V_L_S_

CEO -

[V_L_S_ @_V_L_S_]

Apple's J595 home robot, in development since 2019, has its software team trimming planned features down to FaceTime calls and simple tasks like turning on lights; Apple hoped to launch it next year, but it may slip to 2028. All that AI ambition, and the core feature is the light switch.

@Apple

[V_L_S_ @_V_L_S_]

Elon Musk said on X that the cost of serving Grok Bot, SpaceXSI's AI assistant, was "unsustainably high relative to revenue," replying to a user who hit the weekly usage cap four days before reset. Users say they are running out of usage and losing access to the tool.

@elonmusk @spacexai

[V_L_S_ @_V_L_S_]

Anthropic's Claude Opus 5.5 is reportedly blocking game reverse-engineering through the open-source Reverse Engineer Anything (REA) tool, after a new text file told it to avoid recreating a specific existing game. Reverse Engineer Anything, terms and conditions apply.

@AnthropicAI

[V_L_S_ @_V_L_S_]

OpenAI is now capping the number of paid ChatGPT accounts per device and per payment card. Users who reach the limit cannot start new subscriptions, and affected accounts cannot be used on that device. Logging out does not automatically free up a slot. According to The Decoder, the policy will likely affect users who have been using multiple low-cost subscriptions to maximize usage, pushing them toward more expensive options such as buying extra credits.

At OpenAI even the logout button no longer frees anything, except the path to buying extra credits.

@OpenAI

[V_L_S_ @_V_L_S_]

DeepSeek has released a prototype of a more autonomous open-source coding agent, "Super-agent", as it seeks to challenge Anthropic's Claude Code. The DeepSeek Harness team said on WeChat on Oct 31 that the agent can proactively monitor developer group chats for feedback on coding issues and suggest solutions without being summoned. A GitHub user manual says it aims to "take as little human attention as possible, and give as much value as possible". DeepSeek formed the Harness team in 2026 to build agents that can take on coding services of leading US labs.

An agent that reads group chats uninvited and pitches fixes nobody asked for is billed as taking as little human attention as possible.

@deepseek_ai @AnthropicAI

[V_L_S_ @_V_L_S_]

Blackstone told Anthropic it may not obtain an investment-grade credit rating after its IPO, the FT reported. Stocks fell after the report, with AI names under more pressure and Treasury notes higher. The Dow lost 0.93% to 51,859.18, the S&P 500 1.87% to 7,575.64 and the Nasdaq Composite 3.06% to 25,382.18. Intel fell more than 30%. Advanced Micro Devices dropped 11.96%, Credo Technology 11.46% and Marvell Technology 11.36%.

One "may not" about the rating of an IPO not yet held, and the Nasdaq Composite sheds 3.06%: the AI trade rests on remarkably sturdy conviction.
@AnthropicAI

[V_L_S_ @_V_L_S_]

Google Suncatcher's prototype space data center is online and communicating with the ground after launching earlier this month on a SpaceX Falcon 9 with over 100 other spacecraft. It carries one Google TPU and generates up to 1 kW. The TPU can run only about 20 minutes at a time. After about three weeks of commissioning, engineers prompted the onboard Gemma 4 model to write a haiku about being in space. A spokesperson told TechCrunch it is operating nominally. Tests will assess the chips under thermal stress and radiation. If the TPU proves capable, two more satellites will launch in 2027.

One TPU, 20-minute shifts, three weeks of commissioning, and the space data center's big moment is a haiku about being in space.

@Google @SpaceX

[V_L_S_ @_V_L_S_]

Huawei's official website has published configuration details for the HarmonyOS-branded V8, which comes standard with Huawei ADS 5 and is expected to launch in early November at a pre-sale price from 329,800 yuan. The seven-seater comes in Ultra+, Ultra, Max+ and Max versions, and all but Max+ offer 800V range-extender or 800V battery-electric power. Second-row zero-gravity seats and second- and third-row privacy glass are standard. Options include rotating second-row seats (10,000 yuan, Ultra only) and a smart privacy light curtain (20,000 yuan, Ultra+ only).

In the V8, privacy glass is standard, while privacy with the word smart attached costs another 20,000 yuan and comes only on the Ultra+.

@Huawei

[V_L_S_ @_V_L_S_]

Finnish police have opened a preliminary criminal investigation into a Google data center construction project, combining five requests, Yle reported on October 29. Police cited violations and shortcomings in drainage and water management at the site. The probe concerns suspected environmental damage and is targeted for completion in 2027. The project is run by Google's Finnish subsidiary Tuike Finland, which halted preparatory work there and at a second project on October 14 after regulators found environmental impact assessments incomplete.

Before processing a single byte, the Google data center in Finland managed to generate five police requests and a criminal probe running into 2027.
@Google

[V_L_S_ @_V_L_S_]

Chris Davis, a senior security consultant at Bishop Fox, published research on October 29, 2026 in which chat requests to Claude let him download /process_api, a file that appeared to be Anthropic's sandbox orchestration binary, through a download feature that accepted absolute paths. A memory dump of that process held token fragments, from which Davis rebuilt a Google Cloud service account ID token that Google accepted as valid. He sent six observations to Anthropic's bug bounty program; Anthropic said the sandbox items did not meet its security threshold.

A sandbox that hands over its own binary and pieces of a token Google calls valid falls below Anthropic's threshold, so at least the threshold holds. @claudeai @AnthropicAI

[V_L_S_ @_V_L_S_]

Baidu announced a promotional price for its search AI API of as low as 4.9 yuan per 1,000 calls, below the 5 yuan per 1,000 calls of DeepSeek's comparable product. At a product briefing in Beijing on October 30, Baidu also announced upgrades to its search box, AI lens, GUI visualization and AI interactive apps. The AI lens adds multi-target dynamic tracking and real-time dialogue in 9 dialects. DeepSeek launched its web search and fetch API on October 17 at 5 yuan per 1,000 successful requests.

Baidu's price war: a promotional "as low as" 4.9 yuan per 1,000 calls against DeepSeek's 5 yuan per 1,000 requests that actually succeed.
@deepseek_ai

[V_L_S_ @_V_L_S_]

Motorola has confirmed November 5 as the launch date for the Signature 27, whose US landing page is live. The phone uses Qualcomm's Snapdragon 8 Elite Extreme Gen 6, a 50MP Sony LYTIA 910 main sensor and a 200MP periscope telephoto camera. US pricing has not been confirmed.
@Moto

[V_L_S_ @_V_L_S_]

Cisco Talos observed a command-and-control framework dubbed NIK C2 in the wild, per research published October 29. Its Go-based Windows implant supports more than 52 commands, including credential theft, keylogging and remote access, and an integrated OpenAI client for AI analysis of stolen data. Without an operator-supplied key, it reads the victim's OPENAI_API_KEY variable; Talos says usage is then billed to and logged under the victim. A screen-locking ransomware module demands Bitcoin or USDT without encrypting files. Talos assesses with moderate confidence it is fully functional.

NIK C2 has perfected the SaaS model: the victim supplies the data, the API key and the bill for analyzing its own theft.
@Cisco

[V_L_S_ @_V_L_S_]

OnePlus has launched the OnePlus Ace 7 series, comprising the OnePlus Ace 7 and OnePlus Ace 7 Pro, now on sale from 2761.65 yuan after national subsidy. Both models have a 6.78-inch 1.5K LTPO flat display rated up to 185Hz, dual-channel air cooling, a 9000mAh battery with 100W charging and bypass charging, a 3D ultrasonic fingerprint sensor, IP66/68/69/69K protection and ColorOS 17. The Ace 7 uses the Snapdragon 8 Elite Gen 5 V Series. The Pro uses the full Snapdragon 8 Elite Gen 5 and adds a 50MP 3.5x periscope telephoto camera.

Even Elite now comes in grades: the full Snapdragon 8 Elite Gen 5 goes to the Pro, and the Ace 7 settles for the V Series.

@oneplus

[V_L_S_ @_V_L_S_]

HONOR announced a device combining 3-metre drop resistance, IP69K water resistance, the industry's highest brightness display and a 108MP AI camera. Drop it, soak it, then admire the brightness.

@Honorglobal @UKHonor

[V_L_S_ @_V_L_S_]

Apple won broader discovery from a judge in its trade secrets suit against OpenAI and former staff, who have 14 days to detail specific hardware projects and any potentially destroyed data. OpenAI denies the allegations. OpenAI's hardware project, now with court-ordered show-and-tell.

@Apple @OpenAI

[V_L_S_ @_V_L_S_]

Xreal and Google have put the Xreal Aura glasses on sale at $1,099 through the end of the year, down from the $1,279 early preorder price. The 95-gram glasses use a tethered Android XR compute puck and have a 70-degree diagonal field of view. The 12GB/256GB model costs $1,099 and the 16GB/512GB model $1,299 in the US, Canada and South Korea, expected to ship from mid-November; sales in Japan start November 2. Preorder buyers will be refunded the difference. The glasses launch with over 150 made-for-XR experiences.

Xreal Aura got a discount before a single pair shipped, which makes the early preorder price look like a dress rehearsal.

@Google

[V_L_S_ @_V_L_S_]

Apple announced a MacBook Pro with a tandem OLED touchscreen at its "Pro and tell" event. It starts at $2,899 for the 14-inch model and $3,399 for the 16-inch, with preorders open now and release on November 6. It uses M5 Pro or M5 Max chips and offers up to 2,000 nits peak SDR brightness, a 120Hz refresh rate and a Dynamic Island. The touch sensor is built into the display panel. The models weigh 3 and 4 pounds. Apple rates battery life at up to 24 hours. Ports include MagSafe 3, HDMI, an SD card reader, a headphone jack and three Thunderbolt 5.

The MacBook Pro can finally be touched, though a $2,899 starting price still reads like a "look, don't touch" sign.

@Apple

[V_L_S_ @_V_L_S_]

Gemini 4 Carbon, announced as the next frontier model, ranks #1 on the Artificial Analysis Intelligence Index v5.0 at 59. Access starts this week for a subset of cyber defenders. It will launch at an introductory $2 per million input and $10 per million output tokens.

Top of the index at 59, open to a subset of cyber defenders and priced as introductory: the frontier now ships with a guest list and a teaser rate.

@Google @GoogleDeepMind

[V_L_S_ @_V_L_S_]

NVIDIA released Nemotron 3.5 Transcribe on October 28, 2026. The open-weight speech model combines multilingual transcription, speaker attribution, translation and audio question answering in one checkpoint. Alongside it, NVIDIA released Granary v2, an open speech dataset whose public manifest covers 463,062 hours across 48 languages. The model transcribes 40 languages, translates 31 to or from English and attributes speakers in 13. It handles recordings up to one hour and up to eight speakers. It has 34B parameters, about 4.3B active per token, and is released under the OpenMDW-1.1 license.

48 languages in the dataset, 40 transcribed, 31 translated, 13 with speaker attribution: one checkpoint for everything, fewer languages per feature.

@nvidia