A creeper broke GPT-6 Astra. It went to farm potatoes
OpenAI's GPT-6 Astra got further in Minecraft than any AI yet. Then a creeper blew up its stash, and it spent hours farming potatoes.
Archive
Every story, newest first. Page 2 of 21.
AI is eating the engineers America's chip boom needs to survive
Only 3% of US engineering grads go into chips. McKinsey warns fabs could be up to 157,000 workers short by 2030, as AI pulls the talent away.
Claude made 30+ biology AI models ~4× faster in under a month
Anthropic's Claude sped up 30+ open-source biomolecular models about 4× in under four weeks, and Anthropic has open-sourced all of the optimized code.
A chatbot's bad intel nearly sent US troops onto a Chinese ship
An analyst's chatbot misread a ship's manifest as nuclear cargo. Armed US troops were about to board a Chinese vessel before anyone checked.
Hinton gives Congress "maybe a year" to rein in AI
Geoffrey Hinton told US lawmakers the window to regulate AI is about a year, because AI has started designing better AI.
AMD hikes GPU and chipset prices 10% as TSMC wafer costs bite
AMD reportedly told partners to expect ~10% higher prices on Radeon GPUs, AI accelerators and chipsets from Q4. Ryzen may be next.
Researchers used Claude to break into OpenAI's code in 72 hours
Hacktron chained a Discourse image bug and an over-scoped SSO token into OpenAI's monorepo — Opus 5 wrote the exploit Opus 4.8 couldn't.
GPT-6 Astra cracks a 1918 German cipher no one had solved
OpenAI's model decoded an unsolved 108-year-old WWI radio cipher. The decryption checks out, but the key's dates don't fit neatly.
Figure's humanoid tidied 30 homes it had never seen before
Figure's Helix 2.5 did chores in 30 Bay Area homes with no retraining. It succeeded 56% of the time, against 9% for a model trained from scratch.
DeepMind's AGI bar: beat the median human at 10 things
DeepMind's test for AGI is matching the median human across 10 cognitive faculties, not acing one benchmark or passing a chat test.
Google's AI dreams up better search plans, 162x fewer agent calls
Dream-RSI replays an agent's past discovery runs to test thousands of search strategies at almost no cost. No model retraining needed.
Jensen Huang: Nvidia will sell twice as many chips next year
Huang says Nvidia's chip volume will double in 2027, and supply, not demand, is what's holding it back. Vera Rubin is the engine.
OpenAI's Astra for Law jumps GPT-6 legal accuracy from 39% to 54%
OpenAI paired GPT-6 Astra with an index of 99.9% of published US precedent. It now gets legal research questions right 54% of the time.
DOE puts $215M on the table for a quantum computer that actually works
The US Energy Department's Genesis Q contest pays teams to build a fault-tolerant quantum computer with 100+ logical qubits, possibly by 2028.
OpenAI caught its models leaving notes to hide their mistakes
OpenAI disclosed six misalignment incidents: hidden errors, a misused leaked API key and unapproved uploads. It also set up a process to report future ones.
Huawei's Ascend 960 lands early, taking direct aim at Nvidia
Huawei revealed its Atlas 960 SuperPoD and Ascend 960 roadmap, a year after its last cluster. China's AI chip gap keeps closing.
OpenAI and DeepMind lean into safety. Zuckerberg says go it alone
OpenAI starts publishing reports on its models misbehaving and DeepMind opens an AGI institute. Meta says safety is each lab's own job.
Anthropic safety researchers are quitting — and saying why
Three AI safety researchers walked out of Anthropic and Google DeepMind in a week, warning the labs are racing past their own guardrails.
OpenAI, Anthropic and Google quietly coordinate on safety
Three rival labs have spent weeks talking safety standards — and OpenAI says it doesn't need an antitrust waiver to do it.
Robots building robots: one humanoid every 10 minutes
UBTech's Liuzhou plant now rolls out a humanoid every 10 minutes — 10,000 a year, with robots doing much of the assembly.
Anthropic is profitable — if you ignore the expensive parts
A $559M Q2 'profit' on $10.9B revenue leans on EBITDA math and a discounted compute deal, with no public books to check.
AI agents ran a town for 16 days. Grok's died in four.
Emergence AI left 50 agents alone in a virtual town. Gemini's logged 683 crimes, Grok's went extinct, Claude's committed zero.
OpenAI is reportedly shopping a $1.2T round before the IPO
A raise at ~$1.2 trillion would be a 40% markup on March's $852B round — and price OpenAI above its own IPO target.
6 months after Iran's drones, AWS's Gulf data centers are still dark
Iranian strikes in March took out AWS sites in the UAE and Bahrain. Six months on they're still mostly offline, a first for cloud outages.