AI Weekly Malaysia

Omdia preliminary estimates show global smartphone shipments fell 20% year-on-year in Q3 2026, worse than the 6% drop in the previous quarter and the lowest Q3 since 2013, driven by rising memory and storage costs. Samsung and Apple gained share to a combined 45% (23% and 22%), while Xiaomi, Oppo, and Vivo fell to 29% combined from 35% a year earlier. Budget and lower-margin brands are under the most pressure as RAM costs squeeze cheaper devices.

The Hacker News reports that a critical authenticated remote code execution flaw in Langflow — the open-source low-code builder for LLM apps and agents — was targeted in 1,423 exploit attempts recorded across honeypots. The excerpt provided is truncated to the headline and site boilerplate, so it does not include the CVE ID, affected version range, patch release, or the researcher/vendor attribution behind the honeypot count.

The Hacker News reports that flaws in AhsayCBS — the Ahsay Cloud Backup Suite server — were exploited to deploy web shells, and that new fixes have now shipped after earlier patches proved incomplete. The excerpt provided contains only the headline and site navigation, so no CVE IDs, affected versions, fixed build numbers, or exploitation counts are available from this text.

At its first Config event in India, Figma announced Figma Specs, a Markdown-based editor where teams define a design language that both Figma's own agent and third-party agents can reference, with support for embedding FigJam diagrams, code layers, design libraries, code libraries and guidelines. The code layers feature, announced in private beta in June, moved to public beta, letting users ask the agent to turn a design into code or modify code to iterate on a prototype. Figma also added a Payload headless CMS integration, and said its on-canvas agent was made available to a wider user base in the past week.

Anthropic reportedly told investors its third-quarter revenue hit $16.8 billion, up from about $11.6 billion in Q2 and $4.7 billion in Q1, ahead of an IPO reportedly slated for November. The company also said it was profitable on an adjusted basis and cash-flow positive, generating $1.70 in revenue for every dollar spent on computing power, according to the Financial Times. Investors expect Anthropic to reach $110 billion in annualized revenue by year-end, per The Wall Street Journal, with the figures attributed to people familiar with the matter.

BYD Malaysia confirmed Sime Motors' Inokom Corporation as its local CKD (completely knocked-down) assembly partner, formalised via an MoU signed at BYD's Shenzhen headquarters. Inokom's Kulim, Kedah plant already assembles fully electric models including the Chery Omoda E5 and BMW i5, and BYD says it dropped an earlier plan for a dedicated plant in Tanjung Malim because of recent changes to Malaysia's CBU and CKD EV policy frameworks. The announcement names BYD Malaysia MD Jacob Ma, Sime Group CEO Jeffri Salim Davidson, and BYD's Liu Xueling, but does not say which BYD model will be assembled locally first or when production starts.

Ubisoft plans to cut 70 jobs at its Singapore studio, which its website listed as employing over 450 people in 2023, as it pushes back toward profitability. The company posted a net loss of €1.5 billion (S$2.15 billion) for FY2025-26 and now targets at least €200 million in additional spending cuts over two years, on top of €300 million already targeted, and has proposed closing its Finnish studio RedLynx (150+ employees in 2021). Assassin's Creed Black Flag Resynced, a July release and recent project of the Singapore subsidiary, has sold more than 3.5 million copies.

Samsung hosted regional media at 'The Brief' in Hanoi, Vietnam, showing three commercial products at its new Business Experience Studio: an 18kg commercial washer (36-minute cycle, Bubble Wash, VRT+ vibration reduction, on-device Manager Mode, built-in Wi-Fi fleet monitoring) paired with a 14kg vented dryer (45-minute heavy loads), and industrial-scale HVAC built on Samsung's acquisition of European HVAC specialist FläktGroup, explicitly aimed at high-heat environments like data centres. Bernard Ang, described in the article as Regional Head of Home Appliances and Vice President at Samsung Electronics Southeast Asia and Oceania, framed the push around AI-enabled products and reliability for homes and enterprises. The excerpt cuts off mid-sentence in the HVAC section, and no prices, Malaysia availability dates, or specifications for the third solution are given.

Aon projects Singapore salary growth of 4.1% for the year, roughly in line with earlier estimates from Mercer, WTW, Korn Ferry and Robert Walters at around 4%. If that holds, Singapore's gross median monthly income from work — reported by MOM at S$5,775 in 2025 — could cross S$6,000, meaning half of full-time employed residents earn at or above that. The article then breaks down why gross ≠ take-home: employer CPF (17%, about S$872 on S$6,000) comes off first, followed by employee CPF (20%), before any cash reaches a bank account.

Tool: LaTeX renderer Inspired by Paul Ford and Aboard Foundary's new web app Tiny Little Office, which provides a suite of web app content editors (a Word Processor, Outliner, Presentation Builder, Spreadsheet, Calculator, Drawing Tool, Diagram Maker, Animator, Calendar, and Trip Planner). Tiny Little Office's signature trick is to store all of the document data in the #fragment section of the URL, so every document created using it becomes a bookmarkable and shareable link. It also has an llms.txt file telling agents how to construct URLs that will render as documents. Claude used that file to generate this link to a deck that explains how the URL format works: Tiny Little Office inspired me to have Claude Code (Opus 5.5 Extra) build a similar tool that works with LaTeX. Here's the sequence of prompts I used: I want a tool where you can link to it with a fragment hash #code-here and it treats that code as LaTeX and renders it really effectively. It should also support a prefix which means the latex is compressed with a custom compression dictionary optimized for latex. Use the best available latex render library from a CDN (consider WebAssembly too). If loaded without a fragment hash show an editor you can paste latex into and then toggle to a "rendered" view. That rendered view should be bookmarkable. The #fragment view should let you switch to edit source too. Since it was working in my tools repo I told it to take inspiration from the <iframe sandbox="..."> mechanism used by my existing SVG in Markdown tool as protection against XSS risks: Safe from XSS? We could run in an iframe sandbox like we do on the SVG markdown render project I had it copy the Gist/URL loading feature from that tool as well: We should have an option to load latex from a URL (or gist) similar to the markdown SVG tool Here's an example link. The tool uses Pako 2.1.0, MathJax 4.1.3, faster-latex 0.3.1, and DOMPurify 3.4.16. Tags: latex, paul-ford, tools, urls, ai, generative-ai, llms

At TechCrunch Disrupt, Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini described building AI that designs chips, learns from each iteration, and improves subsequent designs to automate and accelerate chip design. Goldie said compute demand could grow at least 10,000x over the next five years—which she called potentially conservative—and argued the response should be more efficient chips rather than 10,000x more data centers. They cited AlphaChip, which they previously developed, as producing an unconventional donut-shaped chip layout that outperformed traditional layouts.

At TechCrunch Disrupt 2026 on Oct. 14, Flock founder/CEO Garrett Langley defended Flock’s license plate cameras and recording devices amid backlash, saying “we shouldn’t pause progress” and framing privacy as a trade-off for lower crime. The report says the criticism has led to cancelled contracts and layoffs, while Langley largely responded with anecdotes from police chiefs and city leaders rather than tangible solutions. It explicitly compares the Flock backlash to AI backlash over consent and sharing data with AI agent makers.

TechCrunch Disrupt 2026's Startup Battlefield narrowed 20 pitching startups down to five finalists across four categories — space and physical AI, energy and manufacturing, health tech and life sciences, and enterprise and security — with the finals at 11:30 am Pacific on Thursday and a $100,000 equity-free prize plus the Battlefield Cup for the winner. The named finalists in the excerpt are Apate.ai, which runs thousands of AI agents posing as scam victims (different ages, genders, accents, languages) to answer scam calls, follow phishing links and infiltrate social media groups; Flox Intelligence, which uses motion-triggered edge devices running on-device wildlife models to emit species-appropriate deterrent sounds; and Intero Biosystems, which is growing human intestinal organoids to study drug responses.

Lowyat.NET reports, dated 15 October 2026, that the government is considering an update to the electricity tariff mechanism next year. The excerpt supplied contains only the headline and site navigation — no figures, rates, effective dates, or named officials appear in the text. Beyond the title's claim, nothing concrete about what would change is available to summarise.

OpenAI published a customer story on Sullivan & Cromwell's Agreement Analyzer, a legal review tool for draft agreements. Partner John Savva built it first as a single-user proof of concept using Codex, then worked with OpenAI to improve output quality, add parallel processing, and make it usable by multiple lawyers at once. The tool combines firm precedents, negotiating playbooks, and ChatGPT Enterprise, runs multiple review cycles from different perspectives, and returns a memorandum plus proposed tracked changes that a lawyer can review, adapt, or reject. The article is a vendor case study with no accuracy numbers, pricing, latency, or independent evaluation, and the text is cut off mid-sentence in the section about making sources and uncertainty visible.

Anthropic says Code Review is now generally available for Claude Team and Enterprise organizations after leaving research preview. The announcement claims most reviews now finish about 3x faster (~20 minutes down to 6-10 minutes), adds per-repository Quick/Standard/Deep review levels, and requires admins to enable it in Claude Code settings and install the GitHub App so reviews run automatically on new PRs. Eligible Claude Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027, with credits expiring 30 days after claiming.

Anthropic says it has begun giving a team from METR employee-like access—company hardware, office space, and internal tools—to evaluate its frontier models, and METR will publish its risk-assessment results. Anthropic is also working with SecureBio and Accenture as evaluators and is in early discussions with Hugging Face, after signing the White House Accord on Super Intelligence. The post lays out its criteria for independent evaluators: frontier-model expertise, no equity or financial dependence on Anthropic, deep access to training and safeguards, and freedom to publish assessments.

Meta released Muse Spark Open, an open-weight multimodal mixture-of-experts model on Hugging Face with roughly 816B total parameters, ~42B active per token, and a context window up to 1M tokens. It accepts text, images, audio, and video as input but outputs text only, and Meta says it was trained for agentic coding tasks like multi-file refactors, long debugging sessions, tool calling, and browsing with images. Checkpoints ship in BF16 and NVFP4 formats under Muse-Spark-Open-816B-A42B, with 256 experts per MoE layer and 8 selected per token; it shares a tokenizer and chat template with the smaller dense Muse Glimmer released in August.

Black Forest Labs released FLUX.3 Image Klein, and Hugging Face shipped Diffusers support on day one. One model handles text-to-image, multi-reference image-to-image, and a new layout control where you write bounding boxes into the prompt to place image parts or move regions of a reference image. Two open checkpoints are published: black-forest-labs/flux-3-image-klein (distilled, 16 steps, no CFG) and flux-3-image-klein-base (50 steps, CFG), with the distilled version generating a 1024x1024 image in about 2 seconds on an H100. NVFP4 and FP8 weights load via from_single_file for smaller GPUs, LoRA training scripts exist for text-to-image and image editing, and it ships as a Modular Diffusers pipeline. It needs the latest diffusers and transformers (for the Qwen3-VL text encoder).

Gudea, founded in 2023 by Keith Presley, Jonathan Sperber, and James Thomin, pitched on the Startup Battlefield stage at TechCrunch Disrupt 2026 (Oct 14) as a "storm tracker for the internet." It claims to analyze behavior, relationships, and information pathways across more than 500 mainstream and fringe sites and platforms — including X, Reddit, and TikTok — to show who is driving a conversation and how fast a story is spreading, rather than just counting mentions, hashtags, and impressions. The article is pitch coverage: no pricing, funding, API, model details, or customer results are given.

Liquid AI released LFM2.5-24B-A2B, a 24B-parameter reasoning model with 2B active parameters per token, open-weight on Hugging Face, positioned for on-device agents and the largest/final model in the LFM2.5 family (230M to 24B-A2B). It reports up to 1.7x faster on-device decode than comparable MoEs — 184 tokens/s on an Apple M5 Max and up to 21K output tokens/s on a single NVIDIA B300 — and claims competitiveness with Qwen3.6-35B-A3B (roughly 50% more total and active parameters) on 5 of 8 agentic/tool-use benchmarks, including LiveClaw 48.09 vs 38.19 and AutomationBench 27.69 vs 8.72. Changes over LFM2-24B-A2B: context extended 32,768 to 131,072 tokens, vocabulary doubled 65K to 128K for better non-Latin tokenization, pretraining scaled 17T to 33T tokens, and a five-stage post-training pipeline that ran Agentic RL inside harnesses like OpenClaw, Hermes Agent, Pi and OpenCode.

Intero Biosystems, a Denver-based startup spun out of the University of Michigan, sells GastroScreen — human intestinal organoids, lab-grown gut tissue about one millimeter across, shipped roughly 100 at a time in a tube — to pharmaceutical, biotech, food, and chemical companies as a supplement or alternative to animal testing. It also sells the supporting supplies (nutrient mixes that feed the cells, chemicals used in experiments) and runs studies for customers. Co-founder and CTO Madeline Eiken says the organoids contain multiple cell and tissue types found in the human intestine, arguing that most drugs fail clinical trials and that animal models and flat cell layers are not human-relevant.

CNBC's Jim Cramer used his Oct 14, 2026 'Mad Money' segment to argue against selling into the selloff, after the S&P 500 fell just over 1%, the Nasdaq dropped 1.9% and the Dow closed slightly lower despite strong earnings. He cited uncertainty over inflation, interest rates and the war with Iran as reasons the next move is nearly impossible to call, and pointed to a JPMorgan finding that $10,000 invested in the S&P 500 between 2005 and 2024 would have ended up with less than half the wealth if it missed the 10 best trading days. The article also references ASML but the excerpt cuts off mid-sentence before Cramer's next point.

Tiny Little Office is a browser-based suite of small tools — Word Processor, Outliner, Presentation Builder, Spreadsheet, Calculator, Drawing Tool, Diagram Maker, Animator, Calendar and Trip Planner — posted to Hacker News, where it drew 406 points and 62 comments. The submission text points at source files (site/site.ts, tinier/cards.ts) and at an /llms.txt that describes how to make a document as a link, with documents embeddable inside one another as a 'fence named by its app'. The excerpt itself is fragmentary and gives no pricing, release version, or architecture detail beyond those references.

CedarDB, a Munich-based database engine spun out of the Technical University of Munich and built on the Umbra research project, is pitching a single engine that handles both transactional (OLTP) and analytical (OLAP) workloads. Five Umbra researchers are cofounders, and cofounder/CEO Moritz Sichert says the team deliberately made CedarDB PostgreSQL-compatible so customers don't have to rebuild their apps, and because Postgres is the database logic LLMs are most familiar with from training. The excerpt gives no funding amount, pricing, benchmark numbers, or customer names — it's a startup profile, not a measured comparison.