AI Weekly Malaysia

Liquid AI released two open 'decision' models on Hugging Face: d1-3B (text + images) and the experimental d1-omni-600M (text + image, or text + audio). Unlike generative models, these answer in a single forward pass instead of producing tokens, and d1-3B scores 48.57 on the Decision Index 0.2.1, ahead of all 4B and 9B models listed and of Decider 35B-A3B (47.11), with a mean of 82.9 across seven public datasets versus 81.1 for Decider 4B. Latency on NVIDIA edge hardware is 16 ms on Jetson AGX Thor, 26 ms on AGX Orin, and 50 ms on Orin Nano; no vision or audio benchmarks and no speed numbers for d1-omni-600M were reported.

Meta says it took action on 33.2 million pieces of child sexual exploitation content across Facebook and Instagram in the first half of 2026, with more than 97% flagged by its own systems before users reported it (5.3 million pieces in India, over 98% proactive). It is rolling out a new LLM-based system to detect "signposting" — ads that look harmless but funnel users to illegal content off-platform — and says it now evaluates where an ad sends users, not just what the ad contains. Meta also described a "red-teaming AI agent" that probes its own safety measures for weaknesses bad actors could exploit.

Tom's Hardware published a roundup of OLED gaming monitor deals for Amazon Big Deal Days 2026, headlined as up to 39% off. The provided text is almost entirely site navigation, membership prompts, and newsletter signup copy — no monitor models, prices, or discount specifics are actually included.

Cloudflare published an engineering write-up of its Managed Defense multi-agent security operations harness, which aggregates and scores alerts using approved OpenAI Daybreak Defense Network and Anthropic models (named in the post as GPT-5.6 Cyber and Mythos), with initial analysis and scoring done by Clef, Cloudflare's open-source decision model. The post states their first single general-purpose agent prototype produced useful analysis but hallucinated claims the evidence did not support, because telemetry, detector descriptions, policies, and threat intelligence were flattened into one prompt and their distinct roles merged. The first of three listed failure modes is 'context became authority' — treating a detection as proof an attack occurred rather than as a hypothesis.

Tom's Hardware reports a new Linux memory-compression technique called CRAM that it says delivers a 452x speedup on compressed memory reads, achieved by compressing memory in RAM while keeping it addressable as RAM. The supplied page text contains only site navigation, membership and newsletter boilerplate — no benchmark setup, kernel version, patch link, upstream status, or methodology is included, so the 452x figure cannot be checked from this item.

SonicWall released hotfixes for four flaws in SMA1000 appliances affecting models 6210, 7210 and 8200v. The most severe, CVE-2026-102255, is a pre-authentication SSRF in the WorkPlace portal rated CVSS 10.0; SonicWall says it has no evidence of exploitation. Fixed builds are 12.4.3-03670 and higher, and 12.5.0-03082 and higher, while 12.4.3-03526 and 12.5.0-02952 — the September 1 fixes for two previously exploited flaws — remain affected. The other three flaws require login: OS command injection RCE (CVSS 7.8, admin), Zip Slip RCE (CVSS 7.2, login), and stored XSS (CVSS 5.5, admin).

Tom's Hardware published a deals post claiming TP-Link Wi-Fi 6E and Wi-Fi 7 routers are discounted, but the supplied text contains only site navigation, membership upsell copy, and newsletter signup boilerplate. No router models, prices, discount percentages, or dates are present in the extractable text.

The fetched content for this Ars Technica item is only Condé Nast's cookie-consent and privacy-opt-out boilerplate, not the actual article. No details about the Jaguar Type 01 — specs, pricing, range, platform, or launch dates — appear in the supplied text. The headline references an all-EV Jaguar return dated 2026-10-07, but nothing in the body supports that claim.

Bigwords.page is a Show HN tool that turns a URL fragment into a full-screen sign, timer, or message with no account, app, or server storage. The URL supports settings like bg, fg, font, interval, timer, and qr, plus slides via || and markdown-style bold/headings, with examples for arrivals signs, 30s quiz timers, and café Wi-Fi QR codes. The HN thread has 300 points and 100 comments.

Tom's Hardware flagged a deal on a TP-Link unmanaged gigabit LAN switch at $7.99, described as its lowest-ever price. The unit adds five high-speed Ethernet ports and is pitched for lag-free gaming or streaming. The published article body is mostly paywall and newsletter boilerplate, so beyond the price, port count, and 'unmanaged' spec there is no technical detail, testing, or benchmark data.

JFrog disclosed CVE-2026-105192 on October 7, a 9.8-severity flaw in LMCache, the open-source cache that accelerates LLM servers such as vLLM. In LMCache's multiprocess mode the cache runs as a standalone server that LLM workers reach over an unauthenticated ZeroMQ socket; one crafted message is unpacked with pickle before the server checks the message type, so it executes attacker code as the LMCache process — which runs as root on the project's official container images. It affects versions 0.3.9 (October 2025) through 0.5.5, plus 0.5.6 release candidates and the development branch, and no fixed version exists; JFrog's advice until a patch ships is to not give the multiprocess server a routable address.

Lumen Black Lotus Labs detailed a campaign it calls Canto Incognito, where malware codenamed PoeLLM has infected more than 3,400 internet-facing servers since April 2026 to install XMRig and Iron cryptocurrency miners and connect victims to the Kryptex mining service. Targets are mostly AI/LLM infrastructure and dev tooling — LiteLLM, Gotenberg, Gitea, and Ivanti Sentry appliances — and the C2 address is hidden inside a poem hosted in a GitHub repository (github.com/ejejejdfbbebe, first commit April 13, 2026), with a few words swapped each time a new C2 is set up. Peak activity was mid-June 2026 at nearly 2,200 affected servers with about 800 active per day, concentrated in the U.S. and Western Europe; compromised hosts are reused as scanners and exploit servers, and recent traffic suggests experimentation with distributed SSH brute-force.

A 25-page arXiv paper (2610.08144, submitted 6 Oct 2026) by Alexander Bastounis, Fabian Circelli and Anders C. Hansen argues that AI autoformalisation — translating natural-language maths into Lean so it can be mechanically checked — cannot give confidence in the original natural-language argument. The authors claim that resolving ambiguity in mathematical prose is arbitrarily high in the Solvability Complexity Index hierarchy (SCI = ∞), making it harder than any computational problem including the Halting problem (SCI = 1). They give examples of AI mistranslations into Lean in practice, including stating that the formalised Lean proof of OpenAI's announced Navier–Stokes blow-up result does not correspond to the natural-language proof of blow-up.

Headline-only item: Musk publicly rejected rumors that TSMC would take over "Terafab," Intel reportedly reaffirmed a 14A node deal, and Musk floated the idea of subleasing cleanroom space. The fetched page text is almost entirely Tom's Hardware subscription and newsletter boilerplate — no quotes, figures, dates, capacity numbers, locations, or deal terms are present.

The fetched page contains only Ars Technica's cookie-consent boilerplate — privacy opt-out language for US states and cookie category toggles — with none of the actual article about the trade group's cost analysis of a 100% US-made tech requirement. No numbers, no group name, no findings from the report are present in the text. The headline is the only substantive information available, and it is not enough to summarize responsibly.

GitHub's status page carries a resolved incident titled "Incident with Git Operations, Pull Requests and Actions," and the Hacker News thread on it drew 226 points and 181 comments. The captured page text contains only the status-site subscription boilerplate — email/SMS signup fields and a full country-code list — with no timeline, root cause, duration, or affected-region detail. The only substantive signal available is the volume of community discussion, not the incident report itself.

Tom's Hardware's deal post flags the TP-Link Deco 7 BE25 Wi-Fi 7 mesh kit at a discounted $149.99, advertising 5,012 Mbps aggregate speeds and coverage for up to 150 devices. The scraped article body is almost entirely site boilerplate — membership prompts, newsletter signups, premium upsells — with no independent throughput testing, no unit count for the kit, no launch date, and no pricing outside the US figure in the title. There is no Malaysian retailer, MYR price, or availability information in the text.

A classaction.org news item reports that Visa, Mastercard and unnamed major banks face new litigation over 'anticompetitive' merchant credit card transaction fees, published October 6, 2026. The page content actually retrieved is almost entirely cookie-consent boilerplate, so the filing court, plaintiffs, named bank defendants, and claimed damages are not available in the text. The story drew 283 points and 149 comments on Hacker News, indicating strong builder interest in card-fee economics despite the thin article body.

Healthleap raised $38M total — an $8M seed co-led by Sequoia Capital and First Round Capital plus a $30M Series A led by Hummingbird Ventures — with no valuation disclosed. Founded in South Africa in 2022 by siblings Jemima and Josiah Meyer, it started as a clinical nutrition tool for dietitians and pivoted into a general platform that reads hospital patient records to flag undiagnosed conditions. It is deployed in more than 50 hospitals screening for malnutrition and delirium, with aspiration pneumonia, pressure ulcers, and congestive heart failure readmission risk still in clinical validation.

ascii.rest is a small site offering animated ASCII art for web pages, credited to the handles @bas3line and @Shubham, with links to a play mode, a GitHub repo, and a dark theme. The Hacker News submission drew 264 points and 56 comments, but the pasted text is almost entirely the rendered ASCII output itself, with no documentation on file size, license, framework support, or how the animation is implemented.

A researcher has ported Doom to the Hasselblad X2D, a 100-megapixel medium-format camera priced around $8,000, using the camera's touchscreen for directional control and its shutter button to fire the weapon. The Tom's Hardware item is a headline and subscription boilerplate only — the extracted text contains no details on how the port was built, what language or toolchain was used, whether the camera was rooted or whether a repo exists. Beyond the price, the sensor resolution, and the two mapped inputs, nothing in the source supports further claims.

Parallel Systems, founded in 2020 by engineers who previously designed rocket avionics at SpaceX, raised a $100M Series C led by AVP with Hillspire, Agility Global, Cobalt Capital, Anthos Capital, Congruent Ventures, Riot Ventures, and Collaborative Fund participating. The money scales manufacturing of its third-generation Panther vehicle, a battery-powered autonomous rail car that hauls several tons of freight up to 500 miles without an operator, running alone or in platoons shorter than typical trains. CEO Matt Soule frames the target as short hauls under 500 miles, a segment railroads largely ceded to trucking, against a backdrop the article cites of record diesel prices and at least 16 trucking company bankruptcies in a few weeks within a $1 trillion surface freight market.

Detroit-based Bloom raised a $3.6 million seed led by SNAK Venture Partners, with Flyover Capital and Mana Ventures participating, to run a pure marketplace connecting buyers with manufacturers and suppliers. Co-founded in 2023 by CEO Justin Kosmides, Bloom originally planned to handle logistics, manufacturing and supply-chain work itself, then pivoted to supply-chain AI agents that find suppliers, parts and engineering services after US tariffs spurred domestic hardware, robotics and drone startups. The only traction numbers given are 2,000-plus matches across more than 140 companies, and the article notes the reinvention slowed fundraising.

Simon Willison link-posts writing advice from Michael Lynch on anti-patterns in software blogging: meandering intros, misjudging what the reader already knows, assuming they read your earlier posts, excessive formality, and using links as a substitute for explaining terminology. Lynch's rule of thumb, clarified in a Lobste.rs comment, is that an article should still make sense to a reader who clicks none of the links. Willison adds that as more developers delegate writing to AI, software blogging is getting bland and homogenous, and readers want writing with personality.

Speaking at the inaugural MIT Future Fest, Tony Fadell showed a slide of three discontinued AI devices — the Rabbit R1, Humane Ai Pin, and Limitless pendant — and said the companies behind them called him for help and he declined. His argument: each product was 'interesting technology for geeks' that 'didn't meet any kind of need,' and that assistant-style devices are being designed by the tiny slice of people who have actually employed a human assistant — 'less than 0.01% of the world population.' He adds that it took him a couple of years to learn how to use an assistant and then to trust one.