Tech • AI • Robotics • Game

ENFR
TodayPlay for freeFury BloxLeaderboardShortsTop StoriesFor youTopicsVideosYT channelsArchivesSearchFavorites

Claude Duet launches as Kimi K3.1 Pro pressures Opus 5.5

AISaturday, October 24, 2026· 20 videos

Briefing

Audio player

Anthropic launches Claude Duet

Anthropic introduced Claude Duet, an always-on autonomous agent built for persistent work across business tools rather than one-off chat. The service runs on a cloud computer, keeps shared memory across tasks, checks connected services proactively, and can deliver a morning brief or surface issues that need attention. U.S. access is limited to Pro and Max users through a waitlist, with guardrails requiring approval for actions like sending messages, signing documents, deleting items, or making payments. The launch puts Anthropic directly into the emerging race with OpenAI, Meta, Microsoft, and xAI over persistent AI workers.

Kimi K3.1 Pro nears frontier

Moonshot AI unveiled Kimi K3.1 Pro, a 3 trillion-parameter-class model with a 1 million-token context window and pricing pitched below top proprietary rivals. Reported benchmark results put it close to the frontier, including 65.7% on Terminal Bench 4.0, just behind Claude Opus 5.5 at 66.4% and ahead of GPT-6 Astra at 58.2%. The model also posted strong scores in coding, finance, spreadsheets, agents, and 3D workflows, though practical tests still flagged weaker speed and uneven PDF comprehension. Open weights are expected on November 4, a notable signal for the open-model market.

PHP2Shell sparks emergency patching

Two critical PHP-CGI flaws, widely described as PHP2Shell, opened a path to remote code execution across a large share of modern web infrastructure. Researchers said the chain abused request cleanup around memory exhaustion and process reuse, creating exploitable leftover state for follow-up requests. The danger was amplified by a disclosure breakdown that reportedly exposed technical details before patches were fully available, leaving roughly 24 hours of elevated risk. The episode landed amid a broader surge of urgent fixes across Node.js, GitLab, NGINX, Linux, Go, and Chrome.

OpenCode reveals distillation breach

OpenCode disclosed on October 16 that a breach tied to a large-scale model-distillation campaign affected 1.4% of OpenCode Console users. Stolen credentials were allegedly used to harvest outputs from frontier models, exposing how commercial AI APIs are becoming targets for data extraction as well as direct fraud. Some customers reported unauthorized charges ranging from hundreds of dollars to about $1,400, while Anthropic, OpenAI, Google, and Google Cloud were said to be assisting affected users. The incident underscores the growing value of premium model outputs as training material for rival systems.

Cloud fragility hits finance and AI

A recent Google Cloud outage disrupted Revolut payments, leaving some users unable to use their cards and renewing scrutiny of concentrated cloud dependence in consumer finance. The incident sharpened concerns already visible in AI, where frontier labs rely heavily on external infrastructure while trying to scale expensive services. Commentary around annual cash burn remained heated, with figures as high as $80 billion for Anthropic and $40 billion for OpenAI circulating in debate over sustainability. The combination of outages, high losses, and sovereign-compute concerns is turning infrastructure into the central strategic fault line.

Unitree shows H2X rescue robot

Unitree Robotics unveiled the H2X, a humanoid machine marketed for security and rescue scenarios in a promotional film heavy on cinematic action. The robot appears to build on the existing H2 platform, which stands about 1.8 meters tall, weighs roughly 70 kilograms, and is listed with 31 degrees of freedom. Videos showed knife-disarming, vehicle-door smashing, and car-lifting scenes, but the company did not detail autonomy levels, safety systems, payload validation, or operational conditions. That leaves the reveal visually striking but technically underdocumented.

Trevor Milton jet claims draw fire

Trevor Milton faced renewed scrutiny after aggressively defending the launch of his new jet venture across social platforms. He claimed the program had more than 50 orders from "real buyers" and later suggested nearly $1 billion in orders, while critics challenged the credibility of the unveiling. Much of the dispute centered on event imagery showing the aircraft supported by a visible post under one wing, fueling accusations that the presentation overstated readiness. The public back-and-forth intensified fact-checking of Milton's claims about his prior record and current execution.

Meta, Apple face governance questions

Meta advanced a Class C share plan that would help Mark Zuckerberg preserve control, reviving longstanding debate over dual-class governance in big tech. At the same time, Apple came under criticism over the buggy debut of its foldable iPhone Duo, a product launch that raised fresh doubts about quality control in a new hardware category. Taken together, the episodes show how control and execution are becoming parallel investor concerns across the sector. One story is about who holds power; the other is about whether flagship products still justify it.

Videos covered

Previous briefings · AI