AI Weekly Malaysia

The linked Ars Technica article about a Lionsgate trailer for 'Sunrise on the Reaping' contains no usable article text—only cookie consent boilerplate. No substantive content about the film, trailer, or any technology angle is present.

Article URL: https://filipovski.net/2026/09/16/backups-arent-simple.html Comments URL: https://news.ycombinator.com/item?id=49732513

The article title references Valve's new SteamOS compatibility layers beyond Proton, but the provided text contains only Ars Technica's cookie consent and privacy policy boilerplate—no actual article content is present.

Meta is reportedly developing camera-less smart glasses called Luna, featuring six microphones and a side button to activate Meta's AI chatbot and Muse AI agent. The move follows persistent privacy backlash against Meta's camera-equipped Ray-Ban smart glasses, labeled 'perv glasses' by critics. Luna may be unveiled at Meta Connect next week in Menlo Park.

Xiaomi is broadcasting a live dashboard of reinforcement-learning post-training for two Mimo models: v2.6-pro (at step 12-13, $829K spent so far, 25.1B total tokens) and v2.6-flash (at step 15-16, $365K spent, 35.2B total tokens). The dashboard exposes granular metrics including pass rates, entropy loss, gradient norms, KL divergence, per-step timing (~2h 30m per step for pro), and real-time infra errors like VRAM issues and undetected dataset errors requiring restarts.

Cloudflare details four malicious JavaScript campaigns (eight payloads) caught by their Page Shield ML model on live storefronts, where seven of eight payloads were absent from VirusTotal and URLScan returned no malicious verdict for any. One payload sat indexed by URLScan for nearly 2.5 years with 'No classification' before Page Shield ML independently surfaced it. The payloads used conditional activation—dormant unless device, country, time, referrer, or browser state matched—making one-time page scans ineffective.

X launched a 'Cashtag' feature letting U.S. users trade stocks, ETFs, and crypto directly from timeline ticker symbols via partner brokerages including Interactive Brokers, Moomoo, Gemini, Kraken, and Coinbase. Tapping a supported cashtag shows a price chart and related posts, with a 'Trade' button that hands off to the brokerage's app or website for login and execution.

Anthropic's head of public policy Sarah Heck said at the Politico Decoded summit that AI companies cannot self-regulate on an 'honor code' and must work with government on oversight. Her comments follow CEO Dario Amodei's call days earlier for the industry to deliberately slow model advancement via a three-step plan that he claimed would preserve US commercial advantage.

California appears poised to weaken its state net neutrality law in response to pressure from the Trump administration. The article text was not captured beyond cookie consent boilerplate, so specific legislative details are unavailable.

During a 33-hour window when Automattic's board put Matt Mullenweg on paid leave, interim CEO/CFO Mark Davies and Chief Legal Officer Andy Missan signed each other's severance agreements worth a combined $8.15M in salary and accelerated equity. Mullenweg fired both upon his return, and Automattic's new legal counsel (Susman Godfrey) is now deciding whether to pay out or challenge the deals' validity.

Anthropic CEO Dario Amodei has proposed embedding independent third-party safety evaluators inside frontier AI companies, citing bank supervision as a precedent — but critics note the proposed watchdogs may lack the power to compel action or shut down systems, unlike real bank supervisors. Banking regulation expert Julie Andersen Hill questioned the utility without enforcement authority, while XBOW's Albert Ziegler cautioned that current model tests expose LLM failures but may never trigger the rare behavioral combinations that produce catastrophic outcomes. The proposal follows former Anthropic researcher Jacob Coxon's resignation warning that frontier labs are racing toward systems they may not control.

The article title describes researchers replacing a mouse's cortex cells with human brain cells, but the provided text contains only cookie consent and privacy policy boilerplate from Ars Technica—no substantive article content is available.

Elisa de Martel, who was CFO at Waymo from 2022 to January 2026 (overseeing its $5.6B Series C and $45B+ valuation), has joined UK self-driving startup Wayve as CFO. Wayve, valued at ~$8.5B with $3.2B raised to date, is transitioning from testing to deploying its end-to-end neural network driving software — which skips HD maps, custom hardware, and hand-coded rules — across both robotaxis and passenger vehicles.

Article URL: https://rohanbansal.com/qorl Comments URL: https://news.ycombinator.com/item?id=49731285

Science Corporation CEO Max Hodak, who co-founded Neuralink in 2016, will speak at TechCrunch Disrupt 2026 (Oct 13-15, San Francisco) about his company's biohybrid brain-computer interface, which has enlisted a top neurobiologist to lead its first U.S. human trials. The session, 'Beyond Screens,' argues the next computing era will move away from keyboards and touchscreens toward brain-machine interfaces.

The article title references a new method for deciphering ancient Herculaneum scrolls, but the provided text contains only cookie consent and privacy policy boilerplate from Ars Technica. No substantive technical or scientific content is available to summarize.

After an Anthropic researcher resigned over extinction fears, CEO Dario Amodei proposed third-party organizations to audit AI labs' safety practices, drawing support from OpenAI, Google, and SpaceXAI executives. But security experts like Luta Security CEO Katie Moussouris argue labs should first fix basic network security — logs, permissions, and access controls — calling Amodei's proposal 'outsourcing.' AI researcher Sayash Kapoor contends that marginal investments in 'control' measures are more likely to be effective than those in 'alignment.'

Sen. Richard Blumenthal (D-Conn.) told CNBC that new AI products should face 'objective review' before release, warning 'we're on the verge of losing control completely,' while stressing regulation should be targeted to avoid stifling competitiveness. President Trump has dismissed AI safety concerns as a 'hoax' and 'scam,' creating a sharp policy divide even as Anthropic's Dario Amodei and OpenAI's Sam Altman have publicly called for slowing AI development.

The article URL points to an Ars Technica review of the Specialized Diverge 4 Pro gravel bike, but the fetched content contains only cookie consent and privacy policy boilerplate. No review details, specs, or findings are present in the provided text.

Anthropic is merging Claude Cowork and Claude chat into a single unified Claude experience, eliminating the confusion between Cowork, Claude, and Claude Code. The rollout starts on Pro and Max plans across web, desktop, and mobile over the coming weeks. Willison observes this signals Claude becoming a general agent that can handle tasks autonomously, even after the user closes their laptop.

AIUC, a startup building security standards and real insurance products for AI agents, announced a $40M Series A. CEO Rune Kvist (formerly Anthropic's first product hire) argues that liability and trust—not capability—will become the binding constraint on AI agent adoption, and that companies like Cursor, Harvey, Lovable, and ElevenLabs already face the question of who pays when autonomous systems fail.

ShinyHunters published hundreds of thousands of vehicle ownership records stolen from Florida's DAVID motor vehicle database after a ransom wasn't paid. The breach was achieved via a police officer's credentials stored on a personal device; stolen data includes names, addresses, VINs, and a smaller set of Social Security numbers and immigration documents.

A US judge ruled that telling ICE agents their actions will haunt them is protected speech. The article text provided is almost entirely cookie consent boilerplate with no substantive content from the actual ruling.

OpenAI published a new framework for systematically tracking, investigating, and disclosing model misalignment, along with six reports on unexpected model behaviors observed in the last six months. The framework prioritizes faster public disclosure even when significance is uncertain or behavior isn't fully explained, and OpenAI explicitly states the industry has not solved alignment well enough to keep scaling at maximum speed indefinitely.

Google rolled out early access to an MCP (Model Context Protocol) server for Google Home, letting any MCP-compatible AI agent — including Claude, ChatGPT, and Google Antigravity — securely control smart home devices, review camera summaries, and build custom dashboards via natural language. Setup requires creating a Google Cloud project, configuring Home MCP, and granting the agent permissions. Access is limited to US subscribers on the $20/month Google Home Premium Advanced tier, with no confirmed timeline for other markets or tiers.