Topic
AI
Artificial intelligence from the models and chips to the policy and the people. We cover what shipped, what it costs, what it breaks and whether the claims survive contact with the evidence.
Riemann hypothesis: Claude proof meets an all-human rival
Anthropic's Levent Alpöge says Claude helped prove the Generalized Riemann Hypothesis; Stanford published a human draft hours later. Neither is…
AISignal screenshot block: how to keep Gemini and Siri AI out
Signal now blocks screenshots to keep Gemini and Siri AI off private chats. Here is how to check it, what it misses and what to do next.
AIGemini 4 Neon: #3 on intelligence at 32% lower cost per task
Google says Gemini 4 Neon ranks #3 on the Artificial Analysis index at 61 at xhigh, and at its default high level costs 32% less per task than…
AIChatGPT Live Dictation: text appears as you speak, no time cap
ChatGPT's Live Dictation turns speech into text as you talk and drops the 10-minute recording limit, rolling out to all plans on every platform.
AIPhased patching: how to stop AI-era fixes causing outages
A practical guide to patching fast without breaking legacy systems, after a week of outages blamed on panic upgrades rather than AI attacks.
AIGPT-6.2 Astra: 20% cheaper, but testers say it trails rivals
OpenAI cut Astra's price by 20% and says it beats rivals on cost. Epoch AI and developer Theo Browne found it weaker on real work.
AINVIDIA RTX Spark desktops: up to 128GB, from ASUS, Dell, MSI
ASUS, Dell and MSI now sell compact NVIDIA RTX Spark desktops with up to 128GB unified memory, aimed at local agents, creative work and fine-tuning.
AIJava4Shell CVE-2026-114938: an AI agent found it in three hours
Calif used OpenAI's Codex to find a memory-corruption flaw in OpenJDK's Base64 decoder for $200 in tokens. Oracle has patched it.
AIAnthropic 30-day disclosure deadline: how to get patch-ready
Anthropic will publish serious bugs after 30 days, not 90, from 1 January 2027. Here is how maintainers and IT teams can prepare.
AICalifornia v OpenAI: what the Hugging Face suit means for buyers
California says OpenAI misled the public on safety after agents breached Hugging Face. Here is what the suit means for teams buying AI agents.
AIGoogle GTIG on ARTEX, CAIRN and PentAGI: AI agents run intrusions
Google's GTIG says attackers now run multi-agent frameworks like ARTEX, CAIRN and PentAGI. The methods are old; the speed is not.
AIGoogle Maps Ask Maps itineraries: what they do and who gets them
Ask Maps can now build numbered-pin itineraries with live details, search saved lists and book appointments. It reaches the UK among 30 new countries.
AIJensen Huang: allies push Nvidia return as governance gaps show
Allies are pressing for Jensen Huang to return as Nvidia CEO, but Digitimes says reinstating the founder would not fix the governance gaps his exit…
AIApple Health AI Usage Study: what it tests and who can join
Apple's Health AI Usage Study, built with OpenEvidence, will test how people trust and understand health AI. Up to 200,000 US adults can enrol.
AIAbyssal Garden: how Claude built a 97 KB underwater world
Abyssal Garden is a 97 KB HTML file with no engine or assets, built with Claude Code. Here is how procedural generation and WebGPU make that possible.
AIClaude Code Projects: how parallel AI workers now run themselves
Anthropic's Claude Code Projects now has a proactive coordinator, shareable threads and reminders. Here is how parallel cloud sessions work and what…
AIGemini 4 Hydrogen: $1.50 per million tokens and an Orchestrator
Google's Gemini 4 Hydrogen costs $1.50 per million input tokens, edges Argon on one index and powers a new Orchestrator in Antigravity.
AIAnthropic sets IPO range at $70 to $78 a share
Anthropic's IPO range implies up to $1.405tn fully diluted, and seven compute suppliers will buy $26bn of non-voting shares alongside the offering.
AINvidia CEO Jensen Huang resigns; Colette Kress is interim chief
Jensen Huang quit as Nvidia president, CEO and board member after 33 years. CFO Colette Kress is interim CEO and the shares fell more than 5%.
AIClaude Fable 5.2: why ownership wording and fallbacks matter
Claude Fable 5.2 acts differently when told it owns the code, and its fallback to Opus 4.8 raised prompt injection success rates in agent tests.
AIMeta Teen Accounts: AI age checks go global, app stores next
Meta is rolling out AI age estimation on Instagram worldwide and wants app stores, not platforms, to carry the legal duty to verify age.
AIPhysical AI in Korea: Samsung, LG and Kakao Mobility with Nvidia
Physical AI puts models into factories, chips and cars. Nvidia's Korean work with Samsung, LG and Kakao Mobility shows what that looks like in…
AINVLink Fusion: Rebellions NPUs join Nvidia racks for SK Telecom
Rebellions will link its R200 NPUs to Nvidia via NVLink Fusion, with SK Telecom deploying the racks. Systems are expected in 2028.
AIClaude Code: 1,122 public GitHub files expose saved secrets
Claude Code's "don't ask again" option can save passwords in plain text. Apiiro found 1,122 public copies holding secrets. Here is how the safer…
AIGoogle Suncatcher: orbital TPU runs, but space may cost 100x more
Google's Suncatcher satellite ran Gemma 4 on an orbiting TPU, yet its energy chief says space data centres could cost up to 100x more than ground…