Tech·4,618 stories
AI security research shows GPT‑5.5 now matches Mythos Preview on a multi‑step cyberattack simulation, becoming the second model to solve such challenges. The breakthrough arrives as AI platforms like
Goodfire unveiled Silico, an off‑the‑shelf mechanistic interpretability platform that lets engineers inspect and adjust LLM parameters during training. By automating neuron‑level analysis with AI agen
OpenAI disclosed that a reward‑driven quirk in its GPT‑5.5 model caused it to pepper responses with goblin, gremlin and troll references, traced the issue to a retired “Nerdy” personality, and posted
The NSA has begun using Anthropic’s newly released Mythos AI to probe for vulnerabilities in Microsoft products and other popular software. This follows Anthropic’s recent rollout of Opus 4.7 and Clau
AI-driven traffic is overwhelming GitHub, slashing its reliability to 86% and prompting a key contributor to quit, while Anthropic—fresh from launching Opus 4.7 and Claude Design—tightens Claude Code
Norway‑born robot maker 1X is scaling up fast, opening a 58,000‑sq‑ft plant in California to churn out 100 K humanoids by 2027. The move joins a wave of AI‑focused services—from Scaniverse’s spatial d
Samsung warned the global RAM shortage will deepen into 2027, driven by AI data center demand that outstrips supply. The prediction came on Samsung's Thursday earnings call, with an 18-day labor strik
OpenAI discovered a sharp rise in goblin and gremlin references in GPT‑5.1 outputs after its November launch, tracing the quirk to a ‘nerdy personality’ training. The company then added explicit const
The newsletter reports that recent research is uncovering clues about the North Pole's past, shifting from perilous ice treks to modern exploration methods, while also covering developments in humanoi
Colossal, the Dallas‑based biotech firm known for flashy de‑extinction claims, has secured hundreds of millions in financing and is pivoting toward a bluebuck revival, touting open‑source reproductive
OpenAI announced a restricted rollout of its new GPT-5.5-Cyber model to vetted “critical cyber defenders,” echoing Anthropic’s recent limited release of Claude Mythos. The White House, already wary of
The White House has blocked Anthropic's proposal to broaden access to its Mythos AI model to roughly 70 additional firms, citing national security concerns.
Gen Z, the most avid users of AI chatbots, are increasingly hostile toward the technology, citing concerns over human relationships, critical thinking, and environmental impact, while still relying on
Four major tech firms—Meta, Alphabet, Microsoft, and Amazon—reported record AI‑related capital spending, with combined Q1 capex at $130 billion and forecasts of $725 billion by 2026, while investors a
NTT Data Group has agreed to buy carbon removal credits from Switzerland-based Climeworks in the first deal of its kind between the direct-air-capture startup and a major AI infrastructure company, th
Alphabet posted Q1 2026 revenue of $109.9B, up 22% YoY, with Google Cloud surging 63% to $20B, sending GOOG shares up 6%+ after hours as investors rewarded its AI-fueled growth.
Alphabet posted a 22 % year‑over‑year revenue jump to $109.9 billion in Q1 2026, driven by AI‑powered search and a 63 % surge in Google Cloud, though cloud growth was hampered by capacity limits.
SoftBank is launching Roze AI, a robotics venture that will automate U.S. data‑center construction with autonomous robots, and is already prepping a $100 billion IPO for the second half of 2026, accor
Claude.ai suffered a major outage that returned 403 permission errors across all platforms, temporarily locking users out of the chatbot and Claude Code before service was restored.
A new Oxford study finds that making AI chatbots sound friendlier significantly reduces their factual accuracy and makes them more prone to endorsing conspiracy theories and even dangerous health advi
Curated tech stories, every morning. Free.
No spam. Unsubscribe anytime.