Amazon

Amazon unveiled the Ring Drone Cam and Ring Rover Cam, AI-powered security devices built on an AI Pilot system that combines vision models, on-device sensors and cloud processing for autonomous patrolling and obstacle avoidance. The products are slated to ship from late 2026 to early 2027, and are likely to reignite privacy concerns over indoor flying cameras.

Amazon has introduced a self-flying indoor security drone and a ground rover for its Ring ecosystem, both dispatched automatically to preset waypoints when an alarm trips so video may already be streaming before the user opens the alert. The drone rests in its dock with the camera off, records only during flights, uses deliberately loud propellers as a deterrent, and ships with Ring's strongest default encryption.

Amazon-owned Ring has formed a robotics division, Ring Robotics, unveiling a $270 home security patrol drone and a $1,000 enterprise patrol robot called Rover Cam, both supporting scheduled patrols and 2K video streaming. The move extends Ring from fixed cameras to mobile autonomous security hardware, though no new AI algorithms or autonomy stack details were disclosed.

Ring has confirmed its first indoor flying camera, the Drone Cam, will launch on January 27, 2027, for $249.99, nearly six years after the product was first announced in 2021. The AI-powered device responds to triggers from Ring sensors such as motion, door openings, water leaks, or alarms, flying preprogrammed routes to a set location to livestream and record via the Ring app.

Anthropic has launched its Nasdaq IPO roadshow with a price range of $70 to $78 per share, aiming to raise up to $90.4 billion through public shares and a private placement at a targeted $1.02 trillion valuation. Q3 revenue grew 47% to nearly $17 billion with a narrower net loss, though growth decelerated from the prior quarter. The offering faces headwinds from roughly $900 billion in compute commitments and mounting Wall Street skepticism over AI spending.

Amazon's internal ticketing system swapped Claude Sonnet 4.6 for Strands Decider, a decision model launched a month ago, achieving comparable labeling quality at 98% lower token cost and saving millions of dollars annually. The key insight is reframing open-ended text generation as a constrained decision problem over fixed labels, letting a specialized decision model replace a general-purpose LLM.

An Amazon research team introduces HOPE, a higher-order expert pruning method for Mixture-of-Experts LLMs that removes redundant experts by accounting for inter-expert coupling rather than scoring them independently. The work targets the memory bottleneck where expert MLPs dominate inference footprint: Qwen3.5-122B has 256 experts per layer across 48 MoE layers, with 95% of its ~250 GB parameters in expert MLPs while only 8 experts per layer activate per token.

A new Amazon paper audited nine public rule-following benchmarks such as RuleArena and LegalBench and found that their reference answers and grading code contain errors, causing correct model responses to be marked wrong. In LegalBench's tax scorer, decimal points were stripped, turning $259,487.13 into 25,948,713; after the fix, three models' accepted answers rose from 8-9 out of 96 to 87-91. The authors urge checking scorer code and hand-reviewing failed answers before trusting low scores.

Amazon has assembled a new team to develop an unprecedented home robot that can adapt to changing household environments and operate autonomously around family members. The group draws on former Astro staff and an MIT robotics expert, though product form and capabilities remain undisclosed. Astro itself shipped only to a few thousand invitees and its commercial variant ended within a year, yet Amazon is clearly not abandoning Bezos's long-term home robotics bet.