|
📖 Read In Depth
|
Google plans to invest up to $40B in Anthropic
Google is committing up to $40B in Anthropic, structured partly as a circular deal through TPU compute purchases. The HN comments reveal interesting subtext: Anthropic was apparently becoming severely capacity-constrained and had to sign somewhat adverse contracts with both Amazon and Google in quick succession — suggesting the investment is as much about locking in compute supply as it is a pure bet on the company. The deal reframes Anthropic's independence story significantly.
hn/Best Stories
|
Amateur armed with ChatGPT solves an Erdős problem
An amateur mathematician used ChatGPT to solve a 60-year-old Erdős problem — the HN thread includes the actual prompt used, which is careful and adversarial (no internet, demand unconditional proof or disproof). This is a concrete data point on what LLMs can do in formal reasoning when steered well, not vibe-coded. Worth reading for the actual math and methodology.
hn/Best Stories
|
The AI industry is discovering that the public hates it
Examines growing public hostility to the AI industry — not just from Luddites but from people who feel the technology is being imposed on them rather than built for them. The HN comments are a useful counterpoint, with developers split between genuine productivity gains and concerns about vibe-coded garbage proliferating. Good reality check on the gap between industry self-perception and public reception.
hn/Best Stories
|
Leaked Code for Anthropic’s Claude Code Tests Copyright Challenges in A.I. Era
Leaked source code for Claude Code raises questions about whether copyright law can even function when AI makes reproduction trivially fast and attribution ambiguous. This sits at the intersection of IP law, AI training data, and software ownership — and has direct implications for anyone building or licensing AI coding tools.
nyt/Technology
|
How Elon Musk Used SpaceX to Benefit Himself and His Businesses
NYT investigation into how SpaceX has functioned as a financial instrument for Musk personally — providing loans and cross-subsidizing his other ventures. With a SpaceX IPO approaching, understanding the entanglement between Musk's personal finances and company governance matters for anyone tracking the broader tech-power-finance nexus.
nyt/Technology
|
Why has there been so little progress on Alzheimer's disease?
Freakonomics digs into why Alzheimer's research has been so unproductive despite massive investment — touching on the amyloid hypothesis failure, institutional incentives, publication bias, and how scientific monocultures form. A case study in how systems that should produce knowledge can systematically fail, with obvious parallels to other fields including AI research.
hn/Best Stories
|
Replace IBM Quantum back end with /dev/urandom
Someone replaced IBM Quantum's backend with /dev/urandom and it still 'recovered' a 17-bit elliptic curve key — the same result that won a 1 BTC prize for 'the largest quantum attack on ECC to date.' A delightful empirical skewer of quantum computing hype: when circuits are longer than the coherence time, the quantum computer is already producing noise indistinguishable from random. The HN comments connect this to a prior April Fool's paper that made the same point theoretically.
hn/Best Stories
|
|
âš¡ FYI
|
Elon Musk and Sam Altman’s Epic Fight Heads to Court
The Musk vs. Altman/OpenAI jury trial is starting, with Musk seeking billions in damages. Outcome could reshape OpenAI's governance and the broader AI race — particularly around whether OpenAI's nonprofit-to-for-profit conversion was a breach of its founding commitments.
nyt/Technology
|
OpenAI Unveils Its New, More Powerful GPT-5.5 Model
OpenAI released GPT-5.5, described as more powerful and with a notably more open approach to cybersecurity than competitors. The timing — coinciding with the Musk trial and Sam Altman's pressure to monetize — makes the model release politically as well as technically significant.
nyt/Technology
|
Google to invest up to $40B in Anthropic in cash and compute
Duplicate of the Bloomberg story on Google-Anthropic investment — skip in favor of the Bloomberg piece, but noting the TechCrunch version adds context on cash vs. compute breakdown and TPU capacity commitments.
hn/Best Stories
|
How Do You Measure A.I. Firms’ Gargantuan Energy Plans? In ‘Bragawatts.’
Coins the term 'bragawatts' for AI companies' unverified boasts about power capacity. The piece documents how datacenter energy announcements have become competitive signaling with little grounding in actual contracted supply — useful lens for evaluating hyperscaler and AI infrastructure claims.
nyt/Business
|
Trump fires NSF's oversight board
Trump fired the NSF's National Science Board, removing the independent oversight body that sets NSF's priorities and approves its budget. Combined with earlier cuts to federal research funding, this is a significant structural attack on U.S. basic science infrastructure — with downstream consequences for the academic pipeline that feeds ML research.
hn/Best Stories
|
Tim Cook Will Step Down as Apple C.E.O.
Tim Cook is stepping down as Apple CEO, replaced by John Ternus, the head of hardware engineering. Ternus's background is deeply technical, which may signal a shift in Apple's priorities toward hardware and silicon after years of software/services focus under Cook — relevant to the broader Apple Intelligence and chip strategy story.
nyt/Technology
|
U.S. Sanctions Zigzag in New World of Economic Warfare
The Trump administration is running a contradictory sanctions regime — simultaneously pursuing and easing sanctions on Russia and Iran based on oil price management rather than coherent geopolitical strategy. The piece frames economic warfare as increasingly improvisational, with real consequences for supply chains and energy markets that affect everything from jet fuel to chip manufacturing inputs.
nyt/Business
|