The stories, launches and policy moves that matter in AI — verified against primary sources and updated as they break.
40 stories tracked and verified by the StackHK news desk.
GitHub graduates Copilot's autonomous coding agents from preview to GA with new concurrency limits and strict enterprise security controls. Here's what changed.
Anthropic's Claude app adds a desktop widget and quick-prompt bar, narrowing the gap between your calendar, browser, and the assistant you actually use.
Nvidia's new RTX AI Workstation line brings up to 96GB of GDDR7 to desktops for local LLM development, from $2,999. Here's what the three tiers offer.
Amazon's Titan 3 family reaches general availability on Bedrock, bringing multimodal and on-device inference options to AWS's mainstream enterprise customers.
Grok 4 can now answer questions about what's on your screen or camera in real time, extending xAI's live-vision push to mobile and desktop apps.
Google rolls deep Gemini integration into enterprise Gmail: thread summaries with citations, AI drafting, and agentic actions under strict admin controls.
Notion is rolling out AI database agents that crawl linked sources to fill rows, dedupe records, and refresh statuses on schedule. Here's when you get them.
Meta's enterprise AI Studio brings custom assistant-building to business teams, adding admin controls, usage analytics, and managed access across Workspace and developer APIs.
OpenAI's next o-series reasoning model ships with adjustable thinking budgets and lower per-token pricing, targeting agentic coding, math, and research work.
ElevenLabs ships Voice Island, a standalone app for text-to-speech, voice cloning, and audio editing in one timeline. Here's full pricing and rollout.
Meta, Microsoft, Nvidia, IBM and Palantir formed a coalition with Hugging Face and the Linux Foundation to push back on open-weight restrictions — the counter-lobby to the safety letters.
While writing a $17.1 billion settlement check, Meta also expanded its Nvidia deal to millions more AI chips. The bill for the AI future arrives on two tracks.
Fifty-nine feature changes in one month across Excel, PowerPoint, Word, Chat and Cowork. The Copilot fast-follow strategy is working exactly as designed.
“Automated AI R&D” may become a major risk as AI research starts compounding. The safety case is coming from inside the lab.
The Claude lab is hiring for custom silicon — the next stage of the compute vertical-integration race that Nvidia’s customers all eventually start.
Vertex customers get higher-fidelity generations with director-level camera movement — our testing shows it rivaling Sora shot-for-shot.
OpenAI's video model now generates dialogue, effects, and ambient sound matched to the visuals — the missing piece for production use.
We built three real apps in three weeks — heres where autonomous building shines and where it breaks.
Kuaishous generator pairs near-top-tier quality with the most generous free credits in the category.
Live social awareness lets Grok catch sentiment shifts hours before manual monitoring — for better and worse.
Three months after Sundar Pichai declared the agentic Gemini era at I/O, the demos have become defaults across Search, Workspace and Android.
Prompt-to-presentation in minutes — we built 15 real decks to find where it still falls short.
Our 200-image benchmark shows Flux matching or beating Midjourney on realism and text rendering.
The open-source model family keeps closing the gap with closed frontier models — self-hosters are the big winners.
The agentic editor now plans and executes larger cross-file changes with fewer stale-context errors.
The free VS Code agent plans, edits, runs, and self-corrects — with any model you connect.
Revenue up 106% year over year, EPS beating estimates, and $442 billion added to the market cap in a day. The AI infrastructure trade just posted the biggest quarter in chip history.
116 companies signed an open letter calling for a global surge in cyber defense — signed by the same industry that spent the year racing model capabilities.
Claude is moving into the world’s dominant CRM. Salesforce calls it Claudeforce; enterprises will call it a land-grab.
Forty-seven states, one landmark number, and product changes to Facebook and Instagram. One of the largest tech settlements ever clears Meta’s biggest legal overhang.
OpenAI just gave IT departments a control plane for ChatGPT Work and Codex. The AI land-grab has officially moved to the admin console.
The signers come from OpenAI, Anthropic, Google and Meta — and the CEOs signed too.
Two releases, five days apart, bracketing the open-weight market from both ends. Alibaba is pressing Meta from the top and the bottom at once.
For the first time, downloads of Chinese open-weight models have overtaken American ones. Stanford’s numbers say the model race is no longer a two-horse sprint.
Every Claude model launched after August 2 embeds machine-readable markers in its output. Detection is becoming a default feature, not an add-on.
V4-Pro is out of preview. The Flash build quietly learned to speak Codex. DeepSeek is betting price-performance is the wedge into agentic coding.
xAI’s coding-focused update ties GPT-5.6 Sol Max on the AA Intelligence Index, ships into Copilot and Microsoft Foundry — and Grok 5 is still waiting.
OpenAI’s August 6 refresh sharpens the paid tier and hands the faster Luna model to every free account. Two tracks, one funnel.
Meta, Anthropic, OpenAI and Google will submit models to government-led evaluations. It is the most concrete US alignment on AI safety this year — and it stops short of law.
A handset that books, buys and coordinates on your behalf. WAIC 2026’s headline act marks the moment agent software left the browser tab.