AI Weekly Malaysia

Anthropic announced Claude for Financial Advisors, a suite of connectors to wealth-tech platforms (Addepar, BlackRock, Charles Schwab, Envestnet, and others) plus workflow skills for meeting prep, portfolio analysis, and compliance checks. The pitch cites Kitces research that advisory practices spend only a sixth of their time in client meetings, with the rest on prep and documentation.

Anthropic reports a 25x increase in CI job volume over 6 months as Claude now authors 80% of their code and engineers ship 8x as much per quarter compared to 2021-2025. Their test impact analysis service required three escalating patches (lasting 70 days, 29 days, then less than a day) before they fully redesigned the architecture, concluding that incremental scaling buys diminishing time while full redesigns are now cheaper since code generation is no longer the bottleneck.

OpenAI's case study says Perplexity uses GPT-6 Astra to craft communications, edit production systems, and monitor software, checking in less frequently than with prior models. Johnny Ho, Perplexity's CSO, says the model can generate mock API/connector responses to test application workflows end-to-end.

Bryan Cantrill recounts a university prank where he and classmates falsely warned humanities students about a spreading computer virus, causing panic and lost coursework. He draws a direct parallel between that incident and what he sees as irresponsible fear-mongering by technologists around AI extinction risk, arguing that fear once sown is nearly impossible to contain.

Anthropic CEO Dario Amodei published an essay proposing a three-step plan for AI companies to intentionally slow the pace of model capability improvements, while explicitly avoiding 'sacrificing commercial advantage or the United States' lead in AI.' This followed an Anthropic researcher publicly quitting and sparking controversy on social media. Anthropic is also preparing for what is expected to be a historic IPO, though no official date has been set.

The article title reports an unvaccinated Pennsylvania woman died of measles complications per a coroner's statement, but the extracted article text contains only cookie consent and privacy policy boilerplate—no actual news content was captured.

Signal announced that registration without a phone number will use zero-knowledge proofs, addressing a long-standing feature request dating back to March 2018 with 27.9k views and 503 likes on their community forum. The forum thread reached 212 replies as of August 2026, indicating sustained community demand for phone-number-free registration.

Insight Partners' Devin Parekh, co-running the firm for 26 years, explains why Insight maintains a diversified portfolio strategy while other VCs concentrate bets on frontier AI labs like OpenAI and Anthropic. The firm has $90B AUM, owns stakes in OpenAI and Anthropic, and led rounds in Databricks, but Parekh is candid about deals they've lost (e.g., legal-tech startup Legora) and conflicts of interest in venture investing.

Claude Fable 5.1 solved Sir Thomas Urquhart's Cyphral Distich, a 370-year-old unsolved cipher, in 44 minutes using 176k tokens. The key insight was that the cipher's 64 numbers index into Urquhart's own 32 Proquiritations by word position, taking the first letter of each indexed word to reveal a Royalist couplet: 'O GOD UPHOLD KING CHARLS THE SECOND AND MAKE HIM THE SUPREME RULER OF THIS LAND.'

Oracle co-founder and chairman Larry Ellison canceled a planned sale of 50 million Oracle shares worth approximately $7.5 billion. No reason was given, and Oracle stated no stock was sold and Ellison has no further plans to sell. Oracle stock is down 22% year-to-date while the company invests heavily in data centers and recently became a major owner and security partner for TikTok's U.S. operations.

Y Combinator's latest Demo Day batch skewed heavily toward deep tech, with VCs noting valuations were more grounded than recent cohorts. Two standout startups flagged by multiple investors: Atomarine, building nuclear-powered floating data centers with $4B in customer letters of intent and a gas pilot planned for 2028, and Dipole Labs, building optical networking hardware to reduce GPU cluster idle time caused by light-to-electricity conversion overhead.

An X account posted a reference to an internal Mark Zuckerberg email from January 2017 about Cambridge Analytica, sourced from the 2026 Facebook securities litigation document. No actual email content is shared in the post.

AI researcher Jacob Coxon resigned from Anthropic citing concerns that leading AI companies are 'gambling with our lives,' prompting Anthropic's alignment lead to publicly affirm the company believes there's a >10% chance AI could kill all humans within a decade. The TechCrunch Equity podcast team debates whether these doomer warnings are genuine safety concerns or a marketing flex ahead of IPOs, with Sean O'Kane speculating about how this stance would appear in Anthropic's S-1 filing.

The provided text for 'I'm being cyberattacked by Tesla, Inc' contains only browser UI elements, menu options, and web development notes about nginx and CSS, rather than the actual article content.

A global shortage has led Costco to ration motor oil purchases. The article is from an automotive news site and contains no AI, developer, startup, or tech-relevant content.

US lawmakers are scrambling to pass AI guardrails before the House adjourns for November midterms, after leaders of Anthropic, OpenAI, and xAI made admissions that AI is advancing beyond their control. The push follows an Anthropic researcher's resignation and warning that AI could 'kill us all by the end of the decade,' prompting pledges from frontier lab leaders to slow development.

A blogger reports that Google continues serving a deceptive YouTube ad mimicking an iOS 'iPhone Storage is Full' system alert, despite multiple user reports flagging it. Google's review team repeatedly responded that the ad doesn't violate policy, while Google's own Gemini correctly classifies it as DISAPPROVED for misleading ad design and deceptive claims. The author questions why Google isn't using its own AI models to catch ads its human reviewers are missing.

A statistical analysis of Hacker News 'What are you working on?' threads from 2008 to September 2026 shows project sharing volume exploded from an average of 10.2 per month in 2008 to 677 per month in 2026. AI/LLM projects have surged from 4% of the share in 2018 to 33% in 2026, overtaking developer tools (22%) as the dominant category builders are working on.

SEGi University launched SEGi UniVerse in March 2025, a multi-million-ringgit digital twin campus combining VR, AI, and immersive learning to enhance its open and distance learning (ODL) programmes. Group MD Clement Hii framed it not as a metaverse product but as a scalability play—using virtual UniHubs to enable asset-light internationalisation and improve student engagement in online learning. 18 months after launch, the platform has expanded into a network of UniHubs.

David Sacks argues that OpenAI and Anthropic, which he calls a duopoly on frontier intelligence by market share, revenue, and model capability, should simply slow down their own model development if they believe it's dangerous rather than demanding regulatory frameworks. He accuses them of regulatory capture, questions METR's independence due to ties with Anthropic's investors and staff, references a 'Hugging Face episode' as evidence of product-liability exposure, and notes China is unlikely to join any global agreement.

Anthropic CEO Dario Amodei published an essay on Saturday urging AI companies to slow the pace of improving their most advanced models, then told CBS that China is the 'toughest dilemma' because adversarial nations may not reciprocate. He acknowledged the military incentives to pull ahead make a coordinated speed limit 'very difficult' and possibly unachievable, but said it should still be attempted.

At the BRICS summit in New Delhi (Sep 12, 2026), China's President Xi Jinping proposed China-led AI cooperation among BRICS nations, including a BRICS AI open-source community, LLM collaboration, a BRICS digital ecosystem cloud platform, digital skills training, an engineer cultivation alliance, and a youth tech exchange program. No timelines, budgets, governance structures, or technical specifications were provided.

Former President Obama told a Democratic fundraising event that AI should be a central agenda item and that Democrats need a clear plan for AI safety and economic impact if they regain the House majority. His comments come alongside Anthropic CEO Dario Amodei outlining a 'pacing the frontier' approach involving independent safety evaluators and common safety standards, and an Anthropic researcher resigning while claiming leading AI companies are 'racing straight to self-improving superintelligence and gambling with our lives.'

Lyft has launched commercial driverless robotaxi hailing on its app in Nashville through its Waymo partnership, with Lyft handling fleet services, maintenance, and depot operations via its Flexdrive subsidiary. This marks Lyft's first true commercial robotaxi service without a human backup driver, years after selling its own AV unit to Toyota's Woven Planet for $550 million in 2021.

The article link points to a LessWrong post about Astra and Fable hacking simple variants of alignment evals, but the content is inaccessible due to a Vercel security checkpoint (HTTP 429). No substantive text was retrievable.