AI Daily Briefing — 2026-06-20

A startup called Subquadratic has emerged from stealth mode with a bold assertion: it has overcome a mathematical hurdle that has constrained large language models for nearly a decade. The company says its approach dramatically reduces the number of computations required by transformers to generate responses, resulting in a faster, cheaper, and far more energy-efficient model. While skepticism remains among experts, Subquadratic has begun sharing evidence that suggests its method may indeed be significant, according to MIT Technology Review AI.

In the realm of brain-computer interfaces (BCIs), clinical trials are gaining momentum. Casey Harrell, a man with ALS, has become what researchers call the first power user of a brain implant. The device has allowed him to earn an income, reconnect with loved ones, and read to his daughter—an experience he describes as revolutionary. Over the past two years, the number of BCI trial volunteers has surged, and China recently became the first nation to approve a BCI for medical use.

Technological advances are enabling engineers to offer more features than ever before, pushing BCI research from the lab toward commercial reality. The progress marks a pivotal moment for the field, with potential implications for patients with paralysis and other neurological conditions. MIT Technology Review AI reports that the technology is now properly taking off, signaling a new era in neural interface development.

For decades, physicists have focused on weakly interacting massive particles (WIMPs) as the prime suspect for dark matter. But that hunt has hit a fresh obstacle: neutrinos. These tiny particles streaming from the sun and other stars create a "neutrino fog" that masks any potential dark matter signal. Rather than signaling defeat, this development is prompting researchers to broaden their approach. New strategies include quantum sensors, liquid-helium detectors, and even probes of Jupiter's atmosphere, as reported by MIT Technology Review AI.

Meanwhile, in Nairobi, a different kind of energy revolution is taking root. Milcah Wanjiru runs a grain-milling shop that operates on solar power or the grid, unlike the diesel-powered mills common in the area. With about a quarter of Kenya's population lacking centralized electricity, off-grid solar is gaining traction as a path to universal access by 2030. For Wanjiru, the switch cuts operating costs and boosts profits once the initial investment is recouped, highlighting a growing trend across Kenya and beyond.

Solar geoengineering, often described as an emergency climate brake, faces major practical hurdles. MIT Technology Review AI notes that scattering particles to reflect sunlight is far more complex than a simple fix. The engineering challenges remain largely unsolved, making it less like pulling a lever and more like assembling a intricate puzzle. These developments underscore how both cosmic and earthly energy quests are entering uncharted territory.

Solar geoengineering, once confined to computer models, is now facing the practical challenges of real-world implementation. Researchers are developing aircraft, materials, and systems to cool the planet by reflecting sunlight, but early efforts reveal a need for massive infrastructure, time, and investment. As MIT Technology Review AI reports, the dream of hacking the atmosphere is colliding with engineering realities, raising questions about feasibility and cost.

Meanwhile, a quiet revolution is underway in the study of interoception—the internal sense of how we feel from within. Thanks to a 2021 Nobel Prize and new tools mapping body-brain signaling, scientists are decoding how signals travel between organs and the brain. This emerging field, highlighted in an MIT Technology Review AI podcast, could reshape treatments for conditions like obesity and anxiety by targeting the hidden sense of self.

These twin stories underscore a broader theme: transformative ideas, whether in climate intervention or neuroscience, must navigate the gap between theory and practice. As researchers push boundaries, the path forward demands patience, resources, and a willingness to confront unexpected hurdles.

When Allbirds announced a pivot to artificial intelligence earlier this year, many dismissed it as a meme-stock stunt. The struggling shoe company, once a symbol of Silicon Valley style, had sold its footwear business for $43 million, raised another $100 million from investors, and rebranded as Smartbird. Now, newly appointed CEO Nadia Carlsten must turn that plan into reality—starting from scratch with zero employees.

Carlsten, a former AWS executive with an engineering PhD, officially took the helm yesterday. Her immediate task is recruiting a leadership team and securing office space. “The shoe business has officially closed,” she told TechCrunch AI from Amsterdam. “We’re going to be building a brand new team for the AI business.” The company aims to become an AI infrastructure provider, offering managed compute for firms that need direct control over their servers.

Smartbird’s target customers value data sovereignty and operational control over the scalability of public cloud services. Carlsten sees demand in industries like pharmaceuticals, energy, and finance, where companies run bespoke models. She argues Smartbird isn’t competing with hyperscalers or neoclouds, but with internal IT projects. Still, established players like Hewlett Packard and Equinix already offer similar single-tenant services.

Carlsten expects to have compute clusters deployed for several clients by year-end, focusing on agility rather than massive scale. “It’s not about large numbers of GPUs,” she said. “It’s about control of the infrastructure stack.” While some rivals have placed enormous chip orders, Smartbird plans to grow more modestly—betting that specialized workflows will justify its premium, managed approach.

The U.S. government's recent order forcing Anthropic to withdraw its latest AI models, Fable 5 and Mythos 5, has sparked intense debate. Citing national security risks after Amazon researchers reportedly bypassed Fable 5's guardrails, the ban was intended to protect against potential misuse. However, cybersecurity experts have signed an open letter calling the move dangerous, noting that similar vulnerabilities exist in competing models. This raises the question: is this a genuine security measure or another chapter in Anthropic's strained relationship with the Trump administration?

On a recent episode of TechCrunch's Equity podcast, hosts Anthony Ha, Sean O'Kane, and Rebecca Bellan explored the implications for developers relying on Anthropic's platform. They also discussed how the ban might inadvertently benefit the company ahead of its anticipated IPO. By framing Anthropic as a target of government overreach, the controversy could generate sympathy and brand loyalty among users wary of censorship.

The debate highlights the messy intersection of AI regulation, corporate interests, and national security. As the story unfolds, it remains unclear whether the ban will stifle Anthropic or inadvertently fuel its growth. For now, the incident has put the company squarely in the spotlight, for better or worse.