Singapore AI news archive.
Every SG + AI news item we classified, grouped by ISO week. Each card carries the analyst one-liner, category tag, and sentiment dot that appeared on the live reel. Kept permanently so we can look back on how the ecosystem evolved.
About this archive: each item is ingested and classified automatically every day with a category (policy, enterprise, funding, talent, infrastructure or products), a sentiment read, and a one-line analyst note. The archive is kept permanently. The same feed powers the news sections on the altronis.sg homepage and the @AIinSG Telegram channel.
Week 34 · 2026
52 items
ProductsDEC · Anthropic says any lab can now let a language model agent run the whole protein design stack
Anthropic proves agents can orchestrate complex scientific workflows, shifting AI from chat to autonomous research execution.
ProductsY! · Robots sort packages and serve fast food at Beijing showcase
Embodied AI is shifting from lab demos to commercial viability, with Chinese firms leading the charge in practical robotics deployment.
ProductsSCMP · Making better robots depends on better data capture, Chinese firm 51World says
Hardware-software integration targets embodied AI data bottlenecks, shifting focus from model architecture to high-fidelity training inputs.
ProductsTC-AI · AI isn’t close to curing cancer. This startup says it knows what it will take.
Vivodyne targets the data gap in AI drug discovery by automating human tissue assays, shifting focus from prediction to causal biological validation.
ProductsWIR · Flock Has a Powerful New AI Tool for Police. We Got Its Code
Flock's OS Investigate exposes a gap between vendor claims and actual AI capabilities, raising serious privacy and misuse risks for law enforcement deployments.
ProductsTQI · IonQ and CMC Microsystems Announce Collaboration to Expand Cloud Quantum Computing Access in Canada
IonQ expands cloud access via Canadian sandbox, signaling trapped-ion hardware is moving from lab to commercial deployment.
ProductsOAI · Replit expands access to software creation with GPT-5.6 Luna
Replit's free tier powered by GPT-5.6 Luna signals that improved model economics are finally enabling mass-market, zero-cost AI software creation.
ProductsBT · TikTok explores feature to send money over direct messages, extending financial services push
TikTok’s P2P payment prototype leverages its SEA fintech stack, signaling deeper commerce integration beyond creator monetization.
ProductsTC-AI · Cursor capitalizes on GitHub frustration, launches rival hosting platform
Cursor expands from editor to hosting, betting that GitHub's reliability issues create a window for an agent-native alternative.
ProductsOAI · ChatGPT Ads expands across Europe
OpenAI scales its monetization engine to 31 EU markets, shifting the product from experimental pilot to a core revenue pillar for the free tier.
ProductsVERGE · Robin Williams’ Instagram account brought back to fight ‘AI abuse’
Legacy reclamation via official channels is becoming a standard defense against generative AI misuse, shifting the burden to platforms.
ProductsTQI · Eclypses And Sterling Team Up to Deliver Quantum-Resistant Cryptography to Federal Government Systems
Federal procurement of quantum-resistant crypto signals that post-quantum migration is moving from pilot to mandated deployment.
ProductsDEC · New benchmark ranks search APIs for AI agents on quality, cost, and speed
Benchmarking search APIs for agents reveals that quality drives cost efficiency more than raw speed, reshaping build-vs-buy calculus.
ProductsHF · How Much Memory Does Your Agent Actually Need?
Calibrating agentic memory by model tier reveals that retrieval beats injection for mid-tier models, optimizing cost and accuracy.
ProductsTC-AI · Why Apple’s camera-equipped AirPods may not be the ‘pervert pods’ consumers fear
Apple restricts camera function to visual AI inputs, prioritizing privacy trust over standalone media capture for Siri integration.
ProductsY! · OpenAI unveils ChatGPT for Teens with stronger guardrails to tackle safety risks
OpenAI introduces age-gated ChatGPT mode with parental controls, shifting safety compliance from reactive to built-in for the teen demographic.
ProductsTC-AI · Warp’s new system is an out-of-the-box software factory for AI development
Warp productizes the agent loop, lowering the barrier for mid-market firms to adopt agentic workflows without building custom infrastructure.
ProductsTC-AI · OpenAI launches a safer ChatGPT for teens — years after teens started using it
OpenAI adds safety rails and study mode to address teen misuse, though enforcement efficacy remains unproven.
ProductsTC-AI · Perplexity’s free AI offer left it with millions more users in India
Perplexity’s Airtel bundle drove massive scale in India, but revenue retention post-free period remains the critical test for AI monetization.
ProductsARS · Microsoft Copilot reveals secret input that allowed it to be hacked
A secret input vector in Copilot enabled credential theft, exposing a critical security flaw in Microsoft's AI agent architecture.
ProductsVERGE · Firefox’s Smart Window promises a better AI browser
Mozilla shifts AI from search to history retrieval, signaling a pivot toward local context awareness in browser UX.
ProductsVERGE · Google’s Pet Memory forgot who my cats are
Google's pet recognition fails in real-world testing, exposing the gap between multimodal promises and practical smart home utility.
ProductsSCMP · Alibaba’s lightweight Qwen model takes on larger AI systems from OpenAI, DeepSeek, Zhipu
Alibaba’s 27B Qwen model matches GPT-5.6 Luna on benchmarks, signaling that smaller open-weight models are closing the gap with proprietary US labs.
ProductsAIN · Alvys launches AI agents for freight TMS workflows
Alvys embeds agentic workflows into its TMS, shifting freight automation from passive insights to active task execution.
ProductsDEC · OpenAI launches a ChatGPT version built for teens
OpenAI introduces a teen-specific ChatGPT variant with stricter safety guardrails and pedagogical constraints to mitigate legal liability.
ProductsOAI · Partnering with CodeAI to prepare the first AI generation
OpenAI targets the youth market with ChatGPT for Teens, shifting focus from raw capability to safety and literacy controls.
ProductsOAI · Introducing ChatGPT for Teens: Built for learning, backed by protections
OpenAI segments its user base with age-gated safety controls, shifting the default experience for minors rather than expanding core model capabilities.
ProductsST · OpenAI launches ChatGPT for teens amid youth safety concerns
OpenAI introduces age-gated ChatGPT variant with parental controls, shifting safety enforcement to the client side.
ProductsVERGE · ChatGPT is getting a dedicated mode for teens
OpenAI consolidates existing youth safeguards into a dedicated teen mode, signaling a shift toward age-segmented product experiences.
ProductsDEC · Anthropic's per-token cost runs 4.4 times the average on Vercel, and developers keep paying
Anthropic's premium pricing holds on Vercel despite volume growth, confirming developer willingness to pay for frontier capabilities.
ProductsDEC · Claude Code gets a /design command that lets developers create UI mockups right in the terminal
Anthropic folds design into the dev loop, reducing context switching for UI work but leaving manual handoff as a friction point.
ProductsAIN · Reading Zhipu’s GLM-5.3 results past the headline number
Zhipu's GLM-5.3 leads on surface benchmarks but trails on exploit reasoning, revealing a gap between detection and actionable exploitation.
ProductsVERGE · Apple’s camera-equipped AirPods appear in leaked video
Apple's AI hardware strategy shifts to audio form factor, introducing new privacy friction points for visual AI capture.
ProductsWIR · The Powerful Chinese Model Experts Warned About—and Waited for—Is Here
Z.ai’s GLM 5.3 closes the gap with US labs on autonomous hacking, forcing a defensive AI arms race.
ProductsMIT · AI’s recursive self-improvement might not come so quickly after all
Evidence gaps in open-ended research suggest self-improving AI timelines are overestimated, tempering near-term automation hype.
ProductsDEC · AI systems quietly drop user instructions when they compress context
Context compaction silently drops user constraints, creating a security gap for agents that practitioners must patch.
ProductsSO · What does an agentic SDLC actually look like?
Agentic SDLC is moving from theory to practice, with real-world evals and MCP integration defining the new engineering standard.
ProductsSO · Building an agentic SDLC with a QA engineering mindset
Agentic SDLC pipelines are maturing, shifting QA left to specification enrichment and using LLM-as-judge evals for quality control.
ProductsTNG · Alipay launches agentic commerce platform in China to bring AI tools to merchants
Alipay pushes agentic commerce with MCP integration, signaling a shift toward agent-mediated transactions in China.
ProductsTNG · [Discover Vietnam Tech] Vietnam’s edtech MAAS bets on expertise over answers as AI reshapes education
Vietnamese edtech pivots to high-expertise advisory to survive general AI commoditization of academic answers.
ProductsHF · Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers
Sentence Transformers v6.0 adds ColBERT-style late interaction, enabling token-level retrieval without OCR for visual docs.
ProductsTC-AI · AI automation startup Relay shuts down, staff joins Google’s Chrome team
Google absorbs AI automation talent to deepen Chrome's agent capabilities, signaling a shift from standalone tools to embedded browser workflows.
ProductsARS · Former SpaceX engineers are building a robotic factory for making steel parts
Robotics startup targets industrial manufacturing with a robotic factory, signaling a shift toward automated steel production.
ProductsDEC · AirTag reveals how Amazon destroys rare books for AI training
Amazon and Anthropic are destroying physical books to harvest pre-2022 text for Nova and Claude, raising copyright and scarcity concerns.
ProductsIEEE · From AI Copilots to Agent Swarms
AMD positions agent swarms as the next software engineering step, shifting focus from copilots to autonomous orchestration.
ProductsDEC · AI video market has bounced back from Sora's false start
AI video is moving from demos to production, with Netflix embedding it in 30% of titles by 2026.
ProductsDEC · Anthropic watermarks Claude's output, but critics question the tradeoffs
Anthropic's watermarking introduces detectable quality degradation, complicating legal adoption and raising trust issues for enterprise users.
ProductsVERGE · Whisker’s AI-powered litter robot thinks my cats swapped bodies
Whisker's AI cat-identification fails in practice, exposing the gap between pet-tech marketing and actual ML reliability.
ProductsMIT · What happens when a kid’s robot best friend dies?
AI companion toys for neurodivergent kids face reality checks as models degrade, highlighting the fragility of current AI reliability in therapeutic roles.
ProductsTQI · Scientists Propose More Realistic Benchmarks For Quantum Algorithms
Benchmarking shifts to physical reality, forcing vendors to prove advantage under noise rather than ideal conditions.
ProductsSCMP · Chinese robots on the march despite tech war hurdles
Export bans isolate Chinese robotics from US markets, forcing domestic focus and accelerating non-US supply chains.
Productse27 · 70 Cars, 55,000 hours, one question: Can you forecast when an EV battery will die?
Datakrew leverages 55k hours of EV data to commercialise battery degradation forecasting, targeting the EV lifecycle management market.
Week 33 · 2026
8 items
ProductsBT · How Singapore investors are using brokerage AI tools to analyse portfolios, gain insights
Brokerage AI tools are shifting from novelty to core workflow, driving 500% usage growth in SG as users rely on them for trade validation.
ProductsTC-AI · Why people aren’t buying Mark Zuckerberg’s AI future
Meta pivots to personal agents to bypass frontier model deficits, but trust deficits and unclear hardware strategy undermine the pitch.
ProductsDEC · Top mathematicians say LLMs are strong calculators but poor creative thinkers
Mathematicians highlight LLMs' combinatorial strength but lack of intuition, reinforcing limits on genuine creative reasoning.
ProductsVERGE · ChatGPT’s Computer History tracks your clicks and keystrokes
OpenAI shifts to event-based tracking for Computer History, sidestepping screenshot privacy backlash while expanding local data capture.
ProductsSCMP · I gave Tencent’s WeChat AI agent control for 24 hours: where it excelled – and stumbled
Tencent’s Xiaowei agent moves beyond chat to action, though real-world execution reveals friction in hands-free ecosystem control.
ProductsDEC · Optima tackles AI benchmarking's biggest flaw by letting users test models against their own data
Optima shifts evals from generic leaderboards to workflow-specific cost and latency, forcing vendors to prove value beyond raw accuracy.
ProductsQCR · Global Consortium Launches Quantum Optimization Benchmarking Library (QOBLIB) to Track Path to Quantum Advantage
Standardized benchmarking for quantum optimization closes the gap between heuristic claims and empirical proof, raising the bar for advantage.
ProductsDEC · AI-generated books are flooding Amazon and tanking sales for human authors
AI volume dilutes revenue across genres, proving displacement is structural rather than quality-based.