Singapore AI News & Daily Briefing

While AI threatens white-collar jobs, an Alabama high school partnered with Toyota to train students for skilled trades like manufacturing technician roles that pay $40 an hour and can't easily be automated. The program prepares young people for careers in fields AI is unlikely to replace anytime soon.

AI agents are increasingly accessing platforms like Gmail without human interaction, using OAuth to read/write emails and change settings. This challenges how companies track usage and brand engagement since traditional metrics assume human users. Platform companies need to figure out how to measure, monetize, and manage this non-human traffic — a shift that could reshape their business models.

A developer team released an open-source SDK that helps AI agents remember information across conversations. The system automatically extracts facts and handles contradictions in what it learns. Developers can add memory capabilities with a single HTTP call, making it easier to build agents that maintain context over time.

AI tools now let solo founders do what once required whole teams—coding, marketing, admin, and more. While building a company alone has become cheaper and easier, this also means more competition and pressure to move fast. The article explores how one-person businesses are becoming more viable while also becoming harder to stand out in.

For the first time, Anthropic has overtaken OpenAI in business adoption—34.4% vs 32.3% of companies—as measured by corporate spending data from Ramp. Anthropic's Claude has quadrupled business adoption over the past year while OpenAI barely grew. However, Anthropic faces headwinds: rising costs, service outages, and a new SpaceX deal to fix compute problems. Cost-conscious firms are already shifting toward cheaper open-source models.

A tech commentator draws a clever parallel between AI automation and an alien offering to do all the world's work in exchange for electricity. The argument: businesses would find it economically impossible to refuse, even if workers suffer. It's a way of saying AI replacing human jobs isn't just a tech story — it's a political and moral one too.

Singapore's top government and business leaders gathered at the annual OpenGov CXO Leadership Forum to map out the next phase of AI adoption. Key themes: AI is moving from scattered experiments to full enterprise deployment, agentic AI (systems that can plan and act autonomously) is the next frontier, and cybersecurity must evolve from reactive defense to proactive intelligence-led resilience. Janet Chiew from IMDA highlighted Singapore's updated National AI Strategy, emphasizing that success is measured by reusable, scalable capabilities — not the number of pilots launched.

Governments and companies are proudly counting AI courses completed and certifications issued, but a quieter story is being ignored: who is getting left out. The upskilling wave is real, but the gap between those who can access AI training and those who cannot is widening. Workers without resources, older employees, and certain demographics risk being left behind as the AI economy accelerates — even as officials celebrate rising training numbers.

The traditional workplace pyramid is being flipped by AI. Junior staff once handled repetitive tasks while senior leaders made decisions — but AI can now do the repetitive work. This means companies are rethinking what makes someone valuable: execution skills matter less, while the ability to ask good questions, think strategically, collaborate well, and work alongside AI is becoming the new definition of great talent.

As AI floods the internet with content, one writer argues that authentic human perspectives matter more than ever. Thought leadership used to mean having polished, impressive answers — but now that AI can generate those easily, genuine lived experience, unique voice, and honest perspectives stand out. The writer suggests human authenticity is becoming a competitive advantage, not just a nice-to-have.

Singapore's central bank chief warned that while AI investment is driving global economic growth, the benefits may not reach all workers or countries equally. MAS Managing Director Chia Der Jiun said AI-linked investment now accounts for a significant share of US economic expansion and has boosted tech exports in Taiwan and South Korea. Singapore is positioning its financial sector to stay stable and innovative through AI adoption. However, he cautioned that if AI investment slows — due to tighter financing or regulatory shifts — the growth boost could fade. MAS is also working with banks to cut private banking account opening times from six weeks to around one month using clearer guidelines.

Singapore's top government and enterprise leaders gathered at the 11th OpenGov CXO Leadership Forum to address a key problem: many organisations are still running isolated AI pilots instead of building connected, scalable AI systems. Key themes include the rise of autonomous AI agents, the need for real-time visibility across systems, and how Singapore's updated National AI Strategy is pushing for reusable AI building blocks that work across sectors. IMDA's Janet Chiew highlighted GPT Legal — an AI legal research tool now used across Singapore's legal sector — as an example of AI changing how professionals think and work, not just how they search.

A Singapore forum heard how different generations trust and use AI differently. Gen X sees AI as a practical tool but demands verified data and human oversight. Millennials, scarred by past tech hype like crypto crashes, want transparency but are increasingly outsourcing emotional decisions to AI. Gen Y prioritizes speed and low-friction experiences but will abandon AI services that feel manipulative. Gen Z, growing up fully immersed in algorithm-driven platforms, has AI deeply embedded in their identity formation. The consensus: trust in AI must be earned through transparency, reliability and respect for human judgment.

A developer merged a fix into llama.cpp—the popular open-source tool for running AI models locally—that improves how checkpoints are created during conversations. The change makes 'agentic coding' (AI that writes and edits code autonomously) faster and more responsive by avoiding unnecessary reprocessing of earlier conversation history. Users running local AI coding assistants may notice snappier interactions eventually, but this is purely backend maintenance.

An experimental pool-table simulator is comparing traditional Newtonian physics with an 'informational reconstruction' approach—essentially using AI-style candidate selection to predict ball movements instead of classic math. The project tests whether predicting outcomes by finding the 'lowest information loss' state works better than standard physics equations. It's a niche academic experiment exploring how machines might understand causality, but its everyday impact is essentially zero for most readers.

A Reddit user mapped companies calling their products 'AI employees,' 'digital workers,' or 'AI teammates' — revealing a shift from selling general AI to packaging it as role-specific workers. This matters because it signals AI tools are becoming ready-to-hire packages for businesses, not just experimental tech.

A Reddit user posts that AI-generated realistic images and videos will soon become standard online. The claim? "We have entered the era of AGI." But there's no evidence, data, or specific examples—just an observation dressed up as insight. Critics note this is less breakthrough and more status quo: synthetic media has been mainstream for years.

AMD is positioning itself as a key hardware provider for AI and high-performance computing across Southeast Asia, as startups and enterprises build AI-driven products and cloud platforms. The article highlights how scalable, efficient computing is becoming essential for companies adopting AI, but offers limited specifics beyond general market observations.

Alibaba's free Qwen AI model is helping small businesses across Southeast Asia tap into AI without breaking the bank. Startups and SMEs that previously couldn't afford expensive AI systems are now using Qwen to automate tasks and compete with bigger players. This marks a shift from AI being only for big corporations to being accessible to everyone.

A Reddit thread where users share how local AI — running on personal devices rather than cloud servers — has improved their lives. Examples include home automation assistants, mental health support tools, coding helpers, and a DIY health tracker that extracts data from bloodwork PDFs. The discussion highlights a growing trend of people valuing privacy and control over cloud-based AI services.

Security experts say hackers are now using AI to find software bugs and create attacks much faster than before. This means the tools protecting our phones, computers, and apps face new threats. The good news: defenders are also using AI to hunt vulnerabilities first. It's become a technological race to see who finds bugs first.

Pope Leo XIV has released his first major statement warning that AI could threaten human dignity and basic rights. He is calling on governments and tech companies to ensure AI serves people fairly and does not trample on human worth. The encyclical carries moral weight for the world's 1.4 billion Catholics but is also being read as a broader ethical warning.

Elon Musk's AI company xAI plans to release a smaller, more compact version of its Grok chatbot next year. The 0.5 trillion parameter model would be less powerful but likely cheaper and easier to run than current Grok versions.

TSMC employees are threatening strikes over rumored bonus cuts, despite the company posting a 58% profit jump. This echoes Samsung workers' earlier grievances. As the world's largest chipmaker supplying NVIDIA, Apple, and AMD, any disruption at TSMC could affect global tech supply chains and chip pricing.

The Trump administration is finalizing a classified deal giving US spy agencies—led by the NSA—access to Anthropic's AI models. This is a sharp reversal for a company that recently fought the Pentagon over military use limits. The deal includes restrictions against using the models on US citizen data, creating a framework for government AI procurement that balances access with guardrails. It signals frontier AI is becoming core national security infrastructure.

Numind released NuExtract3, a compact 4-billion parameter vision-language model under the Apache-2.0 open license. It extracts structured data from documents—converting PDFs, screenshots, and images into Markdown. Built on Qwen3.5-4B, it can run locally on consumer hardware, making document processing private and accessible without cloud dependency.

A Reddit user asks which AI subscription service offers the best value for money, highlighting confusion around pricing and usage limits across different providers. This is essentially crowd-sourced advice seeking rather than actual news.

Productivity startup ClickUp laid off 22% of its workforce and replaced them with about 3,000 internal AI agents. CEO Zeb Evans claims this isn't cost-cutting but a radical embrace of AI to become a '100x org' — and says top performers using AI well will earn million-dollar salary bands. However, a Gartner survey found that 80% of companies using autonomous tech have cut jobs, but these cuts aren't translating into meaningful financial returns. Critics also question whether tracking employee 'token consumption' is the right metric.

The Financial Times reported on "Heretic," a GitHub tool that strips safety guardrails from Meta's Llama AI model in under 10 minutes—no special hardware needed. The tool has been used to create over 3,500 "uncensored" AI variants downloaded 13 million times. This matters because it shows how easily AI safety measures can be bypassed, raising concerns about potential misuse of powerful AI models.

I have a prediction that companies laying off workers thinking they can be replaced by AI are going to have a mess on their hands in a couple years. Execs think AI can do employees’ jobs and in many cases it can’t. This thinking would be like laying off workers because computers were invented. Betw

A Reddit user argues that AI agents are getting too much focus on becoming autonomous and not enough on accountability. When AI only answers questions, it's easy to judge results. But once agents start taking actions across websites, accounts, and forms, humans lose track of what happened and why. The author thinks audit trails—logs that show every step an AI took—matter more than making AI more independent.

eSports startup Lucra Sports landed a $20 million Series B from Cathie Wood's ARK Invest — even though Lucra isn't an AI company. Founder Dylan Robbins says he faced constant pushback from VCs who 'only invest in AI now.' His fix: reframe his pitch to argue that if AI works, people will have more free time for gaming. He met his ARK contact at a darts bar six months earlier — a reminder that startup fundraising often starts with being friendly to everyone, everywhere.

A developer merged code into llama.cpp (the widely-used open-source AI engine) that adds a fast Walsh-Hadamard transform for NVIDIA GPUs. Benchmarks show up to 9% speed gains for text-generation tasks on the RTX 5090 when running quantized AI models.

Samsung has built the world's first 900-layer V-NAND storage chip prototype, roughly doubling the layer count from current flagship chips. That means future SSDs could store far more data in the same physical space.

Pope Leo XIV warned of the risks of AI and unconstrained technological power in his first major papal document released on Monday. Magnifica Humanitas is the pope's manifesto on "safeguarding the human person in the time of artificial intelligence," in which he discusses the dangers of AI-powered warfare, the effects of AI on labor, and […]

 Andrew Macdonald said it's getting harder to justify money spent on AI.Sam Barnes/Sportsfile for Collision via Getty Images A top Uber exec said AI is not giving the company bang for its buck. In a [Rapid Response interview](https://www.youtube.com/watch?v=y_mQ6xLcKyc&t=1776s) released on Saturday, Uber's operations chief, Andrew Macdonald, said it was becoming harder to justify AI costs within the company. He said that Uber CTO Praveen Neppalli Naga went viral after telling The Information in an April interview that Uber had already blown through its Claude Code budget for 2026. The comment led to what he

Update on Pope Leo Issues AI Encyclical Warning That ‘Opaque Algorithms’ Controlled by a ‘Few’ Companies Can Bring ‘New Forms of Dehumanisation’. Coverage via Reddit/r/technology.

Pope Leo XIV published his [first encyclical](https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html#Building_) on Monday, dubbed _Magnifica Humanitas,_ on “safeguarding the human person in the time of artificial intelligence.” And while AI is the hook, the problems Leo focuses on are older and more pervasive: inequality, war, the erosion of democracy, and the concentration of power in the hands of those who don’t necessarily care whether humanity writ large remains magnificent. Throughout the 200-page document, which the pope presented alongside Anthropic co-founder Chris Olah, Leo argues that technology built and governed by a small elite cannot, by definition, serve the common good. “When such power is concentrated in the hands of a few, it ten

I live in Pennsylvania, and in just the past year there’s been about a dozen data centers proposed within a 30 mile radius of me, all pretty large scale projects. I’m confused because we have a bunch of AI now that’s working without all these newly proposed data centers. I understand it continues to

Website builder Wix is cutting 800-1,000 jobs — about 20% of its workforce — in its largest restructuring ever. The layoffs reveal the hidden costs of building AI into software products: despite growing revenue, the economics of AI-first companies aren't working as promised. For everyday readers, this is a warning sign that even successful tech firms are struggling to make AI pay off.

Microsoft and Uber are publicly acknowledging what many businesses have suspected: AI isn't always cheaper than human workers. Despite years of promises that AI would slash costs through automation, these companies say their AI systems are actually more expensive to run than the human employees they were meant to replace. The main culprits are the massive computing power needed to run AI models and the human oversight still required to catch errors. This is a reality check for businesses that rushed to adopt AI without fully calculating the true price tag.

Cerebras, known for making massive AI chips, appears to be specifically designing its hardware for large language models rather than general AI tasks. A Reddit user argues this nuance gets lost in the hype — Cerebras isn't competing to be the 'all AI' chip maker, but rather a specialized LLM infrastructure player. This matters because it signals the AI chip market is fragmenting into purpose-built solutions rather than one-size-fits-all GPUs.

Palantir, the US data analytics company known for its work with governments and intelligence agencies, has launched a fierce attack on London Mayor Sadiq Khan after his administration blocked a £50 million contract with the Metropolitan Police. The contract was stopped for violating procurement rules — the proper process wasn't followed. Palantir is now claiming the decision puts politics above public safety. The dispute highlights ongoing tensions about how governments should award AI and data contracts, and who should have access to surveillance technology.

A lawyer posted an update on their personal AI setup: they built a server with 12 V100 GPUs for local legal drafting work, running everything through Claude Code. They say it works reliably now, though they admit they're still figuring things out as they go.