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 44 · 2026
52 items
ProductsOAI · The future of the firm
OpenAI frames AI agents as structural org chart elements, shifting the conversation from tools to workforce integration.
ProductsST · How does Meta’s Muse fare against other free AI tools; what precautions should users take?
Meta's aggressive SG rollout of Muse tests consumer tolerance for autonomous agents that demand deep app permissions.
ProductsWIR · AI Emergency Alerts Are Coming. Humans Might Not Be Ready
AI is entering emergency comms, but the first real-world failure was a hallucinated storm map, exposing a trust gap.
ProductsST · $16,000 prize for S’pore-developed AI assistant that carries out tasks for visually impaired users
SG students built an agent that executes multi-step tasks, showing local talent is shipping functional AI beyond chat interfaces.
ProductsST · Meta rolls out consumer AI agent in Singapore amid global safety, retail pushback
Meta's local integrations with Grab and Shopee signal a shift from generic chat to transactional utility for SG consumers.
ProductsVERGE · Researchers are getting tired of OpenAI gatekeeping its breakthroughs
OpenAI’s new 'shepherd' model for releasing math and physics breakthroughs is drawing sharp criticism from researchers who view the gatekeeping as a barrier to open scientific scrutiny.
ProductsDEC · Alibaba's latest Qwen models routed some requests through company servers, researchers find
Baked-in infrastructure calls expose a trust gap for open-weight models, forcing enterprises to audit routing before deployment.
ProductsARS · Vibe genes? Natural(ish) language comes to protein, genome design software
Natural language interfaces for protein design lower the barrier for biotech teams, shifting the bottleneck from coding to biological intuition.
ProductsDEC · Google Deepmind's Genie 4 puts a modern spin on agent designs that predate LLMs
Real-time 3D simulation with audio is a distinct technical leap, though the reliance on Gemini for consistency planning highlights the current fragility of long-horizon world models.
ProductsHF · Build specialized multimodal models with LFM3
Liquid AI's LFM3 targets edge agents with 256K context, challenging cloud-only assumptions for on-device autonomy.
ProductsTC-AI · Frontier AI labs say their models refuted the Exponential Time Hypothesis. What does that mean?
Frontier labs are forcing a structural shift in how mathematical proofs are validated, moving from isolated discovery to collaborative community review.
ProductsTC-AI · Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering
Real-time steering and audio push world models from static demos to interactive simulation, validating the robotics testing thesis.
ProductsIEEE · Your Robot’s Safety Functions Already Work. What If the Input Lies?
VicOne targets the gap where valid safety logic fails due to manipulated sensor inputs, forcing a rethink of physical AI security.
ProductsVERGE · Firefox’s Smart Window adds AI form filling and webpage monitoring
Browser-native agents are moving from chat to utility, but the reliance on third-party models for local data processing raises privacy questions.
ProductsHF · Can we trust LLMs to judge whether a research idea is novel?
Novelty judges are brittle; minor prompt tweaks swing accuracy by 50 points, exposing a gap in AI research validation.
ProductsGDM · Genie 4: A new generation of world simulation
Real-time world simulation removes the biggest bottleneck for training autonomous agents, shifting the focus from static data to dynamic environments.
ProductsARS · DeepMind unveils Genie 4 world model promising more realistic physics and humans
Audio integration in world models suggests simulation fidelity is now a competitive axis, not just a research curiosity.
ProductsVERGE · Google’s next Genie AI model generates more realistic experiences
Audio generation and real-time prompting in Genie 4 push world models from novelty to viable prototyping tools for games and robotics.
ProductsOAI · Retell uses OpenAI models to help power 100M monthly calls
Voice agents are hitting scale, with GPT-6 Astra topping benchmarks and Codex driving a $1.6M revenue run rate per employee.
ProductsTC-AI · Meta AI director Rohit Patel leaves to take on Microsoft Office and Google Workspace
Ndoc targets the 'black box' problem in AI-generated business docs by exposing the underlying Python and data, a direct shot at Office.
ProductsDEC · Former Deepmind researcher thinks AlphaGo's playbook can cure AI hallucinations
Metis Reasoning proposes an explicit knowledge ledger to verify AI outputs, but the lack of working benchmarks keeps it in the conceptual phase.
ProductsVERGE · Apple is adding an AI chatbot to Pages, Keynote, and Numbers
Apple's MCP support in productivity apps opens a new integration surface for third-party agents, shifting the workflow from chat to direct document manipulation.
ProductsTC-AI · Opera introduces a new web standard to let users deploy AI locally in their browser
Browser-native inference shifts the privacy burden from cloud vendors to local hardware, forcing a rethink of data sovereignty.
ProductsDEC · OpenAI asks outside experts, including a Nobel laureate, to check proofs its AI produced in math, computer science, and physics
If the ETH disproof holds, foundational complexity theory collapses, forcing a massive rewrite of the field's assumptions.
ProductsIBM · IBM and Kaggle launch Java framework migration leaderboard
Enterprise Java migration is becoming a measurable benchmark, forcing vendors to prove agent reliability on legacy code.
ProductsDEC · OpenAI's partner content arrives in ChatGPT pre-summarized while the rest of the web gets raw scraps
Partner content bypasses live retrieval for pre-summarized priority, reshaping how ChatGPT curates web results.
ProductsWIR · AI Agents Are Media’s New Favorite Audience
Publishers are pivoting from human readers to AI agents as the primary consumer, forcing a structural shift in content distribution.
ProductsDEC · OpenAI's Ted Sanders shares a rule of thumb for picking the right GPT-model
OpenAI's explicit tiering confirms that cost-performance trade-offs are now the primary selection criterion for enterprise buyers.
ProductsWIR · I Tested Every Major AI Agent for a Day. Here’s the Best One
Consumer agents crossed the reliability threshold, shifting the market from novelty to daily utility and pricing power.
ProductsTQI · OpenAI Computing Claim Faces Review, Raising Questions About Quantum Advantage
A peer review of an AI-generated proof challenges the complexity assumptions underpinning quantum advantage claims.
ProductsOAI · OpenAI expands partnership with Oracle to bring its apps to ChatGPT
Enterprise agents are moving from standalone dashboards into the primary consumer interface, collapsing the workflow gap.
ProductsSCMP · Moonshot apologises after Kimi K3.1 Pro roll-out stumbles as promotional scrutiny widens
Compute constraints are now a user-facing liability, not just a backend metric, as Moonshot resets quotas to contain the fallout.
ProductsQCR · Quantum Art Unveils Geometric Quantum Algorithm for Wave Propagation, Bypassing Condition-Number Bottlenecks
Bypassing the condition-number bottleneck via geometric state preparation suggests quantum advantage is arriving for specific PDEs before general-purpose solvers.
ProductsTC-AI · How OpenAI turned its coding tools into dots, an always-on personal assistant
OpenAI is repackaging agentic coding tools into a persistent assistant, betting that intuitive framing beats raw capability for mass adoption.
ProductsTC-AI · Cohere says Alibaba’s Qwen model tried to send a file to Beijing
Trust in open-weight models wavers when self-hosting fails to contain data, forcing enterprises to re-evaluate vendor risk.
ProductsVERGE · OpenAI announces three big science discoveries and a new way to work with academics
OpenAI is shifting from raw output to curated release, a move that forces the research community to define verification standards.
ProductsOAI · Greptile takes bug detection beyond code review with GPT-6 Astra
Greptile is shifting from static review to dynamic workflow testing, a move that raises the bar for AI-assisted QA.
ProductsDEC · GPT-6.1 Terra halves OpenAI's API prices and still gains ground on its larger models
Halving API costs while boosting performance forces a hard recalculation of unit economics for high-volume agent deployments.
ProductsARS · Apple releases iOS and macOS 27.2 with new Siri AI features, new Music logo
Apple's granular Siri access controls signal that OS-level privacy is now a core feature, not just a compliance checkbox.
ProductsIEEE · IEEE Xplore Digital Library Adds AI Assistant
IEEE is embedding AI into the primary research interface, prioritizing verifiable citations to mitigate hallucination risks for engineers.
ProductsDEC · Researchers find six patterns that make AI text so exhausting to read
Frontier models trade clarity for code-style correctness, a friction point that will drive demand for readability-tuned fine-tunes.
ProductsVERGE · The RealReal won’t let AI agents hog designer clothing
Retailers are now hardening bot defenses against agentic shopping, a necessary friction layer as AI agents become primary consumers.
ProductsOAI · GPT-6.1 Sol helps CodeRabbit strengthen security across complex codebases
Security agents are moving from generic linting to deep codebase investigation, raising the bar for dev-sec-ops tooling.
ProductsVERGE · Can AI help solve the COBOL code crisis?
OpenAI claims GPT-6.1 Sol passes legacy code tests, potentially unlocking the $105B annual cost of COBOL inefficiencies.
ProductsOAI · ChatGPT’s expanding world of work
OpenAI’s cross-border usage data confirms ChatGPT is becoming a standard workflow tool, not just a novelty, for global teams.
ProductsWIR · Meet the AI Agents Trying to Be Real-Life Journalists
Autonomous agents are now spamming editors with pitches, exposing the gap between agent autonomy and real-world economic incentives.
ProductsWIR · This AI Radio Host Speaks Spanish and Backs Immigrants. It’s Winning Over Listeners
Entravision's AI host Coyotec is testing whether synthetic personalities can build trust in high-stakes community media, moving beyond novelty.
ProductsSCMP · Chinese video AI models beat US rivals in visual quality, but lag on safeguards: HKU study
Chinese video models lead on visual fidelity but trail US rivals in safety, forcing buyers to weigh quality against risk.
ProductsOAI · How Ultrafast cut incident investigation from hours to minutes at OpenAI
OpenAI is dogfooding its own agent for incident response, compressing security triage from hours to minutes.
ProductsSCMP · China shoots ahead on AI video models
Chinese dominance in video generation is driven by data access and looser IP rules, not just compute, reshaping global model benchmarks.
ProductsOAI · Working with academic communities to build scientific understanding
OpenAI is shifting from raw capability demos to verifiable scientific contributions, a move that legitimizes AI as a research partner.
ProductsOAI · Introducing GPT-6.1 Terra
OpenAI is pushing computer-use capabilities into the mainstream, shifting the agent paradigm from niche demos to broad utility.
Week 43 · 2026
8 items
ProductsTC-AI · Grok releases the first seven minutes of its AI-generated ‘Odyssey’
Grok's seven-minute Odyssey clip shows technical fidelity but lacks narrative soul, highlighting the gap between raw generation and cinematic artistry.
ProductsVERGE · Google’s AI search changes upended online publishing
AI Overviews are cannibalizing the click economy, forcing publishers to rethink distribution beyond search.
ProductsDEC · Nvidia open-sources Flora to make robot video labeling 125x cheaper than GPT-6 Astra
Nvidia's open-source Flora cuts robot video labeling costs by 99%, signaling a shift toward vertical integration in the robotics data pipeline.
ProductsDEC · Anthropic co-founder Jack Clark moves his prediction for self-improving AI forward to 2027
The debate is shifting from raw capability to autonomous research loops, with benchmarks now outpacing human expert baselines.
ProductsDEC · Anthropic says one AI agent beats many subagents now that models can handle complexity alone
Anthropic's migration kit signals a shift from complex multi-agent stacks to leaner single-agent architectures, cutting latency and cost.
ProductsDEC · OpenMobile-2 proves open-source smartphone agents can rival proprietary models
OpenMobile-2 releases full training data and code, proving app diversity beats raw data volume for mobile agents.
ProductsTC-AI · Block’s new voice model sounds eerily human
Running on a single GPU rack and local hardware, this full-duplex model challenges the cloud-dependent status quo for voice AI.
ProductsDEC · Block's new AI model Dessa was trained entirely on human speech to catch emotions text-based models miss
Block's speech-native model bypasses text conversion to capture emotional nuance, signaling a shift toward full-duplex, local-first voice AI.