BriefTechNews
Ring's flying cameras, Anthropic's $1.3T IPO, and the AI timeline that just sped up

Ring's flying cameras, Anthropic's $1.3T IPO, and the AI timeline that just sped up

7 min read · 14 sources

TL;DR
  • Ring's $250 Drone Cam flies up to 8 minutes per patrol and the $1,000 Rover Cam repurposes Amazon's Astro robot for business patrolling.
  • Anthropic plans to sell 825 million shares at $70-78 in its IPO, seeking a $1.3 trillion valuation with $26 billion from Google and AMD.
  • Gemini 4 Carbon ranks #1 on CursorBench 4.0 at more than 2X lower cost per task than the next best external model.
  • Apple's Claris Harbor app clusters Macs over Thunderbolt 5 to pool unified memory for local AI inference.
  • AI Futures forecasts Superhuman AI Researcher by 2027, with a 35-65% chance of top-expert-dominating AI before 2028.

A $250 drone that flies through your house when the alarm trips, and a $1,000 robot that patrols your business with a periscope camera. Amazon’s Ring just turned its surveillance hardware autonomous, and the privacy questions that have dogged the company for years are about to get a lot more physical.

Meanwhile, Anthropic is heading for a $1.3 trillion IPO, Google shipped a model that beats its own frontier at half the cost per task, and a new analysis says the intelligence explosion arrives in 2027. Here’s what actually changed today.

Anthropic expects to raise $26 billion from Google and AMD alongside its planned $1.3 trillion IPO.

Ring's Drone Cam and Rover Cam Put Autonomous Surveillance in the Aisle

Source: cnbc.com ↗

Ring unveiled two new devices that turn the company from a passive camera vendor into an autonomous patrol company. The $250 Drone Cam is an indoor flying camera with LED lights that follows a preset path when alarms trigger, flying up to eight minutes before returning to its dock. The $1,000 Rover Cam is an autonomous robot for businesses, repurposing Amazon’s Astro, with a display and an extendable periscope camera on a two-wheeled base.

The drone was first previewed in 2020 as the Always Home Cam; it’s taken six years to ship. It has two plastic-encased propellers and emits a loud buzz, which is either a feature (deterrent) or a bug (annoying), depending on your threat model. The Drone Cam ships in the US on January 27 pending FCC approval; the Rover ships December 16.

For engineers, the interesting part isn’t the hardware. It’s that Ring - which has a documented history of police partnerships and recent backlash over AI features - now owns autonomous physical presence in your home. The attack surface isn’t just a camera feed; it’s a flying device with rotors that can be hijacked. The privacy calculus changed the moment the camera could move itself to where you are.

Apple's Harbor Turns Macs Into a Private AI Cluster

Source: macrumors.com ↗

Apple’s Claris subsidiary is testing Harbor, an app that turns Apple silicon Macs into private AI inference servers. The headline feature: connecting multiple Macs over Thunderbolt 5 to pool unified memory, letting you run larger models with higher concurrency than any single machine could handle. The app automatically detects Thunderbolt 5-connected devices, lets you browse and load open-weight models, and requires no manual configuration.

Harbor is in early access with a removed App Store listing, and its sign-up page asks for a company name - suggesting a business focus. It’s a direct competitor to tools like Exo and LM Studio, but with a key difference: Apple’s software stack, Apple’s hardware, and presumably Apple’s polish.

The engineering implication is straightforward. If you’re running local models today, you’re manually sharding across machines or renting cloud GPUs. Harbor promises to make a Thunderbolt 5 cluster of Mac Studios a viable private inference tier - no cloud, no data leaving the building, no per-token billing. The question is whether Apple prices it as a developer tool or as an enterprise product.

XPeng Says Its Flying Car Launches in China Next Summer

Source: theguardian.com ↗

XPeng’s CEO Deli Zhao says the company’s $300,000 two-seat “flying car” could be ready for recreational and emergency services use in China by the second half of 2027, following approval from China’s civil aviation authorities. The vehicle, about the size of a Fiat 500, is lifted by six battery-powered propellers and steered by a joystick, or flies autonomously to a programmed destination. The price includes a six-wheeled van that unfolds it.

Zhao claims anyone can learn to fly it in five minutes, and he dismissed Elon Musk’s flying Tesla prototype as “hopping or jumping” rather than real flying. Flight range is 30 kilometers, which XPeng says will soon double.

Don’t get carried away. This is a recreational and emergency services vehicle, restricted to approved park areas. It’s not a commuter product. But the technical integration - drone-style propulsion, autonomous flight, and a ground vehicle that deploys it - is the real engineering story. The hard part isn’t the flying; it’s the certification, the battery density, and the regulatory framework that lets it operate.

Source: pcmag.com ↗

The Starlink Mini 2 is the first dish to support SpaceX’s swappable battery, which lasts up to 4.5 hours. Early adopters pay $999, which includes a full year of Starlink Roam Unlimited - usually valued at $2,100. Shipments begin mid-December.

So the battery is effectively free if you were going to buy the roaming plan anyway, but the upfront cost is triple what the original Mini cost. For field engineers, remote workers, and disaster-response teams, a hot-swappable battery on a dish that’s already tiny is genuinely useful. The question is whether the $999 price point makes sense versus a $50 power bank and a USB-C adapter. For now, SpaceX is betting on convenience over cost.

Gemini 4 Carbon: Frontier Benchmarks at Half the Cost

Source: blog.google ↗

Google released Gemini 4 Carbon, its strongest model for cybersecurity and long-horizon agentic workflows. It’s available via API in Google AI Studio and Gemini for Business at the same introductory price as Gemini 4 Argon: $2 per million input tokens, $10 per million output tokens.

Carbon ranks #1 on benchmarks like CursorBench 4.0 and WANDR at more than 2X lower cost per task than the next best external model. It powers a new /swarm feature in Google Antigravity for coordinated agent teams, and rolls out to Google AI Ultra subscribers.

The economics are the story. Frontier accuracy at half the cost per task changes the math for production workloads - code review, debugging, scientific reasoning - where you previously had to choose between capability and budget. Carbon says you can have both.

Modern Web Guidance 1.0 Gives Coding Agents Browser Expertise

Source: developer.chrome.com ↗

Chrome announced the stable release of Modern Web Guidance 1.0, a set of skills that embed web platform expertise and browser compatibility data into coding agents. The release includes a new modernization workflow, an expanded guide database, deeper human expert review, and a public eval dashboard. Firefox and Edge joined as cross-browser contributors.

Installation is npx modern-web-guidance@latest install, and it works with any coding agent. If you’ve ever had an LLM generate code that breaks in Safari or ignores accessibility, this is aimed directly at that failure mode. The eval dashboard means you can actually verify the guidance works rather than trusting the marketing.

AI Futures Says the Intelligence Explosion Arrives in 2027

Source: blog.aifutures.org ↗

A new analysis from AI Futures argues AI progress is accelerating faster than expected, citing a METR report estimating Anthropic is making progress about twice as fast as historical rates, with ~80% of the acceleration due to AI. They now forecast Superhuman AI Researcher (SAR) arriving in 2027, with a 35-65% chance of Top-Expert-Dominating AI (TED-AI) before 2028.

The authors urge an industry-wide pause or pace on frontier AI development, warning of risks from misaligned SAR and potential theft of training recipes by China. Whether you buy the timeline or not, the underlying claim - that AI is accelerating AI development - is measurable and hard to dismiss. For engineers, the practical implication is that capabilities you’re planning for in 2029 might arrive in 2027.

Anthropic Files for a $1.3 Trillion IPO

Source: wsj.com ↗

Anthropic plans to sell 825 million shares at $70 to $78 a share in its IPO, expected to start trading Thursday next week. The company expects to raise an additional $26 billion from big investors, including Google and AMD. Some executives and founding members plan to sell around 23.6 million shares. At the planned valuation, it won’t top SpaceX’s record-breaking offering in June.

A $1.3 trillion valuation for a company that, by most accounts, is still burning cash at a staggering rate tells you everything about where the market thinks AI is headed. The fact that Google and AMD - both of whom have their own AI ambitions - are writing checks says even more.

The Rest: Glue Work, Reddit Reviews, Google Fi, and Agent Traffic

Source: seangoedecke.com ↗

Glue work is still harmful in the age of AI, argues Seán Gedecke, and AI actually reduces its importance. Agents cut coordination needs, gatekeeper engineers are rarer post-ZIRP, fewer juniors mean less onboarding, and LLMs automate write-only reports. But communication with other engineers remains irreplaceable.

Reddit launched Reddit Reviews, an AI feature organizing user conversations about ~10,000 products into structured reviews with highlights and key takeaways, rolling out to all US users. It turns casual research into a structured product, and raises the stakes for fake review detection.

Google Fi introduced a “super network” that automatically switches between two carriers for better speed and reception, rolling out first to Pixel 9+ with eSIMs, plus Wi-Fi Auto-Connect+ with automatic encryption. Free for Unlimited Premium and Flexible plans in 200+ destinations.

Finally, Mintlify open-sourced agent-traffic, a zero-dependency classifier that distinguishes agents, AI crawlers, search crawlers, and bots from people, verifying crawlers against operators’ published IP ranges. Install with npx skills add mintlify/agent-traffic. If you run a documentation site, you’ll want this before the bots eat your analytics.

Get the brief

Liked this one? Pick the beats you care about and get just those, five minutes a morning. Free, no hype, one click to leave.

About Me Author

My name is

BriefTechNews

A daily digest of what actually moved in AI, tech, crypto and fintech, assembled and written with AI, with every item linked to its source. Read More
Tags

You May Also Like