OpenAI Math Breakthroughs Signal AI’s New Role in Pure Research
OpenAI Math: OpenAI released results on 10 long-standing open problems in mathematics and theoretical computer science, spanning geometry, cryptography, and mor
Murtuza Merchant is the founding editor of Fathom. A blockchain industry analyst by background, he writes on artificial intelligence and on the points where AI and digital assets genuinely meet, which he treats as a narrower set of stories than the sector's enthusiasm suggests. He set up Fathom on the conviction that the difference between a journalist and a content writer matters more as the volume of published words rises, and that the reporters displaced by newsroom cuts were rarely the ones who could not write. As founding editor he covers across beats rather than within one, generally taking the story that sits between two of them. He is most interested in the questions that only become answerable after the announcement cycle has moved on, and in what the money is actually doing as opposed to what it is said to be doing.
OpenAI Math: OpenAI released results on 10 long-standing open problems in mathematics and theoretical computer science, spanning geometry, cryptography, and mor
A leaked Apple memo and fresh hardware rumors confirm the iPhone 18 Pro's biggest upgrades yet, but a potential $200–$300 price hike is raising serious question
AI code review: Coldcard's maker warns AI tools failed to detect the software flaw that let hackers drain 5,200 Bitcoin wallets, raising urgent questions about
6B Model: Liquid AI's LFM2.5-2.6B runs on a single phone chip and beats models up to 18B parameters on key benchmarks, reshaping the edge AI deployment race.
Almost No: Meta posted a record $60.8B in Q2 revenue but AI infrastructure spending hit $31.1B, collapsing free cash flow to just $784M. Here is what happened.
Apple's Q3 2026 earnings topped Wall Street estimates, with record revenue and growing AI income capping Tim Cook's final quarter at the helm.
Kimi K3: A new independent study ran 12 frontier AI models through real production code tasks.
Gemini managed agents: Google's Gemini API Managed Agents now run on the faster 3.6 Flash model with new hooks and triggers, giving developers production-grade
AI Compute War: AMD committed up to $5 billion to deploy two gigawatts of MI450 GPUs for Anthropic. Here is what the deal means for the AI compute…
OKX found itself pulled between Washington-style legitimacy and Asian regulatory pressure this week, adding former New York Governor Andrew Cuomo to its board as South Korea blocked new…