A Seattle-area startup called Pauling.AI is harnessing artificial intelligence to automate the early steps that lead to the discovery of new drugs. The technology can complete tasks in a matter of weeks that previously required three to six months, said founder and CEO Javier Tordable.

Using AI to accelerate research timelines could ultimately spark an exponential increase in new treatments, proponents say.

“The dream of a lot of people in the field would be that, at some point, we’ll go from 30 or 40 new drugs approved every year to 300 or 400,” Tordable said, “and cure all sorts of diseases.”

Tordable launched his company in 2024 after a 16-year tenure at Google, most recently as the technical director of the company’s healthcare and life sciences initiatives. While he doesn’t have expertise in biology or chemistry, Tordable said he’s skilled at building tech tools that can perform complex tasks — such as those required to create new pharmaceuticals.

The startup operates on a “scientist-as-a-service” model, allowing researchers to outsource early steps in the drug discovery process to AI. The platform performs computational chemistry work, engineering drug candidates and modeling how they might interact with molecules and inhibitors within a cell.

The result is a curated list of small-molecule compounds that scientists can then move into a physical laboratory for testing as therapeutics. In the future, the startup would like to produce more complex compounds as drug candidates, such as antibodies.

To accomplish all of this, Pauling is building automation tools that engage with existing large language models and databases from numerous sources.

The startup has six employees who work remotely. Its leadership includes Chief Scientific Officer Oleksandr Savytskyi, a computational biologist who worked in academia in Ukraine and did research at the Mayo Clinic.

Pauling has secured an undisclosed amount of pre-seed funding from Flex Capital and angel investors. It currently serves less than a dozen customers, including several high-profile academic institutions, Tordable said.

The company joins a burgeoning field of AI-biotech ventures, with numerous Pacific Northwest startups: Variational AI in Vancouver, B.C.; Seattle-based Potato and Synthesize Bio; and Xaira Therapeutics, which is based in San Francisco and has labs in Seattle. Additionally, FutureHouse is a California nonprofit in this sphere.

Ultimately, Tordable hopes that by shrinking the time and cost of drug development, it will become economically feasible to tackle rare diseases that are typically not served by big pharma, providing overlooked patients with treatments and cures.

“The nice thing of working in this field is that we’re not necessarily doing it just for economic returns,” Tordable said. “There’s also an enormous benefit to humanity.”

Michael Lucarelli is looking for Seattle food recommendations — after relocating to the city earlier this year and moving his company with him.

RentSpree, which got its start in Los Angeles but is now headquartered in downtown Seattle, has built a profitable business helping landlords and real estate agents screen tenants, collect rent, sign leases, and manage rentals online.

The company, founded in 2016, serves more than 4 million users and is growing without relying heavily on paid advertising, said Lucarelli, CEO and co-founder of RentSpree.

Lucarelli, a former real estate agent, said Seattle stood out because of its concentration of real estate and proptech companies such as Zillow, Redfin, and Opendoor. The company this month hired former Redfin exec Alex Berezhnyy as chief technology officer, further anchoring its presence in the region, where more than half of the executive team is now based. It has more than 30 employees in Seattle.

“Seattle is really great for talent that balances both an aggressive growth perspective, but also building sustainable companies over time,” Lucarelli said.

RentSpree targets “DIY” landlords, typically individuals who own one to four rental units and still rely on paper applications and manual rent collection. The company’s software helps them manage the entire rental process online, from applications and leases to monthly payments.

While landlords and real estate agents are RentSpree’s core users, the company makes most of its money from renters, who pay application fees and small convenience fees for rent payments. That model has helped fuel its payments business, which is now growing about 150% year-over-year and processing hundreds of millions of dollars annually.

RentSpree also recently launched a banking-as-a-service offering that lets landlords open bank accounts through the platform, earn interest, and track expenses — pushing the company further into fintech territory.

The company’s real advantage is it’s distribution, Lucarelli said. Instead of relying on digital ads, RentSpree partners with MLS systems, Realtor associations, and real estate software platforms to reach landlords where they already work.

More than 10,000 landlords and agents use RentSpree each month. The company has rolled out new AI features to help streamline filling out forms and listing properties.

“We’re focusing on the important jobs that they’re trying to accomplish, or things that they’re doing already — and how we can make it vastly easier by utilizing AI for them,” Lucarelli said.

The company has raised $28 million to date and employs 135 people across the company in the U.S. and Thailand.

In a remarkable pivot, Athira Pharma is shifting its primary focus from Alzheimer’s to oncology following a multi-year period marked by clinical failures and leadership turnover.

The Bothell, Wash.-based company announced today that it has licensed a Phase 3 breast cancer drug from Sermonix Pharmaceuticals, supported by $90 million in funding from a group of healthcare investment firms. It could land an additional $146 million if the research yields promising results.

Athira will simultaneously continue researching ATH-1105, its own drug candidate for treating ALS.

Athira President and CEO Mark Litton called the development “exciting and transformative news.”

“By securing rights to this late-stage program — while also advancing pATH-1105 for ALS — we are building a pipeline that we believe has the potential to change lives and create enduring value,” Litton said on LinkedIn. “We are honored to have the backing of some of the most respected biotechnology funds in the industry.”

Following the deal announcement, the company’s stock rose 70% to $7 per share.

Athira has been through a tumultuous few years:

• Its Alzheimer’s disease drug candidate, called fosgonimeton, stumbled and then failed a Phase 2/3 trial last year.
• That sank Athira’s stock price and triggered a layoff of 49 employees, or about 70% of its workforce in September 2024.
• In 2021, Athira CEO and President Leen Kawas resigned after it was confirmed she had altered images in scientific papers from her graduate studies that helped form the company’s foundation.

Details on the deal

Under today’s deal, Athira secures an exclusive license to develop and commercialize the breast cancer therapeutic for nations outside of Asia and select Middle Eastern countries. The drug, called lasofoxifene, is currently in a clinical trial that has enrolled over half of its target patient population, with initial results expected in mid-2027.

The agreement provides Sermonix with 5.5 million shares of Athira’s stock. The Seattle company has also committed to paying Sermonix up to $100 million plus limited royalties if certain commercial targets are reached.

Three lead investors are backing the research: New York’s Commodore Capital, biotech hedge fund Perceptive Advisors, and California-based TCGX. Additional participants are ADAR1, Blackstone Multi-Asset Investing, Kalehua Capital, Ligand Pharmaceuticals, New Enterprise Associates (NEA), Spruce Street Capital and 9vc.

Athira raised $90 million by selling stock and warrants to investors, extending its cash runway into 2028. If investors choose to exercise those warrants in the future, the company could receive up to an additional $146 million to fund its clinical programs.

“We’re proud to support this evolution and excited by the opportunity to deliver meaningful impact for patients and shareholders alike,” said Joseph Edelman, founder and CEO of Perceptive Advisors, in a statement.

Regarding its own ALS drug candidate, Athira this year successfully completed Phase 1 safety trials, and plans to start Phase 2 trials early next year.

Tin Can is answering the call from investors.

The Seattle startup behind a landline-style, Wi-Fi-enabled telephone for kids raised $12 million in new funding, the company announced Thursday.

The seed round was led by Greylock Partners with participation from Lateralus Holdings and existing backers. Tin Can previously raised $3.5 million in pre-seed funding in September from PSL Ventures, Newfund Capital, Mother Ventures, and Solid Foundations.

Tin Can’s colorful screen- and text-free phones aim to help kids connect without the pressures and addictive pull of the digital world. The devices operate on a private network and include a companion app and modern safeguards.

Since launching its flagship product earlier this year, Tin Can quickly went “viral,” sold out its first two production runs and built a near-six-figure waitlist. The momentum comes amid growing concern about the effects of smartphones and social media on kids’ mental health, attention and development.

Co-founder and CEO Chet Kittleson told GeekWire that his team is “pretty elated” to attract investors who are also parents, who care about what Tin Can is building and want to help the company keep pace with demand.

“They really care about us and about the customer and about the world that we want to see, and they’re also really technically proficient,” Kittleson said about David Shuman, founder of Lateralus Holdings, and Mike Duboe, general partner at Greylock.

“Mike is like a growth machine. He was head of growth at Stitch Fix for a long time,” Kittleson said. “And David just knows everything you could ever want to know about supply chain and manufacturing, and cash flow. He’s already helped us a million different ways.”

Duboe said in a news release that Tin Can isn’t just building a product; they’re leading a movement.

“In an age defined by digital noise, they’ve created a joyful alternative that redefines how we view modern connection,” he said. 

Kittleson, along with co-founders Graeme Davies and Max Blumen, previously worked at Seattle real estate startup Far Homes. He was recognized this month as one of GeekWire’s six “Uncommon Thinkers,” honoring innovators driving positive change in the world.

“I’m so grateful that this is the hit, that it worked,” Kittleson said in a GeekWire profile tied to the honor.

Tin Can employs 17 people and is “growing at a pretty fun clip,” Kittleson said.

The startup plans to use the new funding to scale production, add engineers and customer support, and prepare for international expansion.

Coming up: Tin Can CEO Chet Kittleson will be our guest this weekend on the GeekWire Podcast. Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.

Amazon is quietly rolling out the last big pillar of its AI-powered Alexa+ vision: the Alexa.com website, bridging the gap between its Echo devices, mobile app, and consumer desktops.

The web portal means users can now interact with Alexa through a keyboard and mouse: accessing and continuing past Alexa chats, starting new ones, going back and forth between voice conversations in the living room and typed chats in the home office, etc. 

(Alexa.com is available initially to a subset users in the Alexa+ early access program, with access likely to expand in the coming weeks, so if you’re not seeing it yet, stay tuned.)

I’ve been trying it out, and I’m already finding it quite useful as an extension of the Alexa experience. In addition to expanding the chat functionality to the browser, the web interface offers fine-grained control over reminders, calendar appointments, uploaded files, and smart home devices. 

For example, I was able to edit a family reminder: changing the assigned person, adjusting the date and time, setting it to repeat weekly, and adding a flag for Alexa to follow up until it’s complete. All of this happened through simple clicks, much easier than talking Alexa through the details, in my experience.

The rollout of Alexa+ earlier this year also introduced the ability to email or upload documents to Alexa for summarization and reference. This made Alexa a lot more useful on its own. Now, with the online portal, it’s much easier to upload, access and delete files.

There’s also some nice smart home integration, with the ability to control lights and plugs, for example. It’s similar to the Alexa app, but nice to be able to access on the computer. 

In short, it’s a level of point-and-click precision that voice commands and the mobile app can’t offer. Within a few minutes of using Alexa.com, I had this sense of liberation, being able to interact with Alexa in the same way as anything else on the computer. What a concept! 

That said, I couldn’t help but wonder how much I’ll actually use it.

Three years after the launch of ChatGPT, my AI routines have become relatively entrenched. I’m having a hard time envisioning going to Alexa.com on my computer to start a chat rather than Gemini, NotebookLM, Claude, Perplexity or other AI tools that have become daily habits. 

Then again, for me, those tools are about work and individual tasks. Alexa is really the digital hub for my family. The introduction of better AI with Alexa+ has improved that experience over the past few months, and the web portal adds a whole new dimension. Family is why I’ll use it.

The Alexa+ integration goes even further for me since I’ve been talking to Alexa more and more via my Amazon Echo Buds on my phone when I’m out and about, although this is probably more illustrative of me being an edge case than anything else.

Within Amazon, Alexa+ sits at the top of the AI stack that Amazon CEO Andy Jassy talks about. It’s the consumer-facing layer for a company that has made a much bigger mark in AI with its cloud infrastructure and platforms. Alexa’s web launch fills a gap that’s been glaring for a while.

One thing I’d like to see is true Ring integration into the web experience — the ability to see and access smart home cameras from Alexa.com in addition to the existing Ring.com interface. 

The more Amazon can unify everything inside the web portal, the more useful it will be. No doubt I’ll come up with other feature requests as I continue to use it.

But for now, the simple act of using an AI-powered Alexa in a web browser is so mind-blowing, in such a basic way, that it’s hard not to wonder how much further along Amazon would be in the world of consumer AI if it had been able to make this happen a long time ago. 

What happens if you let teens craft the rules that dictate their use of phones at school? You get policy ideas with a nuanced, holistic perspective that rival those being officially issued by the adults in leadership.

The University of Washington’s Youth Advisory Board, a group of approximately 20 teens from Seattle-area schools, recently published its first memo tackling this contentious issue. The memo weighs the pros and cons of phone bans and offers recommendations on how schools should draft and communicate their policies.

“The whole point of the memo was to bring teen experiences into real policy conversations,” said Jaden Hong, a sophomore at Eastlake High School and board participant. “I think it matters that our ideas get into the hands of the principals, district leaders and even state-level decision makers or legislators who are actively shaping phone and tech rules.”

The Youth Advisory Board’s memo was informed by a UW study and questionnaires on the impacts of phone rules at middle and high schools in Washington. The regulations ranged from all-day bans to restrictions during lunch and passing periods. The board’s key suggestions for high school policies include:

• Compromise: Preferred policies allow phone use during breaks between classes and lunch, but not during academic time, as opposed to all-day bans.
• Reframing: Use neutral language around the policy, avoiding polarizing terms like “ban” or “phone free.”
• Inclusion/communication: Input is needed from students, parents and teachers, and should include polls and classroom discussions to get buy-in. Clearly communicate the policies.
• Consistency: Make the rules school-wide and don’t vary them by teacher or class.
• Diverse needs: Students with responsibilities outside of school (like some jobs) or with medical needs require leniency.
• Social engagement: Educators need to foster social engagement during class lessons as well as structured social activities outside of academics.
• Digital wellness: Beyond tech literacy, teens welcome classes on digital wellness and the healthy use of devices.

What the research showed

Lucía Magis-Weinberg, a developmental psychologist and head of the International Adolescent Connection and Technology Laboratory at the UW, conducted the surveys that helped inform the students’ opinions. Roughly 4,400 students, teachers and parents responded to the initial inquiry.

In the answers to questionnaires, teachers emphasized that with limited phone access, there are fewer distractions in the classroom, more social engagement and less bullying. Teens said the restrictions reduced the amount of cheating.

On the downside, teens and parents were concerned that communications were more difficult, such as friends making plans, scheduling with family, or in the case of an emergency. Teens and teachers noted that phones had positive instructional uses and could aid students with specific academic or language challenges.

“As a student, sometimes it’s hard to look outside of yourself,” said Abbie Huang, a board participant who also attends Eastlake. She said that reading teachers’ comments on student engagement and realizing that a lot of students are OK with phone restrictions broadened her opinion.

“It was really cool to see other schools and the way they approached it, and just other people’s perspectives that I didn’t think about before,” she added.

Current policy landscape

The Washington Office of Superintendent of Public Instruction allows local districts to set their own phone policies. The office reported that 75% of the state’s districts were implementing restrictions — either banning phones during class time or throughout the school day.

Oregon, by contrast, took a statewide approach, prohibiting phone use during school hours in the state’s K-12 public schools.

Seattle Public Schools has not issued a district-wide policy, though at least three public middle schools in the district have banned phones at school, and at least one high school prohibits their use during classes.

UW researchers shared the Youth Advisory Board’s memo at last week’s Washington Educational Research Association conference in Tacoma.

Broader tech concerns: AI and social media

Board participants agreed that student input is equally crucial for other pressing tech issues, including rising teen use of artificial intelligence and chatbots, as well as ongoing concerns about social media’s impact on young people.

“I really want to highlight how important it is to get the youth voice in there,” said Rotem Landesman, a UW graduate student in the Information School helping lead the Youth Advisory Board. Teens need to be represented in drafting policies and guidelines, she added, as tech is being integrated into schools “at such a rapid pace.”

Recent data from the Pew Research Center highlights the challenge:

• Some 64% of U.S. teens report having used an AI chatbot, and 31% do so daily.
• The vast majority of teens are engaging with social media, with 92% using YouTube and 68% on TikTok.

For both AI and social media, experts worry about mental health harms, misinformation, privacy and other concerns — while regulating the technology’s use remains difficult.

Sirjana Kaur, a senior at Redmond High School and board participant, said that her AP literature course forbids the use of AI due to concerns about cheating, requiring students to do all of their writing longhand and in class. The year-end AP test, which potentially provides students with college credits, will be done on a computer.

“There’s definitely a lot of work” to be done around AI regulations, she said. “I think there’s a balance that needs to be struck between avoiding AI, but also not making things even harder for students.”

Jeff Bezos’ Blue Origin space venture has delayed the first flight of a wheelchair user into space due to a technical issue that came to light just before today’s scheduled launch.

The countdown for Blue Origin’s suborbital New Shepard mission was terminated nearly an hour after the launch window opened at 10 a.m. CT (8 a.m. PT) at Blue Origin’s Launch Site One in West Texas. Blue Origin didn’t specify the nature of the glitch. Launch commentator Tabitha Lukin said only that the flight team “observed an issue with our built-in checks prior to flight.”

“We are assessing our next opportunity for launch,” she said. Later in the day, Blue Origin said the next launch attempt would come no earlier than Saturday.

This will be the 37th New Shepard mission, and the 16th to carry humans on a brief ride above the 100-kilometer (62-mile) altitude level that marks the internationally accepted boundary of space. Eighty people have previously flown on New Shepard, including Bezos. Six have gone multiple times.

The six-person crew for the mission, known as NS-37, includes Michaela “Michi” Benthaus, a German-born aerospace and mechatronics engineer at the European Space Agency who sustained a spinal cord injury in a mountain biking accident in 2018.

In a 2023 interview published by the Technical University of Munich, Benthaus said that she’s always been “mega fascinated” by space, and set her mind on becoming an astronaut when she was 10 years old.

When Benthaus lost the use of her legs, she initially thought her flight into space “was never going to happen.” But in 2022, her hopes got a big boost when she experienced a zero-G flight arranged through AstroAccess, a project that’s dedicated to paving the way for spacefliers with disabilities. Last year, she was the commander of an analog space mission conducted at the Lunares Research Station in Poland.

Now the 33-year-old is on the verge of blazing a new trail for space access.

“This feels like an important step, since space travel for people with disabilities is still in its very early days,” Benthaus said in a LinkedIn post. “I’m so thankful and hope it inspires a change in mindset across the space industry, creating more opportunities for people like me. I might be the first — but have no intention of being the last.”

Blue Origin has been working for several years to improve accessibility at its New Shepard facilities — for example, by adding an elevator to the seven-story launch tower. A business resource group named New Hawking, in honor of the late wheelchair-using physicist Stephen Hawking, has helped lead the way.

During their flight, Benthaus and the rest of the NS-37 crew will experience a few minutes of zero-gravity and views of a curving Earth against the blackness of space. At the end of the ride, the booster will make an autonomous landing while the crew capsule descends to a parachute-assisted touchdown in the West Texas desert.

Benthaus’ crewmates include:

• Joey Hyde, a physicist and quantitative investor who recently retired from his career at Citadel, a leading hedge fund. He lives in Florida with his wife and five children.
• Hans Koenigsmann, a German-American aerospace engineer whose career has been dedicated to advancing reusable spacecraft and launch vehicles, most notably as an early team member at SpaceX.
• Neal Milch, a business executive and entrepreneur who launched his career through Laundrylux, a family-owned business. He now serves as the chair of the Board of Trustees at the Jackson Laboratory, a nonprofit biomedical research institute.
• Adonis Pouroulis, an entrepreneur, investor and mining engineer with more than 30 years of experience in the natural resources and energy sector. He is the founder and chairman of Pella Resources, co-founder of Energy Revolution Ventures, chairman of Rainbow Rare Earths, and the CEO of Chariot Limited.
• Jason Stansell, a computer scientist and a self-proclaimed space nerd rooted in West Texas. He’s been watching from a front-row seat as the space industry has expanded to offer opportunities for commercial spaceflight. 

In addition to the crew, NS-37 will be carrying more than 20,000 postcards submitted by students and others through a program organized by the Club for the Future, Blue Origin’s nonprofit educational foundation. The featured partners for this latest batch of postcards are UNIQLO, Arm & Hammer Baking Soda Rocket Day and Give Kids the World Village.

Blue Origin typically doesn’t reveal how much people pay to take trips on New Shepard. In some cases, crew members have flown as invited guests. On the other end of the spectrum, crypto entrepreneur Justin Sun paid $28 million for a ticket in a widely publicized auction.

The suborbital space program’s previous milestones include flying the world’s oldest spaceflier (former test pilot Ed Dwight, who was 90 when he rode on New Shepard last year); the world’s youngest spaceflier (Oliver Daemen, who was 18 when he flew with Bezos and two others in 2021). and the first married couple to reach the final frontier together on a commercial spaceship (Marc and Sharon Hagle in 2022).

In a Chapter 11 bankruptcy petition filed this week by Rad Power Bikes, the Seattle-based electric bike maker lists creditors holding the 20 largest unsecured claims against the company.

At the top of the list? Not a major supplier, or partner, but U.S. Customs and Border Protection, which is owed more than $8.3 million by Rad for tariffs, according to the filing. The claim is one of several listed as “disputed” by the company.

The situation underscores the financial strain facing Rad and the broader e-bike industry after rapid growth during the COVID-19 pandemic gave way to slowing demand, rising costs and lingering trade pressures.

A Rad spokesperson said Wednesday that the company is not able to comment on specific line items in its filing. In a November letter to employees warning that the company could shut down as early as January, Rad cited “significant financial challenges, including in the form of tariffs and the macroeconomic landscape.”

Tariffs have drawn increasing scrutiny from the e-bike industry. A recent report by The Washington Post, examined how import duties under both the Biden and Trump administrations sent expenses spiraling for Rad and other bike companies that rely on Asian manufacturing.

Tariffs are “stressing U.S.-based companies, in some cases past the breaking point, while not seeming to have much effect on foreign marketplace sellers who are doing business as usual,” Matt Moore, policy and general counsel of the trade group PeopleForBikes, told the Post.

PeopleForBikes said in October that lagging bike sales and consumer pullback were being exacerbated by tariff concerns.

RELATED: The rise and fall of Rad Power Bikes: From breakout success to the brink of shutdown

Rad launched in 2015 with a direct-to-consumer model and sub-$2,000 e-bikes aimed at casual riders. Demand surged during the pandemic, climbing nearly 300%, and in 2021 the company raised more than $300 million, reaching a valuation of $1.65 billion and branding itself as North America’s largest e-bike seller.

That momentum faded in 2022 as demand cooled. In its letter to employees last month, Rad said it did not anticipate “the sudden drop in consumer demand from COVID-era peaks,” leaving the company with excess inventory.

In its bankruptcy filing this week, Rad revealed a steady drop in gross revenue — from $129.8 million in 2023 to $103.8 million in 2024, and $63.3 million so far this year. The company reported total liabilities of nearly $73 million, more than double its assets of $32 million. 

Ed Benjamin, chairman of the Light Electric Vehicle Association, told the Post that tariffs created “confusion and chaos” across the industry, making future purchasing decisions difficult amid uncertainty over costs.

The Post detailed why the Biden administration allowed an exemption for e-bikes from tariffs on Chinese imports — first imposed in 2018 — to expire last year. The e-bike industry’s average tariffs have risen from about 11% to between 20% and 55%, according to PeopleForBikes.

Several industry publications have warned that layered trade policies — including China-focused tariffs, battery duties and steel restrictions — are raising prices and squeezing manufacturers. Numerous e-bike companies, including E-Cells, Kent International, Fuell, Juiced, and Electric Bike Company, have cited tariffs as a factor in shutdowns or bankruptcies.

“There’s no coherent strategy here, just a patchwork of protectionist measures that hurt importers, confuse dealers, and raise prices for consumers,” EV news website Electrek wrote. “If the U.S. wants to promote micromobility and clean transportation, it’s going to need smarter policies than this.”

A day after Rad filed for bankruptcy protection this week, U.S. Customs and Border Protection said it has collected more than $200 billion in tariffs under more than 40 executive orders issued during the Trump administration.

“This figure underscores CBP’s effectiveness in promoting secure, fair, and compliant trade,” the agency said.

The U.S. Supreme Court is weighing whether Trump exceeded his authority in imposing the tariffs. Costco and dozens of other companies have filed lawsuits seeking refunds if the court rules the duties unlawful.

Rohit Prasad, the executive who has led Amazon’s artificial intelligence initiatives and overseen the creation of its homegrown Nova AI models, is leaving the company at the end of the year.

In a memo Wednesday morning, Amazon CEO Andy Jassy named Peter DeSantis, a 27-year company veteran and top cloud infrastructure executive, to lead a new organization that combines its Nova and model research teams with custom silicon and quantum computing.

AI researcher Pieter Abbeel, who joined the company last year when Amazon hired the founders of robotics startup Covariant, will lead the frontier model research team within Amazon’s AGI organization, while continuing his work with the company’s robotics team, according to Jassy’s memo.

News of Prasad’s departure comes two weeks after Amazon unveiled its Nova 2 models at its annual re:Invent conference. The company is attempting to close the gap with AI rivals including OpenAI and Google in the race to develop increasingly capable AI systems.

Prasad joined Amazon in 2013 during the early days of Alexa and was named senior vice president and head scientist for artificial general intelligence in mid-2023 as part of a broader effort to recharge the company’s AI initiatives in the face of stiff competition.

In his memo, Jassy framed the reorganization as an effort to unify Amazon’s most important AI bets at an “inflection point” for the technologies. The new organization will bring together Amazon’s most expansive AI models, including Nova and AGI, with its custom silicon development group, which builds chips including Graviton, Trainium, and Nitro, as well as its quantum computing efforts.

DeSantis, who will report directly to Jassy, has led some of Amazon’s biggest technical initiatives. He launched Amazon EC2, the company’s core cloud computing infrastructure, oversaw the acquisition of chip designer Annapurna Labs in 2015, and most recently ran AWS Utility Computing, which includes compute, storage, database, and AI services.

Jassy called him a leader with “unusual technical depth” and a track record of “solving problems at the edge of what’s technically possible.”

Prasad’s departure was mentioned toward the end of Jassy’s memo, with the Amazon CEO saying that Prasad “has built a strong team, differentiated technology, growing customer momentum, and a culture of ambitious invention.” The memo described the departure as Prasad’s decision, calling him “missionary, passionate, and selfless” and thanking him for “everything he’s built here.”

It’s not yet clear what Prasad will do next.

In a separate memo, AWS CEO Matt Garman laid out a new internal structure consisting of seven AWS groups: Compute, Platform, and AI Services (led by Dave Brown); Storage and Analytics (Mai-Lan Tomsen Bukovec); Databases (G2 Krishnamoorthy); Security and Observability (Chet Kapoor); Agentic AI (Swami Sivasubramanian); Applied AI Solutions (Colleen Aubrey); and Infrastructure and Region Services (Prasad Kalyanaraman).

Amazon has positioned itself as a major player in enterprise AI through its Bedrock platform. Its Nova models are competitive on industry benchmarks. The Nova Forge service, launched at re:Invent, lets businesses and developers customize models using their own data.

Separately at re:Invent, the company unveiled a series of “frontier agents,” aiming to get ahead of the industry’s push toward autonomous AI systems for businesses.

But Amazon is still generally viewed as a fast follower in generative AI, trailing OpenAI, Google, and others in the perception of frontier model capabilities. Amazon has partnered closely with Claude maker Anthropic as a counterpunch to Microsoft’s partnership with OpenAI.

More recently, Amazon has also been in discussions to invest $10 billion or more in OpenAI, according to a report this week from The Information, citing three people familiar with the talks.

In an interview with GeekWire at re:Invent, Prasad gave no hint that he was preparing to leave. He described Nova Forge as “a game changer” and said the company was focused on proving that AI could deliver real value for its business customers.

Prasad took a pragmatic view of artificial general intelligence, pushing back on what he described as Silicon Valley’s tendency to portray AGI as “some kind of a god power.” He spoke instead of developing “generally intelligent systems that you can specialize for your purpose.”

A fresh analysis of tidal perturbations on Titan challenges a long-held hypothesis: that the cloud-shrouded Saturnian moon harbors an ocean of liquid water beneath its surface ice. But the scientists behind the analysis don’t rule out the possibility that smaller pockets of subsurface water could nevertheless provide a home for extraterrestrial life.

“The search for extraterrestrial environments is fundamentally a search for habitats where liquid water coexists with sustained sources of energy (chemical, sunlight, etc.) over geological time scales. Our new results do not preclude the existence of such environments within Titan, but rather, further support their plausibility,” University of Washington planetary scientist Baptiste Journaux, a co-author of the study published in Nature, told GeekWire in an email.

Journaux acknowledged that the results don’t match up with conventional wisdom. He said they represent a “true paradigm shift” in how scientists think Titan is put together.

“When the first indications from the new data analysis suggested the absence of a global ocean within Titan, the result prompted extensive discussion, careful double- and triple-checking, and contacting colleagues outside the team for critical feedback, even before submission for anonymous peer review,” he said. “We were all surprised, to say the least.”

The hypothesis about Titan’s hidden ocean goes back to NASA’s Cassini mission, which gathered data about Saturn and its moons between 2004 and 2017. “The Cassini spacecraft’s numerous gravity measurements of Titan revealed that the moon is hiding an underground ocean of liquid water,” according to the current version of NASA’s webpage about Titan.

Journaux and his colleagues used improved, up-to-date techniques to put Cassini’s radiometric measurements through a fresh round of analysis — and came to a different conclusion.

The earlier round of research proposed that there was a layer of liquid water potentially measuring hundreds of miles in thickness, sandwiched between Titan’s outer shell of low-pressure ice and a denser layer of high-pressure ice. That hypothesis was based on the best information available at the time about how tidal stresses propagated through Titan’s interior.

In contrast, the newly published research finds insufficient evidence for a liquid layer that large. Instead, it suggests that there’s an upper layer of low-pressure ice, roughly 106 miles (170 kilometers) thick, which transitions into a 235-mile-thick (378-kilometer-thick) layer of high-pressure ice.

Pockets of slush and liquid water could exist within and between layers of ice, or between the deepest layer of ice and Titan’s core. That gives Journaux cause for hope.

“Even a conservative melt fraction of 1% of the hydrosphere (to account for the observed tidal dissipation) would still correspond to total volumes of liquid water inside Titan comparable to that of the entire Atlantic Ocean, implying the presence of vast potential habitable spaces,” Journaux said.

Journaux pointed out that ice tends to exclude salts and other dissolved materials as it freezes, which means “these slushy, near-melting environments would be enriched in dissolved species and nutrients for life to feed on, as opposed to a dilute open ocean.”

“For these reasons, there is strong justification for continued optimism regarding the potential for extraterrestrial life on Titan,” he said.

Such life would probably be most similar to the types of organisms found in sea-ice ecosystems on Earth. “This realization helps constrain the range of plausible life forms and signatures to target, thereby sharpening and strengthening our search strategies,” Journaux said.

Titan’s interior is by no means the Saturnian moon’s only region of interest: Titan also has lakes of liquid ethane and methane, plus an atmosphere that’s rich in hydrocarbons. If life exists on the surface, most astrobiologists say it would be nothing like life as we know it today.

NASA’s Dragonfly mission, which is due to lift off from Earth in 2028 and touch down on Titan in 2034, could provide new insights about the moon’s surface conditions and its interior structure.

Looking beyond Titan, there are several other icy moons in our solar system that are thought to harbor hidden reservoirs of water, including the Saturnian moon Enceladus and three of Jupiter’s moons: Europa, Callisto and Ganymede. Those three Jovian worlds will get a close look from the European Space Agency’s Juice spacecraft (launched in 2023) and NASA’s Europa Clipper (launched in 2024).

Journaux hopes the results announced today will help other scientists get a better sense of what they should be looking for on all of these icy moons. “As our understanding of their interiors will become much more accurate and refined with upcoming missions … this result shows us how we can, with new measurements, place much stronger and more precise constraints on the types of habitable environments that may exist,” he said.

Flavio Petricca of NASA’s Jet Propulsion Laboratory is the corresponding author of the study published in Nature, “Titan’s Strong Tidal Dissipation Precludes a Subsurface Ocean.” In addition to Journaux, co-authors include Steven D. Vance, Marzia Parisi, Dustin Buccino, Gael Cascioli, Julie Castillo-Rogez, Brynna G. Downey, Francis Nimmo, Gabriel Tobie, Andrea Magnanini, Ula Jones, Mark Panning, Amirhossein Bagheri, Antonio Genova and Jonathan I. Lunine.

Welcome back to another Startup Radar, where we highlight up-and-coming early stage startups across the Seattle region.

This time we’re spotlighting founders working on digitizing golf scorecards, automating port operations, developing protein sequencing hardware, and tackling e-commerce returns.

Read on for brief descriptions of each company — and a pitch assessment from GPT-powered “Mean VC,” which we prompt to offer both positive and critical feedback.

Check out past Startup Radar posts here, and email me at taylor@geekwire.com to flag other companies and startup news.

Barkie

Founded: 2025

The business: Sports tech startup aiming to digitize scorecards for golfers. The goal is to preserve the traditional experience on the course while giving golfers advanced analytics after their rounds. Barkie is developing social features within its app and is launching next month.

Leadership: CEO Dane Renkert was a sales leader at Docugami, Komiko, and Ben Kinney Companies. He’s also a competitive golfer who placed 13th at the 2009 World Long Drive Championships. His former Docugami colleague Zubin Wadia is a technical advisor.

Mean VC: “There’s something charming about modern tech wrapped in the polite culture of golf — your timing’s smart, too, with golf booming post-COVID. But digitizing scorecards isn’t exactly a deep moat, and Arccos, Hole19, and 18Birdies already live here. The social angle could be your wedge in, but it needs to be more than ‘Strava for golfers’ if you want retention after the novelty fades.”

Gatein

Founded: 2025

The business: Software to help ports track containers using automation and AI-powered computer vision technology. The company is initially targeting small and medium-sized ports that it says have historically been excluded from being able to automate port operations. Gatein is bootstrapped and recently began working on a site in Europe. 

Leadership: CEO Bernardo Mendez-Arista is a longtime product leader with stints at Amazon, Volley Automation, RRAI, and Yaskawa Motoman. Co-founder and CTO Michael Pivtoraiko also worked at Volley and previously started two logistics tech companies.

Mean VC: “Smart to go after small and mid-size ports — they’re often overlooked, yet desperate for modernization. But port operators move slow, and convincing them to adopt computer vision and AI is more about trust and track record than flashy tech. If you can show real savings and keep integration friction low, you might quietly build a very sticky business.”

Primary Bioscience

Founded: 2023

The business: Early stage biotech startup developing a protein sequencing device. The company is stealthy for now but its website describes the business as a “comprehensive proteomics platform” for multi-biomarker screening. Primary Bioscience graduated from the Creative Destruction Labs’ Vancouver program earlier this year.

Leadership: CEO Stacy Anderson most recently spent three years at Roche as a protein engineering scientist. She earned a Ph.D. from the University of Wisconsin-Madison.

Mean VC: “Proteomics is white-hot, and if your device can actually accelerate multi-biomarker screening, the upside is massive. But ‘comprehensive platform’ is vague, and stealth mode works better when people already know you’re dangerous. The science background is strong, but the key question is: are you a tool company, a data company, or a pharma play pretending to be a device?”

Return Stack

Founded: 2025

The business: A “reverse logistics” startup that helps online retailers process and restock customer returns. It uses computer vision and AI agents to authenticate, grade, and resell returned items. The startup recently opened its first warehouse in Indianapolis to process returns. 

Leadership: Founder and CEO Mayank Sharma spent more than a decade at Amazon, where he led teams working on last mile logistics and returns. Co-founder Maria Pavlovskaia was an engineering leader at Amazon and Uber. 

Mean VC: “Reverse logistics is messy, costly, and growing — which makes it exactly the kind of boring problem that could be a monster business. Having ex-Amazon leadership is helpful here, but you’ll need to prove your tech handles real-world chaos, not just demo videos. The first warehouse is a good start; now show you can scale ops without your gross margins vanishing into the void.”

Rad Power Bikes filed for Chapter 11 bankruptcy protection even as the Seattle-based company said it’s working toward a sale to keep the popular electric bike brand alive.

In a bankruptcy petition, filed Monday in federal court in Spokane, the company reported total liabilities of nearly $73 million, more than double its assets of $32 million. The filing also revealed a steady drop in gross revenue — from $129.8 million in 2023 to $103.8 million in 2024, and $63.3 million so far this year.

The filing comes three weeks after the Consumer Product Safety Commission (CPSC) issued a warning to consumers to stop using some of the Seattle-based company’s bikes because of danger posed by their lithium-ion batteries.

It follows the revelation, in early November, that the once hard-charging startup was fighting for survival as it faced “significant financial challenges.”

A Rad spokesperson said in a statement provided to GeekWire on Tuesday that the company was navigating an extraordinary period of challenge and change.

“As we work to secure a sustainable future for the Rad brand, Rad has filed for Chapter 11 protection as part of a process to complete a sale of the company within the next 45–60 days,” the statement said. “This step allows us to keep operating in the ordinary course of business while we pursue the best possible outcome for the people who rely on Rad every day.”

Rad said its goal is to keep the company intact and preserve relationships it has built with riders, vendors, suppliers, and partners.

RELATED: The rise and fall of Rad Power Bikes: From breakout success to the brink of shutdown

Rad previously filed notice with the Washington state Employment Security Department in which it said the company could shut down as early as January, and that 64 jobs would be impacted.

The bankruptcy filing shows that the company remains primarily controlled by its founder, Mike Radenbaugh, who holds the largest individual stake, more than 41%.

Institutional investors hold significant minority positions, including VCVC V LLC (6.55%), an investment vehicle associated with Cercano Management, and Durable Capital Master Fund LP (5.79%). Co-founder Ty Collins retains a 4.23% stake.

The company’s largest unsecured debts include nearly $8.4 million owed to U.S. Customs and Border Protection for tariffs, and more than $8 million to overseas manufacturers. Insurance companies and individuals seeking to recover payouts related to Rad bikes are owed about $4.3 million, and two people are each owed $1 million for damages, likely from lawsuits.

Rad was conceived in 2007 by Radenbaugh and Collins, who met as students at Humboldt State University in Northern California and built their first e-bike together. After years of doing custom conversions of traditional bikes to electric, they launched their company as a direct-to-consumer brand in 2015.

Rad saw big demand amid the pandemic as more people bought e-bikes. Its sales and workforce surged and it raised more than $300 million from investors in 2021. The company was valued at $1.65 billion that year, according to PitchBook, making it one of a handful of “unicorn” startups in the Seattle region at the time.

Rad operates out of a headquarters and flagship retail location on NW 52nd Street in Seattle’s Ballard neighborhood.

The company is currently led by CEO Kathi Lentzsch, who previously ran Bartell Drugs as CEO before the company sold to Rite-Aid in 2020. She also led companies including Gump’s and Elephant Pharmacy, and held exec roles at Enesco, Pottery Barn and World Market.

Lentzsch replaced Phil Molyneux, the former Sony president who stepped down earlier this year after leading Rad for more than two years.

The CPSC’s Nov. 24 product safety warning, which listed a variety of Rad bikes and battery models, urged consumers to immediately remove and dispose of hazardous batteries that “can unexpectedly ignite and explode, posing a fire hazard to consumers, especially when the battery or the harness has been exposed to water and debris.”

Rad disputed the CPSC’s findings, saying at the time that the company “firmly stands behind our batteries and our reputation as leaders in the e-bike industry, and strongly disagrees with the CPSC’s characterization of certain Rad batteries as defective or unsafe.”

Rad said the significant cost of CPSC’s all-or-nothing recall demand would force Rad to shut down immediately with no way to support its riders or employees.

On Tuesday, Rad said it was “not giving up” and that it was “focused on doing everything we can to strengthen the future of the Rad brand.”

— Lisa Qian is the first AI Officer for the City of Seattle.

“This strategic leadership position is designed to ensure Seattle harnesses the transformative potential of artificial intelligence while upholding the city’s values and commitment to responsible technology use,” the Seattle Information Technology Department stated on LinkedIn.

Seattle is vying to establish itself as an AI heavyweight, touting the contributions of Amazon and Microsoft, smaller companies and startups, and the University of Washington, among others. In March, the city launched “AI House,” a first-in-the-nation hub designed to bring entrepreneurs, investors, students and community leaders together to propel the field.

Before this role, Qian was at LinkedIn where she served as a senior manager of data science. Past jobs include leadership positions at Seattle logistic company Convoy and at Airbnb.

“As a proud Seattle resident, I’m excited to apply my experience building responsible data science and AI systems toward work that directly benefits our community,” Qian said on LinkedIn.

— Microsoft has recruited a longtime media leader for its AI news product: Julia Beizer, current chief operating officer at Bloomberg Media. Adweek broke the news.

Beizer will report to Microsoft AI CEO Mustafa Suleyman, Adweek states, and will work on products including its Copilot Daily news round up and Publisher Content Marketplace, which pays publishers for content that’s used by AI products. (The New York Times sued OpenAI and Microsoft two years ago, alleging they illegally used the media outlet’s content to train its AI.)

Beizer has been with Bloomberg for eight years, helping drive subscriber growth with a focus on marketing, user experience and customer insights. She previously worked at HuffPost and was with The Washington Post for more than a decade.

— Seattle’s DexCare announced Rakshay Jain as its new chief product officer.

Dexcare’s software platform helps healthcare providers manage their system’s capacity and schedule appointments. The startup launched at Providence, spinning out from the healthcare network’s digital innovation group in 2021.

“What drew me to DexCare is that this team isn’t trying to replace systems already in use, but connect them, and create the navigational intelligence that guides where, when, and how patients access care, no matter where they enter the system,” Jain said in a statement.

Jain joins the company from Innovaccer, a Bay Area company providing software for managing healthcare data. He will work remotely from California.

— Sri Chandrasekar is taking a break after nearly nine years as a managing partner at Point72 Ventures. Chandrasekar, who is based in Bellevue, Wash., appears to have resigned from the board of directors of four startups located across the U.S. and in London, while retaining seats at two others.

“To my Ventures team – It was amazing going to battle with you over the last 9 years. I expect nothing but great things from you in the years to come,” he wrote in LinkedIn. “To our Portfolio companies – Working with you is what made the long nights and the non-stop travel worthwhile. If you need me, you know how to find me!”

Prior to Point72, Chandrasekar was a senior vice president at In-Q-Tel in Menlo Park, Calif.

— Brian Fleming is stepping away from Sucker Punch, the Bellevue, Wash.-based game studio he co-founded 28-years ago. “I’ve decided it’s time to shake up the snow-globe,” he wrote on LinkedIn, adding that he’ll stay at the company — part of PlayStation Studios — through April to assist with the leadership transition. Jason Connell and Adrian Bentley will take over as co-studio heads on Jan. 1.

— Jiphun Satapathy is now chief information security officer for Motive, a San Francisco company providing software tools to make operations safer for construction, field service, energy, trucking and other industries. Satapathy, who is based in the Seattle area, previously worked as CISO for Medallia and has held leadership roles at Amazon Web Services, Snowflake and elsewhere.

— Janet Greenlee, director of communications at Allen Family Philanthropies, is retiring from her role. Greenlee has spent her career in marketing and communications. She worked for more than a decade at the philanthropy, which was launched by Microsoft co-founder Paul Allen and his sister, Jody Allen.

Three U.S. senators have launched an investigation into whether tech giants, driven by soaring AI energy demands, are raising residential power bills. Separately, Amazon released a white paper Tuesday stating that its data centers are not the problem, and that in some regions it actually pays more than required for energy use.

The Democratic senators sent letters to Amazon, Microsoft, Google, Meta and three data center firms, according to the New York Times. The lawmakers raised concerns that energy demand driven by artificial intelligence was forcing utilities to deploy new power plants and upgrade the grid — with local ratepayers helping foot the bill.

“We write in light of alarming reports that tech companies are passing on the costs of building and operating their data centers to ordinary Americans as A.I. data centers’ energy usage has caused residential electricity bills to skyrocket in nearby communities,” the senators said, according to the Times.

Elizabeth Warren of Massachusetts, Chris Van Hollen of Maryland and Richard Blumenthal of Connecticut issued the letters.

Amazon offered a much different take in an analysis that examined the potential benefits, costs and risks that large energy loads created by data centers have on electric utilities.

The Amazon-funded study found that in some locations, the power bills currently being paid by the company more than cover the utility impacts. A typical 100 megawatt data center was estimated to pay an additional $3.4 million beyond the costs associated with its electricity use, which also include a utility’s infrastructure upgrades, deployment of new energy generation, operations and maintenance.

Utilities can use that surplus “to reduce rates for other ratepayers, but how this potential benefit is realized will differ across jurisdictions,” the study stated.

The assessment examined Amazon data center campuses in Oregon, California and Mississippi and was performed by E3, an independent economic consulting firm.

The white paper said the benefit to the utilities and other customers should continue into 2030, but noted that utilities will need to adjust their rates in the future to ensure that ratepayers are not subsidizing tech operations.

“To continue to prevent cross-subsidization, utilities must keep pace and leverage the full range of tools available to them to mitigate these risks…” the document states.

The rapid growth of the sector is central to the debate. A Department of Energy report projected that data center energy use, which was more than 4% of U.S. electricity consumption in 2023, could triple by 2028. This forecast is fueled by tech giants’ expanding investment: Microsoft and Amazon each reported nearly $35 billion in capital expenditures in the third quarter, much of it on data center infrastructure.

The tech giants are also investing globally in new wind and solar power and energy storage, while pursuing more costly power sources including nuclear.

In October, a group of U.S. representatives raised similar concerns to the senators’, asking the Federal Energy Regulatory Commission, Edison Electric Institute and the Data Center Coalition for information regarding data center impacts on residential power bills “to help ensure everyday Americans and small businesses aren’t bearing the brunt of data center energy costs.”

Washington representatives Kim Schrier and Adam Smith as well as Oregon Rep. Andrea Salinas were among the 20 lawmakers who made the request.

The investigations come amid a general rise in household expenses, making the allocation of utility costs particularly contentious. Residential electricity costs nationwide are on the rise, according to federal data. Power bills rose more than 7% on average when comparing September rates to a year earlier.

But the causes of the increase are complicated. A study this month in a peer-reviewed journal concluded that multiple factors impact electricity prices, including inflation, fluctuating gas prices and natural disasters such as hurricanes, storms and wildfires.

Washington and 35 other states reached a settlement with Hyundai and Kia in which the automakers will provide restitution to consumers and fixes to millions of eligible vehicles nationwide that lacked industry-standard, anti-theft technology.

Washington Attorney General Nick Brown’s office announced details of the settlement Tuesday, in which Hyundai and Kia have agreed to:

• Equip all future vehicles sold in the U.S. with engine immobilizer anti-theft technology;
• Offer free zinc-reinforced ignition cylinder protectors to owners or lessees of eligible vehicles, including vehicles that previously were only eligible for the companies’ software updates;
• Provide up to $4.5 million in restitution to eligible consumers whose cars are damaged by thieves; and
• Pay $4.5 million to the states to defray the costs of the investigation. 

Eligible car owners can receive up $4,500 for a total loss or up to $2,250 for a partial loss, according to compensation details on the settlement website. The claim deadline is March 31, 2027.

An engine immobilizer prevents thieves from starting a vehicle’s engine without the vehicle’s “smart” key, which stores the vehicle’s electronic security code. The lack of the necessary tech on cars resulted in “an epidemic of car thefts and joy riding” across Washington and the country,

“Security is a key piece for families looking to buy a vehicle, but Hyundai and Kia spent years selling people cars that lacked the industry’s standard protections,” Brown said in a statement. “Year after year, consumers have been easily victimized because of the automakers’ failure here.”

In late 2020, teenage boys began posting videos on social media describing how to steal the cars simply by removing a plastic piece under the steering wheel and using a USB cord. Posts with the hashtag “Kia Boys” racked up more than 33 million views on TikTok by September 2022, according to CNBC. The videos included teens engaged in reckless driving of the stolen vehicles.

Despite years of evidence, Hyundai and Kia waited until 2023 to launch a service campaign to update the software on most of the affected vehicles, Brown’s office said. The update was easily bypassed by thieves.

Seattle City Attorney Ann Davison filed a similar lawsuit against Kia and Hyundai in January 2023.

“The corporate choices by the automakers to cut corners have had very negative impacts and put the public at risk in Seattle,” Davison said in a statement Tuesday.

She added, “I am confident that we will prevail in our lawsuit, and the car companies will finally be required to help in the fight to improve public safety. My action is not to replace criminal prosecution of car thieves, but to hold corporate actors accountable for making choices that prioritize profit over public safety.”

In May 2023, Hyundai and Kia agreed to a consumer class-action lawsuit settlement worth $200 million over rampant thefts of the Korean automakers’ vehicles. The Seattle City Attorney’s Office said at the time that it was a “good first step for consumers” but that the settlement involving individual owners “does not include the litigation brought by the City.”

Under the new multistate settlement, eligible consumers will be notified by the companies that they will have one year from the date of the notice to make an appointment to have the zinc-reinforced ignition cylinder protector installed at their local Hyundai or Kia authorized dealerships. Consumers are urged to schedule the installation of the zinc-reinforce ignition cylinder protector as soon as possible.

Consumers who previously installed the software update on their vehicles (or were scheduled to do so) but nonetheless experienced a theft or attempted theft of their vehicle on or after April 29, 2025, are eligible to file a claim for restitution for certain theft and attempted-theft related expenses.  For more information about eligibility and how to submit a claim visit these sites for Hyundai and Kia.

How many penguins are in this wildlife video? Can you track the orange ball in the cat video? Which teams are playing, and who scored? Give me step-by-step instructions from this cooking video?

Those are examples of queries that can be fielded by Molmo 2, a new family of open-source AI vision models from the Allen Institute for AI (Ai2) that can watch, track, analyze and answer questions about videos: describing what’s happening, and pinpointing exactly where and when.

Ai2 cites benchmark tests showing Molmo 2 beating open-source models on short video analysis and tracking, and surpassing closed systems like Google’s Gemini 3 on video tracking, while approaching their performance on other image and video tasks.

In a series of demos for reporters recently at the Ai2 offices in Seattle, researchers showed how Molmo 2 could analyze a variety of short video clips in different ways. 

• In a soccer clip, researchers asked what defensive mistake led to a goal. The model analyzed the sequence and pointed to a failure to clear the ball effectively.
• In a baseball clip, the AI identified the teams (Angels and Mariners), the player who scored (#55), and explained how it knew the home team by reading uniforms and stadium branding.
• Given a cooking video, the model returned a structured recipe with ingredients and step-by-step instructions, including timing pulled from on-screen text.
• Asked to count how many flips a dancer performed, the model didn’t just say “five” — it returned timestamps and pixel coordinates for each one.
• In a tracking demo, the model followed four penguins as they moved around the frame, maintaining a consistent ID for each bird even when they overlapped.
• When asked to “track the car that passes the #13 car in the end,” the model watched an entire racing clip first, understood the query, then went back and identified the correct vehicle. It tracked cars that went in and out of frame.

Big year for Ai2

Molmo 2, announced Tuesday morning, caps a year of major milestones for the Seattle-based nonprofit, which has developed a loyal following in business and scientific circles by building fully open AI systems. Its approach contrasts sharply with the closed or partially open approaches of industry giants like OpenAI, Google, Microsoft, and Meta.

Founded in 2014 by the late Microsoft co-founder Paul Allen, Ai2 this year landed $152 million from the NSF and Nvidia, partnered on an AI cancer research initiative led by Seattle’s Fred Hutch, and released Olmo 3, a text model rivaling Meta, DeepSeek and others.

Ai2 has seen more than 21 million downloads of its models this year and nearly 3 billion queries across its systems, said Ali Farhadi, the Ai2 CEO, during the media briefing last week at the institute’s new headquarters on the northern shore of Seattle’s Lake Union. 

As a nonprofit, Ai2 isn’t trying to compete commercially with the tech giants — it’s aiming to advance the state of the art and make those advances freely available.

The institute has released open models for text (OLMo), images (the original Molmo), and now video — building toward what he described as a unified model that reasons across all modalities.

“We’re basically building models that are competitive with the best things out there,” Farhadi said — but in a completely open manner, for a succession of different media and situations.

In addition to Molmo 2, Ai2 on Monday released Bolmo, an experimental text model that processes language at the character level rather than in word fragments — a technical shift that improves handling of spelling, rare words, and multilingual text.

Expanding into video analysis

With the newly released Molmo 2, the focus is video. To be clear: the model analyzes video, it doesn’t generate video — think understanding footage rather than creating it.

The original Molmo, released last September, could analyze static images with precision rivaling closed-source competitors. It introduced a “pointing” capability that let it identify specific objects within a frame. Molmo 2 brings that same approach to video and multi-image understanding.

The concept isn’t new. Google’s Gemini, OpenAI’s GPT-4o, and Meta’s Perception LM can all process video. But in line with Ai2’s broader mission as a nonprofit institute, Molmo 2 is fully open, with its model weights, training code, and training data all publicly released.

That’s different from “open weight” models that release the final product but not the original recipe, and a stark contrast to closed systems from Google, OpenAI and others.

The distinction is not just an academic principle. Ai2’s approach means developers can trace a model’s behavior back to its training data, customize it for specific uses, and avoid being locked into a vendor’s ecosystem.

Ai2 also emphasizes efficiency. For example, Meta’s Perception LM was trained on 72.5 million videos. Molmo 2 used about 9 million, relying on high-quality human annotations.

The result, Ai2 claims, is a smaller, more efficient model that outperforms their own much larger model from last year, and comes close to matching commercial systems from Google and OpenAI, while being simple enough to run on a single machine.

When the original Molmo introduced its pointing capability last year — allowing the model to identify specific objects in an image — competing models quickly adopted the feature.

“We know they adopted our data because they perform exactly as well as we do,” said Ranjay Krishna, who leads Ai2’s computer vision team. Krishna is also a University of Washington assistant professor, and several of his graduate students also work on the project.

Farhadi frames the competitive dynamic differently than most in the industry.

“If you do real open source, I would actually change the word competition to collaboration,” he said. “Because there is no need to compete. Everything is out there. You don’t need to reverse engineer. You don’t need to rebuild it. Just grab it, build on top of it, do the next thing. And we love it when people do that.”

A work in progress

At the same time, Molmo 2 has some clear constraints. The tracking capability — following objects across frames — currently tops out at about 10 items. Ask it to track a crowd or a busy highway, and the model can’t keep up.

“This is a very, very new capability, and it’s one that’s so experimental that we’re starting out very small,” Krishna said. “There’s no technological limit to this, it just requires more data, more examples of really crowded scenes.”

Long-form video also remains a challenge. The model performs well on short clips, but analyzing longer footage requires compute that Ai2 isn’t yet willing to spend. In the playground launching alongside Molmo 2, uploaded videos are limited to 15 seconds.

And unlike some commercial systems, Molmo 2 doesn’t process live video streams. It analyzes recordings after the fact. Krishna said the team is exploring streaming capabilities for applications like robotics, where a model would need to respond to observations in real time, but that work is still early.

“There are methods that people have come up with in terms of processing videos over time, streaming videos,” Krishna said. “Those are directions we’re looking into next.”

Molmo 2 is available starting today on Hugging Face and Ai2’s playground.

Don’t take your eyes off the road to read new research from the University of Washington.

In partnership with Toyota Research Institute, UW researchers are exploring how modern touch screens in cars affect driving now that dashboard knobs and buttons are increasingly a thing of the past. The results could help auto manufacturers design safer, more responsive screens and in-car interfaces.

The team’s study, which was presented this fall at the ACM Symposium on User Interface Software and Technology in Busan, Korea, adds to what we already know about the dangers of distracted driving when it comes to phone use.

Participants driving a vehicle simulator and interacting with a touch screen (see video above) were given memory tests that mimic the mental effort demanded by traffic conditions and other distractions, the UW reported. Sensors tracked their gaze, finger movements, pupil diameter and electrodermal activity.

While driving, participants had to touch specific targets on a 12-inch touch screen, similar to how they would interact with apps and widgets. They did this while completing three levels of an “N-back task,” a memory test in which the participants hear a series of numbers, 2.5 seconds apart, and have to repeat specific digits. 

Researchers found that when people try to multitask behind the wheel, their driving and their ability to use a touch screen both suffer. The simulator car drifted in its lane, and speed and accuracy using the screen declined while driving.

“Touch screens are widespread today in automobile dashboards, so it is vital to understand how interacting with touch screens affects drivers and driving,” said co-senior author Jacob O. Wobbrock, a UW professor in the Information School. “Our research is some of the first that scientifically examines this issue, suggesting ways for making these interfaces safer and more effective.”

Popular Mechanics wrote about the mental bandwidth and finger precision that many modern infotainment screens require in cars.

Based on the UW/Toyota findings, researchers suggest future in-car touch screen systems might use simple sensors in the car — eye tracking, or touch sensors on the steering wheel — to monitor drivers’ attention and cognitive load. Based on these readings, the car’s system might adjust the touch screen’s interface to make important controls more prominent and safer to access.

This story originally appeared on Real Estate News.

Could Google crush the “portal wars” once and for all?

A key Google partner is starting to display home listing details directly in search results, prompting some industry experts and analysts to question what impact the feature could have on the traffic — and financials — of major portal players like Zillow, Realtor.com and others.

A ‘controlled experiment’: In some markets, Google’s data partner HouseCanary and its IDX site ComeHome are beginning to experiment with placing home listings at the top of Google search results, complete with basic details, price, images and a “Request a tour” button. According to HouseCanary, the company is licensed in all 50 states and in Washington, D.C., as a full-service brokerage. 

Real estate consultant and analyst Mike DelPrete was the first to report on the pilot listing initiative. 

HouseCanary offered some insight into the “controlled experiment” via an announcement on LinkedIn this week, suggesting that the company and Google “are innovating” and “pushing into new territory” with the effort.

“Before this test started, we contacted and notified every MLS in the regions included. We are working with those MLSs directly and we have active, ongoing communication with them throughout the test. If an MLS has questions or concerns, we address them directly and promptly,” the announcement reads. 

“The goal is simple: improve how consumers discover listings while staying aligned with the rules and expectations of the MLS community. We are excited about what we are building with Google, and we are equally committed to doing it the right way with the MLSs and other stakeholders. We will continue to communicate directly with the MLSs involved and respond quickly to any concerns.”

Impacts of previous search shifts: The move to incorporate home listing information into Google comes over a year after the search giant started integrating AI summaries directly into the top of search results. A July Pew Research Center study found that web users were less likely to click into other pages — such as news media and other outlets that had long depended on traffic as a metric for determining revenue — since Google incorporated AI summaries.

Some major mainstream news sites have seen traffic drop upwards of 30-40% year-over-year partially thanks to AI summaries, NPR reported in July. 

What analysts are saying: There may already be concern about what kind of impact home listing summaries on Google pages could have on the top portals. As the leader in home search, Zillow would be the site with the most to lose. At the time of publishing this story, Zillow’s share price has dropped more than 8% since the opening bell on Dec. 15. 

But some analysts say the concerns may be exaggerated.

“While we don’t expect a direct near-term impact on Zillow’s business, given that most of Zillow’s traffic is direct (e.g., Zillow.com, StreetEasy.com, mobile apps) and Google’s new product is currently limited to select markets and mobile browsers, we view this development as a long-term risk for real estate portals like Zillow,” Goldman Sachs analyst Michael Ng wrote in a recent note to clients, CNBC reports.

Piper Sandler called the concerns “overblown,” and analysts with Oppenheimer and Wells Fargo also appeared to be less concerned about immediate impacts on Zillow’s traffic and revenue. Instead, they suggest that the experiment may simply present a new opportunity for Google to generate more revenue.

Wells Fargo analyst Alec Brondolo sees “Zillow, Homes.com, Realtor.com, etc. bidding for home listing ad units rather than Google attempting to monetize directly with an ad product sold to agents,” CNBC reported. 

In a blog post, Victor Lund, managing partner of real estate consulting firm WAV Group, highlighted some issues with the pilot and suggested it could overstep existing norms and standards with the IDX protocol. 

“IDX was never designed to allow listings to be turned into paid media inventory on global ad networks. If this practice stands, it redefines IDX from a display-based cooperation agreement into an advertising license, something neither MLSs nor brokers have agreed to,” Lund wrote.

Real Estate News has reached out to HouseCanary for more details on the scope and scale of the experiment and to Zillow for comment on the new Google feature.

Amazon filed a new notice with Washington state Monday morning signaling that it’s cutting 84 jobs, but the individual separations are part of the regular course of business, unrelated to the 14,000 corporate layoffs it announced globally in October.

The company said each of its businesses regularly reviews its organizational structure and may make adjustments as a result. It’s a routine process, the company said, not tied to broader workforce actions.

The notice stems from a new state law that requires employers to disclose all terminations occurring within 90 days of a prior notice under the state’s new “mini” version of the Worker Adjustment and Retraining Notification Act, known as the WARN Act.

“We’ve informed a relatively small number of employees that their roles will be eliminated as the result of individual business decisions,” said Amazon spokesperson Brad Glasser. “We don’t make decisions like this lightly,” he added, noting that the company is providing affected employees with 90 days of full pay and benefits, transitional health coverage, and job placement services.

According to the filing, the separations are scheduled to occur between Feb. 2 and Feb. 23, 2026, across more than 30 Seattle and Bellevue office locations, plus six remote workers based in Washington. They include software development engineers, program managers, recruiters, HR specialists, and UX designers, ranging from entry-level to directors and principals.

Amazon noted in the filing that employees were notified starting in early November and received at least 89 days’ advance notice, exceeding the 60-day minimum required under the law. Those who find internal transfers before their separation date won’t be laid off.

Separately, the company said in October that it was cutting 14,000 corporate jobs globally as part of CEO Andy Jassy’s push to reduce bureaucracy and operate more efficiently. That earlier round included more than 2,300 layoffs in Washington state, according to a filing at the time.

Amazon HR chief Beth Galetti signaled additional cuts could continue into 2026. Reuters has reported the total could ultimately reach 30,000 — which would surpass the 27,000 positions eliminated in 2023 and mark the largest overall layoff in company history. 

Washington state is part of a newly filed lawsuit against the Trump administration, challenging the legality of a $100,000 fee for new H-1B visas that allow highly-skilled individuals to work temporarily in the U.S.

Attorneys general from 20 states claim the U.S. Department of Homeland Security set the fee at an arbitrary amount that does not reflect the agency’s costs, and that the fee was enacted without going through a required notice-and-comment process.

The visa is meant to recruit employees from abroad who have specialized expertise not found in sufficient numbers in the U.S. workforce.

Seattle-based Amazon has roughly 19,100 employees working under H-1B visas nationwide. Microsoft, which is based in Redmond, Wash., nationally employs more than 6,200 H-1B visa holders. Washington’s public universities and agencies have nearly 500 H-1B visa holders on their payrolls, according to federal data and state analysis.

Employers are responsible for paying H-1B fees, which used to run between $960 and $7,595, said Washington State Attorney General Nick Brown’s office. Raising the fees, the state warned, will result in empty university labs and science discoveries “will be made somewhere else.”

“These institutions will lose their competitive edge, particularly in the areas of artificial intelligence, cybersecurity, and medical fields,” said a press release from Brown’s office.

In announcing the increased fee in September, the Trump administration said the visa was being abused by employers to supplant Americans with “lower-paid, lower-skilled labor.”

“The large-scale replacement of American workers through systemic abuse of the program has undermined both our economic and national security,” said a White House memo addressing restrictions of nonimmigrant workers.

Priyanka Kulkarni, CEO of the immigration tech startup Casium, said the H1-B workers are not low paid, noting that the median salary for the visa holders was about $120,000 last year.

“Engineers, scientists, healthcare specialists, and educators recruited from abroad often fill critical gaps that enable companies and institutions to grow, invest, and create jobs locally,” she added via email.

The Trump administration has specifically called out high-tech companies’ use of the program, saying they “have prominently manipulated the H-1B system, significantly harming American workers in computer-related fields.”

Xiao Wang, CEO of the startup Boundless Immigration, noted that while tech giants are targeted for criticism, the visa also allows for doctors, nurses and researchers to work in the U.S. — echoing some of the concerns raised by Washington’s attorney general.

“Adding a $100K fee for all foreign talent trying to enter Washington to work in these fields would all but eliminate this pathway for anyone outside of the most valuable companies in the world and would leave the state with a significant shortage of important roles,” Wang said by email.

He added that putting a nurse and an AI engineer in the same visa category highlights an overdue need for immigration reform.

Wang called on Americans to demand that Congress “pass new immigration regulations to stay competitive as a country.”