|
📖 Read In Depth
|
Claude Opus 4.8
Anthropic released Claude Opus 4.8, described as a 'modest but tangible improvement' on 4.7 with enhanced agentic capabilities. Notably, the announcement teases 'Project Glasswing' and a model called Claude Mythos Preview — a higher-capability class currently in limited cybersecurity use — suggesting Anthropic is quietly building above the current Opus tier. The honest framing of incremental improvement is a refreshing contrast to typical launch hype.
hn/Best Stories
|
The $900 Billion Giant: How Anthropic Got So Big, So Fast
Anthropic raised $65B in a Series H at a $900B post-money valuation, surpassing OpenAI's $730B. The DealBook piece goes into the actual numbers — revenue trajectory, burn rate, and the structural headwinds — making it more substantive than the breaking news version. For someone thinking about AI moats and capital dynamics, this is the right read.
nyt/Business
|
Disagreement among frontier LLMs on real-world fact-checks
Empirical study measuring how much frontier LLMs disagree with each other on real-world fact-checks — not synthetic benchmarks, but actual claims. High disagreement rates across models raise questions about reliability for any single-model fact-checking pipeline, and the dataset is publicly explorable. Simon Willison's comment on HN adds useful methodological context.
hn/Best Stories
|
Building durable workflows on Postgres
Makes the case that Postgres is sufficient for durable workflow execution — the problem space that Temporal, Conductor, and similar systems address. Directly relevant to anyone building agentic systems or long-running pipelines and weighing whether to add infrastructure complexity versus leveraging a database they already trust.
hn/Best Stories
|
Tech CEOs are apparently suffering from AI psychosis
Examines the pattern of executives dramatically overclaiming AI capabilities and making organizational decisions (headcount, roadmap, strategy) based on demos rather than production reality. The HN comments add texture: one notes that AI aligns well with existing management levers (fast, directable, no self-preservation instinct), which explains the appeal independent of actual ROI.
hn/Best Stories
|
In Strait of Hormuz Standoff, U.S. and Iran Deploy Economic Warfare
Deep dive into the U.S.-Iran Strait of Hormuz standoff as economic warfare — neither hot conflict nor resolution, with both sides attempting to strangle the other's economy. Explains the structural stalemate clearly and is more analytically useful than the daily oil price updates.
nyt/Top Stories
|
A.I. Doesn’t Have to Mean Layoffs
Case study of Schneider Electric deliberately deploying AI to augment manufacturing workers rather than replace them — a concrete counterexample to the default automation-equals-displacement narrative. Worth reading for the specific organizational choices made and whether they hold up as a model or are idiosyncratic to French labor law and culture.
nyt/Business
|
Using the results of erdos unit distance problem,mathematicians disprove a long standing mathematics conjecture.
Mathematicians used results from the Erdős unit distance problem to disprove a long-standing conjecture in combinatorics. Links to the arxiv paper. Light on context in the reddit post but the underlying math story — an old open problem's resolution enabling a downstream disproof — is the kind of cross-field result worth knowing about.
reddit/r/singularity
|
|
🎬 Check It Out
|
The first time I watched Mind Game (2004) a year ago, I thought it was a wild, overwhelmingly creative romp. The second time I watched it this week, it cemented as one of my favorite movies of all time.
Mind Game (2004) is a maximalist animated film by Masaaki Yuasa — formally experimental, narratively chaotic in the best way, and genuinely unlike anything else in animation. The thread is a passionate second-watch endorsement from someone whose first watch was overwhelmed by the stimulus. Available on MUBI. If you have any appetite for challenging animation, this is the real thing.
reddit/r/TrueFilm
|
|
âš¡ FYI
|
AMD pulls a bait-and-switch on Linux users with Vivado licensing changes
AMD changed licensing terms for Vivado (their FPGA toolchain) in a way that effectively breaks Linux workflows that relied on the free tier. Relevant to anyone doing hardware development or following the chip ecosystem — another data point in the ongoing story of toolchain lock-in in the semiconductor space.
hn/Best Stories
|
Blue Origin Rocket Blows Up on Florida Launchpad During Test
Blue Origin's New Glenn rocket exploded on the launchpad during a static fire test in Florida. This has downstream consequences: Amazon's Project Kuiper satellite constellation is directly impacted, and AST SpaceMobile dropped 10%+ on the news. A costly setback for Bezos's space ambitions.
nyt/Top Stories
|
Uber president says AI spending is getting ‘harder to justify’
Uber's president says the company burned through its entire annual AI budget in four months and isn't seeing meaningful ROI. Paired with Amazon scrapping its internal AI usage leaderboard, this is a pattern worth tracking: enterprise AI spend is hitting a 'prove it' moment faster than expected.
reddit/r/investing
|
Volkswagen blocks Home Assistant by requiring client assertion
Volkswagen is blocking third-party Home Assistant integrations by requiring client certificate assertions — effectively locking users into their first-party app. A small but telling example of how car manufacturers are using software to eliminate interoperability, relevant to the broader BBC article on car data collection.
hn/Best Stories
|