AI Weekly Malaysia

Hugging Face ran a 19-day hackathon (July 15–Aug 2, 2026) where 1,200+ participants used coding agents (Claude Code, Codex, Cursor, OpenResearch's orx) to reproduce ICML 2026 papers claim by claim, producing 6,816 Trackio logbooks covering 2,226 of the conference's 6,352 accepted papers (~a third). ICML 2026 saw 23,918 submissions and 6,352 acceptances, roughly double the prior year, partly attributed to AI agents accelerating experiment cycles and writing.

Why: If you build or rely on ML research outputs, this signals that a meaningful fraction of top-conference papers may not have been rigorously verified by reviewers—one spotlight paper's proofs went unchecked despite strong scores. Coding agents can now attempt full reproductions in an afternoon that would cost a human reviewer a weekend, so teams shipping ML should consider running agent-based reproduction checks on papers they depend on rather than trusting acceptance as a quality signal.

UK-based developer Alex Morley-Finch built Marlin, an open-source personal search engine that indexed ~560,000 homepages for ~$10 in cloud GPU time using under 1GB of storage. It uses a small OpenAI-compatible local LLM to generate summaries, categories, and tags for each page, with four components: a fetcher, a worker, a steward for filtering bad pages, and an API with web UI. His first crawl went wrong—90% corporate sites and docs—so he added a weighting system to prioritize pages he actually cared about, like portfolios and indie projects.

Why: If you want a search index scoped to your own interests—e.g., Malaysian indie dev portfolios, local startup pages, or niche technical blogs—this shows it's feasible for ~$10 and a weekend, with a crawl-weighting approach to avoid drowning in generic corporate content. The main unresolved pain point is LLM-generated tagging quality, which is worth discussing before you replicate the architecture.

A critical vulnerability dubbed 'Zoomsday' allowed anyone in a Zoom meeting to take over another participant's entire device. AI-assisted security research reportedly needed only 20 prompts to discover a working exploit potentially affecting hundreds of millions of Zoom users.

Why: If you run distributed teams or customer calls on Zoom, patch immediately and treat any unpatched client as a remote-code-execution risk. The 20-prompt discovery angle means AI tooling is materially lowering the cost of finding high-impact exploits in software your team already ships with—factor this into your security review cadence, not just your Zoom update schedule.

Nathan Lambert reflects on writing an AI textbook and argues that LLMs remain stagnant at long-form non-fiction writing, increasing entropy rather than compressing knowledge into insight. He contends that if models can't organize and present established science, they're not ready to autonomously solve open-ended scientific problems, and that progress will look more like low-hanging fruit and cross-field connections than revolutionary breakthroughs.

Why: If you're building AI agents for research, technical writing, or autonomous knowledge work, this argues against assuming models will soon self-organize complex information into coherent long-form output. Plan for human-in-the-loop structuring and editing rather than end-to-end autonomous generation for anything requiring sustained argument or knowledge compression.

Anthropic announced that an unreleased model made progress on the Riemann hypothesis by increasing the lower bound of solutions for which it holds true. A non-mathematician staff member prompted the model to attempt the problem, then the model autonomously coordinated 60 sub-agents over 1.5 days, testing 650 ideas and spending 31 million tokens. Two sub-agents developed the key mathematical ideas, 13 contributed supporting ideas, 30 failed to develop new ideas, 13 validated, and 2 wrote the paper; results were confirmed via the Lean proof assistant.

Why: The concrete takeaway for builders is the multi-agent orchestration pattern: a single prompt spawned 60 sub-agents with distinct roles (generators, validators, writers) that ran autonomously for 1.5 days at 31M tokens. If you build AI agent systems, this is a working blueprint for decomposing hard open-ended tasks into specialized agent roles with built-in validation — though the cost profile (31M tokens for one problem) sets realistic expectations for what autonomous agent swarms actually consume.

Rapid7 researchers disclosed an unauthenticated RCE chain in on-premises Microsoft SharePoint (CVE-2026-55040 CVSS 9.1 for identity bypass, CVE-2026-63520 CVSS 8.1 for RCE via unsafe .NET type instantiation in Business Connectivity Services), affecting SharePoint Server Subscription Edition, 2019, and 2016, plus Project Server 2013 SP1 and Office Web Apps 2013 SP1. A significant portion of the vulnerability research was performed by an AI agent. SharePoint Online is not affected, and the July update breaks the chain while the August fix build numbers are not yet public.

Why: If you run on-premises SharePoint, confirm the July update is installed immediately and watch for the August package—Rapid7 says the chain is fixed but Microsoft had not yet published the patched build numbers at disclosure time. For everyone else, the notable detail is that an AI agent materially contributed to finding a CVSS 9.1 exploit chain, which signals that AI-assisted security research is producing real, high-severity results rather than toy demos.

Researchers from the University of Birmingham and Fuzzware presented a toolkit called CATANA at USENIX WOOT that exploits proactive SIM functionality—specifically the RUN AT command—to hijack cellular modems. Testing 26 devices (18 smartphones, 8 IoT modems), they found 9 exposed an AT command interface to the SIM, enabling code execution, file theft, denial of service, and forced 2G downgrades. Demonstrated attacks include code execution on an Autel EV charger via a Quectel EC25-AFX module and 198 AT commands accessible on an Oppo Reno14 F 5G, including one that forced a stubborn downgrade to 2G that couldn't be reversed by toggling airplane mode or changing network settings.

Why: If you ship IoT devices with cellular modules (especially Quectel modems, which are common in Malaysian IoT and fleet deployments), audit whether your modem exposes the AT command interface to the SIM and whether you can disable proactive SIM commands. The Oppo Reno14 F 5G is a consumer device sold in Malaysia, so the 2G-downgrade and shutdown attacks are directly relevant to local mobile users—worth flagging if you build mobile apps or advise on device security.

Anthropic reports that an unreleased research version of Claude, when challenged to attempt the Riemann hypothesis, unexpectedly improved a longstanding lower bound for the fraction of Riemann zeta zeros on the critical line from 41.6% to 67.2%. Two Anthropic mathematicians validated the result, and Claude produced a formally verifiable proof. The work draws on recent research by Baluyot, Goldston, Suriajaya, and Turnage-Butterbaugh extending Montgomery's 1973 techniques.

Why: This is a vendor showcasing its own model on a benchmark with no direct builder takeaway — you don't need to change any tooling or workflow because of it. The only useful signal is that frontier models are now producing novel mathematical proofs drawing on and extending specific lines of published research, which suggests AI/ML practitioners working on reasoning-heavy tasks should track whether formal-proof and verification capabilities become available as product features.

Google announced that AMIE, its research medical AI system, demonstrated real-time clinical video consultation capabilities in what it calls a first-of-its-kind study. The blog post is primarily navigation boilerplate with minimal substantive detail about the study's methodology, results, or limitations.

Why: This is a vendor announcement of a research system with no product availability, API, or actionable integration path. Builders in health-tech or telemedicine in Malaysia should note the trajectory but cannot adopt or test AMIE; no decision or implementation change is warranted from this post alone.

The article argues that AI-generated content and AI-mediated search are eroding the discoverability and reliability of the web's existing knowledge base, with the replacement being worse than the original. However, the article body was not successfully scraped—only navigation boilerplate is present in the source text.

Why: The source text contains no substantive article content to extract concrete takeaways from. Only the title is available, which signals a broad cultural critique of AI's impact on web search and collective memory, but no actionable details, data, or specific claims can be verified from the provided text.

The article title asks whether peer review can survive being overwhelmed in the AI era, but the provided text contains only cookie consent and privacy policy boilerplate from Ars Technica—no article content is present.

Why: Cannot assess practical impact because the actual article body was not included; only the title and privacy/cookie notice are available. No actionable takeaway can be derived from the text provided.