboard with a tray mount and notice how hard that feels. Your mileage may vary here, of course. At this point, gasket mounts are the industry standard, it seems, but I do look forward to some innovation here.

If all of this sounds familiar, then… yeah… basically everything I said in my review of Keychron’s other — but smaller — Alice-style board, the Q8, still applies here, too. Keychron had some early missteps with its Q1 but to the company’s credit, the team quickly learned from that and the community’s feedback. You can modify this keyboard to your heart’s content, but you don’t have to. Sure, you can improve the stabilizers by lubing them yourself (time to break out that dielectric grease), but the pre-lubed ones are good enough, especially because there’s no large rattling space bar to contend with here anyway.

One nice thing Keychron is doing now with its latest board is pre-modding it. One popular trick among enthusiasts is the tape mod, for example, which involves sticking some painters’ tape at the bottom of the PCB to absorb some of the higher-pitched frequencies as you type. On its new boards, Keychron already applies tape to the bottom of the PCB itself. I think the board could still benefit from some thicker foam at the bottom of the case (but there isn’t a lot of room there, so I haven’t tried yet).

Keychron also uses Gateron G Pro switches, which come pre-lubed (as you can see, there’s a lot of lubing going on in the mechanical keyboard world). And while I continue to hope that Keychron will offer more options than just the linear Red, tactile Brown and clicky Blue switches (yellow would be nice, Keychron…), those are competent options.

My review unit came with brown switches, which some people like, but I find to fall into the uncanny valley between tactile and linear. They just aren’t all that tactile, yet also not quite linear. I prefer linear switches, but if I use a tactile switch, I want to feel that tactile bump as I press down. I had a set of pre-lubed three-pin Gateron CAP Golden Brown switches lying around here (as one does), which are a bit more tactile and made me enjoy the keyboard quite a bit more. They also have a slightly lower pitch than the standard Geteron brown switches.

Keychron’s default keycaps, which come with the pre-built board, are fine. They are double-shot PBT keycaps in the OSA profile. That profile is higher and rounder than you may be used to from Cherry keycaps, but that’s easy enough to get used to. Keychron does make a nice set of $40 Cherry-style keycaps, though, which I prefer, both in terms of profile and sound. Maybe one day, Keychron will offer that as an option.

As usual, the company offers its keyboard as a bare-bones kit. That will set you back $195 plus shipping. That’s also the only way to get the dark purple colorway. Fully assembled, with switches and keycaps (stabilizers are included in both versions), you pay $215 plus shipping. Personally, I’d pay the extra $20 and get a set of switches and keycaps to try, but if you already know your preferences and have boxes full of keycaps just waiting to be used, then the bare-bones is probably the way to go (just make sure your own set supports Alice layouts, with their unusual double spacebars).

Keychron doesn’t really have a lot of competition in this space right now; 75% Alice-style boards are a bit of a rarity, though there is the Feker Alice 75 from the Epomaker group, with similar specs, but a $329 price tag (which includes keycaps but not switches). One advantage here is that it includes Bluetooth support. There aren’t a lot of other in-stock options available, though you may find some 75% Alice group buys every now and then and maybe KBDfans will bring back its Mountain Ergo at some point (but that was well over $450 for the bare-bones kit).

If you want a smaller board, there are a few more options, including Keychron’s aluminum Q8 (which is my daily driver) and the cheaper, non-gasket-mount V8. And like with all Keychron models, a $100 V10 is on its way, too.

Keychron gets it right with its Q10 Alice-style keyboard by Frederic Lardinois originally published on TechCrunch

Online publishing platform Medium, originally created by Twitter co-founder Evan Williams, announced today it’s embracing the open-source Mastodon platform by creating its own instance to support its authors and their publications. The company said it’s launching me.dm, a Mastodon community that will offer reliable infrastructure, moderation, and a short domain name to make it easier for authors to share their usernames, among other things.

Though launched six years ago, Mastodon has more recently gained traction as users flee Elon Musk’s Twitter. Since acquiring the social network, Musk has made a series of controversial decisions, like re-enabling the accounts of white supremacists and former president Donald Trump and banning journalists, amid a reduction in Twitter’s content moderation teams. He’s also mucking around with Twitter’s product, promising to un-verify users who don’t pay for the service, while alienating advertisers, putting Twitter’s future at risk.

That’s left a subset of Twitter users seeking new places to serve as a place to post their thoughts and engage in public conversations. Mastodon is among those platforms that directly benefitted.

The social microblogging platform has grown to 2.5 million monthly active users in recent days, up from just 300,000 in October 2022. The service isn’t exactly a Twitter clone, however, because of its decentralized nature. Instead of joining “Mastodon” itself, members join communities, or instances, which have their own sets of rules, moderation guidelines, and home timelines. While this doesn’t prevent a user from following another user on a different community, it’s expected that many will choose to join a community that reflects their specific interests, like tech, music, security, gaming, or one of many other topics.

That seems to be the case here with Medium’s Mastodon entry. The company will open up Mastodon to Medium writers and readers as an additional perk of Medium membership, creating a place for discussions around Medium’s content. This will result in an “interesting local feed,” the company says. Medium also notes it will make it easier to find people and topics that match your interests when joining Mastodon, as a part of its onboarding workflow.

As a part of this effort, Medium will create a “sign-up with Medium” interface for joining Mastodon, which would help to address some users’ complaints about the difficulties with getting started on Mastodon, which can be confusing because of the instance selection process.

Though it may seem odd for a longer-form blogging platform to embrace a platform designed for much shorter posts, Medium believes it’s worth operating in both spaces.

“We think the mission of Medium — to deepen people’s understanding of the world by helping to share the best ideas and best information — transcends mediums,” company CEO Tony Stubblebine wrote in a blog post. (Williams exited Medium as CEO last year.)

“To date, we’ve focused on article-length writing. It’s how our company got its name: as a home for medium-length writing on the internet. By contrast, Mastodon is primarily for short-form writing of 500 characters or less… Today we are extending what we do into the short-form medium (lowercase m) with an instance on Mastodon,me.dm,” he said.

Medium’s move to embrace Mastodon isn’t only because it sees the platform as “an emerging force for good in social media,” as Stubblebine said.

In reality, businesses like Medium could suffer if Twitter were to experience a decline in the Musk era. To date, Twitter has served as a way for publishers to promote their work, engage in discussions with their readership and grow their followings. If more writers exit Twitter, platforms like Medium could lose traction, as well, if the writer didn’t have a strong flowing elsewhere on social media.

Newsletter publisher Substack recently addressed this concern with the launch of a discussions feature that enables short chats between writers and readers on its own app. Flipboard also last month launched a Notes feature for a similar purpose. And Tumblr owner Automattic’s CEO Matt Mullenweg said in November that Tumblr would adopt ActivityPub, the decentralized social networking protocol that powers Mastodon.

For Mastodon, however, the launch of a Medium instance could help further boost its numbers as Medium today touts a network of over 100 million readers. Even if only a tiny percentage of those joined Medium’s Mastodon, it could represent a decently-sized influx of new fediverse users.

Medium says it will begin inviting select authors and publications tomorrow to be in an early group of testers before inviting all writers and readers in the near future.

Medium embraces Twitter alternative Mastodon with launch of its own community by Sarah Perez originally published on TechCrunch

Life has been pretty brutal in hospitality over the past couple of years, but that hasn’t stopped Mint House from bubbling to the top with a series of tech-forward apartments.

Under the slogan of “The comfort of home. The luxury of a hotel. Tech-enabled and tailored to you,” the company is trying to make being away from home a bit less crappy. Targeting high-end business and leisure travelers, Mint House recently raised a $35 million Series Bfrom Mohari Hospitality, with participation from Revolution Ventures, Allegion Ventures and Ingleside Investors. In this teardown, we’ll take a closer look at the deck the company used to land its Series B.

We’re looking for more unique pitch decks to tear down, so if you want to submit your own, here’s how you can do that.

Slides in this deck

Cover slide
Summary slide
Business model slide
Consumer product slide
Competitor comparison slide
Key metrics slide
Value proposition for real estate slide
Results slide
Corporate partnerships slide
Team slide
Appendices cover slide
Case Study appendix slide

Three things to love

Raising $35 million in the hotel space at the tail end of a pandemic? Good heavens, I’m curious how they pulled that one off. That is doubly true given that the deck has a Texas-sized red flag on what is universally accepted to be one of the most important slides of any deck (more about that in the “things that could be better” section below.

But hey! That’s what we’re here to examine! Let’s start with the big wins:

Excellent 1-page summary slide

I’ve made no secret of loving a good summary slide — it can really set the tone and pace of a presentation. This cover slide isn’t great to be used in a pitch setting, and for presentations, I would probably recommend simplifying it further, but for a send-ahead deck, this gets investors up to speed very quickly and explains the macroeconomics of what the company is doing and its traction and progress to date, and offers a brief summary of what the company is doing and whom they do it for.

I wish it also mentioned how much money it was raising, either here or on the cover slide — but apart from that, this slide is a beacon of perfection.

As a startup founder, the thing you can learn from this slide is how to get an investor filled in on the context for the company and the round at hyper-speed so you can focus the bulk of your pitch on the things that really matter: What you’re raising, what your plans are and where the market is going. This is particularly important for later rounds — dwelling on the past is important only to prove that you know what you’re doing. Investors are investing in what’s going to happen next, after all.

Front-load the longest pole in the tent

As a startup founder, you can probably predict what the biggest challenges are for the company you are about to pitch. In the case of Mint House, the third slide says something about what the company has been pushing against when it comes to investment. Hotels can be a lucrative business, but the business model is well-trodden ground. The company is promising to innovate on that model, and to investors, that’s both the biggest risk and the biggest opportunity.

In the cutthroat world of hospitality, it’s fantastically hard to stand out, and nonspecialist investors often don’t even look at the vertical. Finding a way of telling the story well and clearly differentiating yourself from the status quo is a crucial part of telling the story. Mint House does that beautifully here, touching essentially every part of the business model on a single slide. Is the slide itself good? Not exactly — the text is so small it’s practically unreadable. But it does the job of summarizing the key differentiators really well.

As a startup, what you can learn here is how to differentiate yourself from a wall of incumbents. If you cannot, well, you essentially don’t have a business at all.

Show me the numbers!

Businesses that are in operating and scale-up mode have metrics. Not showing them in your pitch deck is very silly indeed, and just as important as the numbers themselves are the numbers you select to represent your business to your potential investors.

The fact that this slide is here is great and encouraging — but investors are going to be looking very closely. Operating profit margin, revenue per available room, and occupancy numbers are crucial for the operating side of the business. In other words: If you are in charge of growing and developing Mint House for the next three months, those metrics are crucially important.

However, those aren’t necessarily the metrics investors care about. I’d have expected to see graphs here, including revenue, occupancy over time, and perhaps customer acquisition costs and other metrics that show what’s happening under the hood. Operational efficiency and cash flow might be another set of metrics worth a closer look.

The kind of metrics an investor would care about are directly linked to the fundraise itself. You’re raising $35 million? Great. Show what you’re going to do with the money and what the major milestones are for the next six to 18 months in the form of an operating plan.

There’s a deep orange, if perhaps not red, flag here, too, though: The net promoter score is represented as a percentage, which is incorrect. NPS is an absolute number ranging from -100 to 100, calculated using a formula you can find on every growth-hacking blog that ever existed. The fact that the founders (and everyone who looked at this deck) didn’t catch that — along with the strangely not-quite-right selection of metrics above — makes me wonder about the quality of the overall team.

Where my mind goes here is whether the team is made up of tech startup founders or hotel operators. Nothing wrong with being the latter, but Airbnb is able to run with much better profit margins than someone who is operating a suite of dozens of brick-and-mortar buildings. I’m not saying that this wouldn’t be a good investment, but I question whether it makes sense as a VC investment.

In the rest of this teardown, we’ll take a look at three things Mint House could have improved or done differently, along with its full pitch deck— and that big red flag I mentioned earlier.

Three things that could be improved

As good as Mint House’s pitch deck is, there are a few things that made me scratch my head — some mistakes that are serious enough that I was kinda curious how the company was successful in raising at all, if I’m being honest. Let’s take a look.

Pitch Deck Teardown: Mint House’s $35M Series B deck by Haje Jan Kamps originally published on TechCrunch

The emergence in the last week of a particularly effective voice synthesis machine learning model called VALL-E has prompted a new wave of concern over the possibility of deepfake voices made quick and easy — quickfakes, if you will. But VALL-E is more iterative than breakthrough, and the capabilities aren’t so new as you might think. Whether that means you should be more or less worried is up to you.

Voice replication has been a subject of intense research for years, and the results have been good enough to power plenty of startups, like WellSaid, Papercup, and Respeecher. The latter is even being used to create authorized voice reproductions of actors like James Earl Jones. Yes: from now on Darth Vader will be AI generated.

VALL-E, posted on GitHub by its creators at Microsoft last week, is a “neural codec language model” that uses a different approach to rendering voices than many before it. Its larger training corpus and some new methods allow it to create “high quality personalized speech” using just 3 seconds of audio from a target speaker.

That is to say, all you need is an extremely short clip like the following (all clips from Microsoft’s paper):

https://techcrunch.com/wp-content/uploads/2023/01/in1.wav

https://techcrunch.com/wp-content/uploads/2023/01/in2.wav

To produce a synthetic voice that sounds remarkably similar:

https://techcrunch.com/wp-content/uploads/2023/01/outcome1.wav

https://techcrunch.com/wp-content/uploads/2023/01/outcome2.wav

As you can hear, it maintains tone, timbre, a semblance of accent, and even the “acoustic environment,” for instance a voice compressed into a cell phone call. I didn’t bother labeling them because you can easily tell which of the above is which. It’s quite impressive!

So impressive, in fact, that this particular model seems to have pierced the hide of the research community and “gone mainstream.” As I got a drink at my local last night, the bartender emphatically described the new AI menace of voice synthesis. That’s how I know I misjudged the zeitgeist.

But if you look back a bit, in as early as 2017 all you needed was a minute of voice to produce a fake version convincing enough that it would pass in casual use. And that was far from the only project.

The improvement we’ve seen in image-generating models like DALL-E 2 and Stable Diffusion, or in language ones like ChatGPT, has been a transformative, qualitative one: a year or two ago this level of detailed, convincing AI-generated content was impossible. The worry (and panic) around these models is understandable and justified.

Contrariwise, the improvement offered by VALL-E is quantitative, not qualitative. Bad actors interested in proliferating fake voice content could have done so long ago, just at greater computational cost, not something that is particularly difficult to find these days. State-sponsored actors in particular would have plenty of resources at hand to do the kind of compute jobs necessary to, say, create a fake audio clip of the President saying something damaging on a hot mic.

I chatted with James Betker, an engineer who worked for a while on another text-to-speech system, called Tortoise-TTS.

Betker said that VALL-E is indeed iterative, and like other popular models these days gets its strength from its size.

“It’s a large model, like ChatGPT or Stable Diffusion; it has some inherent understanding of how speech is formed by humans. You can then fine tune Tortoise and other models on specific speakers, and it makes them really, really good. Not ‘kind of sounds like,’ good,” he explained.

When you “fine tune” Stable Diffusion on a particular artist’s work, you’re not retraining the whole enormous model (that takes a lot more power), but you can still vastly improve its capability of replicating that content.

But just because it’s familiar doesn’t mean it should be dismissed, Betker clarified.

“I’m glad it’s getting some traction because i really want people to be talking about this. I actually feel that speech is somewhat sacred, the way our culture thinks about it,” and he actually stopped working on his own model as a result of these concerns. A fake Dali created by DALL-E 2 doesn’t have the same visceral effect for people as hearing something in their own voice, that of a loved one, or of someone admired.

VALL-E moves us one step closer to ubiquity, and although it is not the type of model you run on your phone or home computer, that isn’t too far off, Betker speculated. A few years, perhaps, to run something like it yourself; as an example, he sent this clip he’d generated on his own PC using Tortoise-TTS of Samuel L. Jackson, based on audiobook readings of his:

https://techcrunch.com/wp-content/uploads/2023/01/samuel_jackson.mp3

Good, right? And a few years ago you might have been able to accomplish something similar, albeit with greater effort.

This is all just to say that while VALL-E and the 3-second quickfake are definitely notable, they’re a single step on a long road researchers have been walking for over a decade.

The threat has existed for years and if anyone cared to replicate your voice, they could easily have done so long ago. That doesn’t make it any less disturbing to think about, and there’s nothing wrong with being creeped out by it. I am too!

But the benefits to malicious actors are dubious. Petty scams that use a passable quickfake based on a wrong number call, for instance, are already super easy because security practices at many companies are already lax. Identity theft doesn’t need to rely on voice replication because there are so many easier paths to money and access.

Meanwhile the benefits are potentially huge — think about people who lose the ability to speak due to an illness or accident. These things happen quickly enough that they don’t have time to record an hour of speech to train a model on (not that this capability is widely available, though it could have been years ago). But with something like VALL-E, all you’d need is a couple clips off someone’s phone of them making a toast at dinner or talking with a friend.

There’s always opportunity for scams and impersonation and all that — although more people are parted with their money and identities via far more prosaic ways, like a simple phone or phishing scam. The potential for this technology is huge, but we should also listen to our collective gut, saying there’s something dangerous here. Just don’t panic — yet.

VALL-E’s quickie voice deepfakes should worry you, if you weren’t worried already by Devin Coldewey originally published on TechCrunch

Field Programmable Gate Arrays (FPGA), or integrated circuits sold off-the-shelf, are a hot topic in tech. Because they’re relatively affordable and can be programmed for a range of use cases, they’ve caught on particularly in the AI and machine learning space, where they’ve been used to accelerate the training of AI systems.

The global FPGA market size could reach $14 billion by 2028, according to one estimate, up from $6 billion in 2021.

One startup looking to get in on the ground floor is Rapid Silicon, which this week announced that it raised $15 million in a Series A round led by Cambium Capital. Launched in 2021, the goal with Rapid Silicon is to promote, adopt and implement open source tech to address the low- to mid-range FPGA market, according to CEO and co-founder Naveed Sherwani.

“Rapid Silicon’s … software leads the programmable revolution as the industry’s first and only commercial open source FPGA design suite,” Sherwani, previously a GM at Intel and the former CEO of semiconductor startup SiFive, said in an email interview. “The latest round of funding will be used to further invest in Rapid Silicon’s product portfolio, support the launch of its premier low-end FPGA product … and to build on the company’s momentum in leading the adoption of open source software for commercial applications.”

Rapid Silicon is developing two products at present: Raptor and Gemini. Raptor is electronic design automation software with an interface for FPGA application design, while Gemini is a 16-nanometer FPGA with hardware including a dual-core Arm processor, external memory controller and ethernet connectivity. Sherwani emphasized that Rapid is based on open source software — another industry first, according to him — and designed to meet the needs of FPGA developers tackling challenges in sectors such as healthcare, automotive and industry.

“Customers are looking for innovative ways to program FPGAs, reduce support load by leveraging the open source ecosystem of active expertise and development engineers, and shorten time-to-market,” Sherwani added. “With open source software, Rapid Silicon is removing the barriers and providing its customers with a robust end-to-end FPGA design workflow. The open source software enables users to design complex applications quickly and efficiently on our FPGA devices.”

Gemini isn’t commercially available, but Sherwani says he expects the FPGA will come to market by the end of Q1. In the meantime, Rapid Silicon is generating revenue — between $2 million and $3 million a year — from licensing its IP.

The FPGA space has formidable competitors including Intel, which several years ago acquired U.K.-based Omnitek and Altera to double down on FPGA-based solutions for video and AI applications. But Landon Downs, the managing partner at Cambium Capital, said that he sees “immense” potential in Rapid Silicon’s tooling and hardware strategy. While that might sound like hyperbole coming from a VC, Rapid Silicon evidently has investors intrigued; the company expects to close a $15 million extension of its Series A within the next few months at an $80 million pre-money valuation.

“Driven by its purpose and world-class talent, we believe Rapid Silicon is ready to revolutionize design-to-silicon turnaround time and provide solutions that meet and exceed the robust performance, power, area and time-to-market requirements for next-generation applications,” he said in a press release. “We see immense potential in the company’s AI-enhanced EDA tools, and we believe this team has the experience needed to bring these solutions to the global market.”

FPGA startup Rapid Silicon lands $15M to bring its first chip to market by Kyle Wiggers originally published on TechCrunch

This is my week of debriefs. CES has a way of hurling you into the new year, kicking and screaming, and it can be hard to find your bearings as you emerge on the other side. As the dust has cleared, one thing remains very clear: No two people have the same notion of what does and doesn’t constitute a robot.

That’s not a problem, per se. Language evolves and so does technology. I was on the Equity Podcast this week to talk CES with Haje, one of our hardware reporters. He posited that much — or even most — technology is essentially a robot at this point. One can fairly credibly make the argument that “robot” as a term is much broader than how we tend to deploy it. Certainly, machine learning is becoming pervasive in ways few imagined.

The upside of casting as wide a net as possible is that one can also make the case that ubiquitous robotics isn’t some vision of the future. It’s very much already here, and maybe that’s heartening. A stricter definition — and one I’ve tended to prefer — involves some level of autonomy and perception. You know, taking in information from external sources and making decisions accordingly, much like we do.

I’m not wedded to this definition to the point that it precludes me from covering other stories in the space that don’t precisely line up. Nor is everything that qualifies necessarily in my purview. For one thing, we have a couple of automotive reporters who are very good at their jobs, so they take first crack at all of the self-driving car stuff. Truth of the matter is, our definition of what does and doesn’t qualify as a robot is also governed by some editorial decision making. It can also be porous.

This could easily become a newsletter about the eight billion lidar startups out there, but I don’t really want that, and I suspect most of you don’t, either. For me, at least, there’s not that much value in strict adherence to pedantry or orthodoxy when I determine what does and doesn’t make sense on these pages. But it is helpful to have some guardrails.

Seeing a “smart” washing machine in this newsletter might be novel the first time, but you would (rightfully) get annoyed with me if I suddenly started forcing them on you every week. Robotic vacuums, on the other hand, do tick off the robot requirements for many — or most — people, myself include. I absolutely cover them here, but I’ve also been doing this long enough to know that covering every single robot vacuum on the market is a good way to hemorrhage readers.

All of this stuff bleeds together at a show like CES. I spent time with Robosen’s extremely neat Transformer robots, for example. But I think it does a disservice to both parties if we attempt to compare those toys to, say, some industrial fulfillment robot exhibiting at the show. I — somewhat cheekily — suggested that robots are “cool technology used for uncool things.” This is obviously not a guiding principle, but it does point to something worth discussing here.

Weird and wacky robot toys are going to grab the headlines at a show like CES. We get that. They’re good for traffic and they’re fun. I’ve written about plenty of them and will likely continue to do so in the future when genuinely interesting ones surface. Another reason people are drawn to them is that they more closely resemble a kind of platonic robot ideal. Robot toys look how we think a robot should look. Whatever you ultimately think about science fiction’s impact on public perception, it’s important to realize that it will never go away.

As robots become a bigger fixture in our daily lives, however, the impact will increasingly be a dialogue. I paid some ungodly sum to see the new “Avatar” in the theater with all of the trappings. One thing that struck me — and likely you, too — was how much the U.S. military robots on Pandora resemble robots that exist in the world today. Want to annoy your date? Point out where each robot in a sci-fi movie gets its inspiration.

Predictably, the robotics writing I’ve done over the past week has largely revolved around CES because, frankly, most everything I’ve done over the past week has revolved around CES. No one ever said the life of a hardware editor would be an easy one, friends.

But two and a half newsletters is more than ample coverage for the show, so let’s round up a couple of the top non-CES stories, shall we?

Another company graduated the Alphabet X “moonshot” labs this week. Two years after exiting stealth, Mineral is now its very own Alphabet company. The robot aspect largely revolves around data collection here, monitoring crops to give farmers deep and rich actionable information for growing more sustainably and efficiently.

“After five years incubating our technology at X, Alphabet’s moonshot factory, Mineral is now an Alphabet company,” said CEO Elliott Grant. “Our mission is to help scale sustainable agriculture. We’re doing this by developing a platform and tools that help gather, organize and understand never-before known or understood information about the plant world — and make it useful and actionable.”

File this one under: not the best look. Home robots and privacy are coming to an inevitable head. MIT Review rounded up images from Roomba that eventually made their way onto social media; iRobot responded that the shots were all taken with user consent. This is an increasingly important discussion as we bring robots and camera-sporting tech into our homes. And it will likely be pointed to as a compelling argument against Amazon’s acquisition of the firm.

One user called it “a clear breach of the agreement on their side […and] also a violation of trust.” The big question that immediately springs to mind here is what is the reasonable expectation of privacy when inviting this sort of technology into your home?

That’s it for this week. Bit of a short one, but I’m gonna do my best to sleep through the weekend and be back with you this time next week. Meantime, please subscribe to Actuator here, if you haven’t already.

Robot or fauxbot? by Brian Heater originally published on TechCrunch

Learning navigation platform Career Karma has laid off another 22 people across its global and domestic workforce, less than five months after it cut 60 staff members, according to sources. CEO and co-founder Ruben Harris confirmed the workforce reduction to TechCrunch.

The cut shows that even as many edtech companies attempt to right-size their staff, there’s more work to do. Harris’ e-mail to remaining staff underscores the tension of today: once eager enterprise customers are still making up their minds on whether or not to sign up for new tools, leading to extended sales cycles and uncertainty.

“Last year, we made the decision to right size the company so that we can orient Career Karma towards working with employers and now that we have started to sign customers it’s clear that we made the right decision,” Harris wrote in the email. “What’s unclear is how Fortune 1000 companies will be responding to the macroeconomic environment and it’s important for us to give ourselves time to work with them to figure that out.” As the market evolves, Career Karma’s service of matching employees or professionals to tech bootcamps is being put in a difficult place. Just last month, BloomTech, a coding bootcamp previously known as Lambda School, cut half of staffin pursuit of profitability.

During Career Karma’s last cut, Harris emphasized that the layoff and its previously-closed $40 million Series B would extend the startup’s runway to three years. After laying off staff this week, Career Karma now has five years of runway.

As TechCrunch has discussed in the past, the strategy of “extending your runway” always comes into vogue whenever investors slow down investing. Career Karma’s shift from the basic three-year rule of thumb to five years shows how that rule may become even more conservative as the downturn continues. Over email, Harris tells TechCrunch that he “always [wants] to have the option to raise, I just don’t want to be forced to raise.”

With 80 staff now remaining at Career Karma, Harris confirmed that no C-suite executives were impacted by the layoff. Those impacted were offered two months of severance as well as extended benefits. The career navigation platform also, fittingly, offered career navigation support to its new alumni.

Current and former Career Karma employees can reach out to Natasha Mascarenhas on Signal, a secure encrypted messaging app, at 925 271 0912. You can also DM her on Twitter, @nmasc_.

Career Karma’s latest layoff underscores edtech’s new challenge by Natasha Mascarenhas originally published on TechCrunch

Actually, there is one industry that could use web3, and that industry is fashion.

Hear us out. Fashion is one of the most polluting sectors in the world, and, according to the United Nations Environmental Programme, is responsible for up to 10% of the world’s carbon dioxide output, more than the international flights and maritime shipping industries combined. Eighty-five percent of clothes in the United States alone end up in landfills, and at least 20% of all water pollution results from textile dyeing. The ravenous appetite of fast fashion shoppers isn’t settling anytime soon, and fashion’s supply chain remains quite arduous on the environment.

A possible step toward finding the multiple solutions needed to fix this damaging sector is, well, embracing more web3.

Web3 could help fashion become more sustainable by Dominic-Madori Davis originally published on TechCrunch

TikTok is the latest tech giant to be schooled by France’s data protection watchdog for breaking rules on cookie consent.

The €5 million penalty announced today by the CNIL relates to a cookie consent flow TikTok had used on its website (tiktok.com) until early last year — in which the regulator found it was not as easy for users to refuse cookies as to accept them — so it was essentially manipulating consent by making it easier for site visitors to accept its tracking than to opt out.

This was the case when the watchdog checked in on TikTok’s process, in June 2021, until the implementation of a “Refuse all” button on the site in February 2022 — which appears to have resolved the matter. (And may explain the relatively small fine levied in this case, along with the number of users and minors affected — as well as the enforcement relating only to its website, not its mobile app.)

Tracking cookies are typically used to serve behavioral advertising but can also be used for other site activity, such as analytics.

“During the check carried out in June 2021, the CNIL noted that while the companies TikTok United Kingdom and TikTok Ireland did offer a button allowing cookies to be accepted immediately, they did not put in place an equivalent solution (button or other) to allow the Internet user to refuse their deposit just as easily. Several clicks were necessary to refuse all cookies, against only one to accept them,” the watchdog notes in a press release [translated from French with machine translation].

“The Restricted Committee considered that making the refusal mechanism more complex actually amounts to discouraging users from refusing cookies and encouraging them to favor the ease of the “Accept all” button,” it added, saying it found TikTok had therefore breached a legal requirement for freedom of consent — a violation of Article 82 of the French Data Protection Act “since it was not as simple to refuse cookies as to accept them”.

In addition, the CNIL found that TikTok had not informed users “in a sufficiently precise manner” of the purposes of the cookies — both on the information banner presented at the first level of the cookie consent and within the framework of the “choice interface” that was accessible after clicking on a link presented in the banner. Hence finding several breaches of Article 82.

The French enforcement has been taken under the European Union’s ePrivacy Directive — which, unlike the EU’s General Data Protection Regulation (GDPR), does not require complaints that affect users across the bloc to be referred back to a lead data supervisor in an EU country of main establishment (if a company claims that status — as TikTok does with Ireland for the GDPR).

This has enabled the French regulator to issue a series of enforcements over Big Tech cookie infringements in recent years — hitting the likes of Amazon, Google, Facebook and Microsoft with some hefty fines (and correction orders) since 2020, following a 2019 update to its guidance on the ePrivacy Directive which stipulated that consent is necessary for ad tracking.

France’s activity to clean up cookie consent looks like an important adjunct to slower paced cross-border GDPR enforcement — which is only just starting to have an impact on ad-based business models centred on consent-less tracking, such as the final decisions against Facebook and Instagram issued by the Irish Data Protection Commission earlier this month.

If tracking-and-profiling ad giants are forced to rely on gaining user consent to run behavioral advertising it’s critical that the quality of consent gathered is free and fair — not manipulated by deploying deceptive design tricks, as has typically been the case — so the CNIL’s ePrivacy cookie enforcements look important.

Only last summer, for instance, TikTok was prevented from switching away from relying on user consent as its legal basis for processing people’s data to run ‘personalized’ ads to a claim of legitimate interest as the legal basis (implying it intended to stop asking users for their consent) after intervention by EU data protection authorities who warned it such a move would be incompatible with the ePrivacy Directive (and likely breach the GDPR too).

While enforcements under ePrivacy only apply in the regulator’s own market (France, in this case), the impact of these decisions may be wider. Google, for example, followed a sanction from the CNIL by revising how it gathers consent to cookies across the EU. That may not be how every company responds but there is a likely to be a cost associated to applying different compliance configurations for different EU markets — vs just applying one (high) standard in all EU markets. So ePrivacy enforcement may help set the EU bar.

TikTok was contacted for comment on the CNIL’s sanction. A spokesperson for the company sent us this statement:

These findings relate to past practices that we addressed last year, including making it easier to reject non-essential cookies and providing additional information about the purposes of certain cookies. The CNIL itself highlighted our cooperation during the course of the investigation and user privacy remains a top priority for TikTok.

TikTok fined in France for manipulative cookie consent flow by Natasha Lomas originally published on TechCrunch