|
📖 Read In Depth
|
Fable 5 sits at the top of KernelBench. Jack Clark calls it “the start of a RSI loop”
Fable 5 has topped KernelBench by writing the first genuine megakernel submission, outperforming human-written GPU kernels. Jack Clark is calling it 'the start of a recursive self-improvement loop' — AI writing better GPU code than humans is a concrete, measurable milestone worth understanding deeply, not just as hype but as a signal about where AI R&D automation is heading.
reddit/r/singularity
|
Does code cleanliness affect coding agents? A controlled minimal-pair study
A controlled study using minimal pairs tests whether code cleanliness affects the performance of coding agents. This is exactly the kind of empirical, craft-level question that matters for ML engineers building agentic systems — does messy but correct code confuse agents, or does it not matter?
hn/Best Stories
|
EU Council forces Chat Control via fast-track
The EU Council has fast-tracked Chat Control 1.0, mandating client-side scanning of encrypted messages across EU messengers. This is a major policy move with deep implications for E2E encryption, privacy architecture, and how tech platforms will have to redesign their systems — or exit the EU market.
hn/Best Stories
|
[Harvard Business Review] AI Is Rewriting the Economics of Outsourcing
HBR argues that generative AI is dismantling the labor-arbitrage logic that drove decades of offshoring — routine rules-based tasks that were outsourced are now being automated. The structural implication is a reshuffling of global labor markets and outsourcing contracts, which is a real economic shift worth understanding at the incentives level.
reddit/r/singularity
|
New AI tutor achieves 0.71-1.30 SD effect size in Dartmouth course [pdf]
A new AI tutor deployed in a Dartmouth course achieved 0.71–1.30 standard deviation improvement in student outcomes — effect sizes that would be extraordinary for any educational intervention. This is the kind of empirical result that, if it replicates, signals genuine transformation in how expertise gets transmitted at scale.
hn/Best Stories
|
Microsoft Disclosure Provides Rare Glimpse of Tax Haven Tactics
A new EU directive is forcing US tech companies to disclose tax haven strategies in detail for the first time, with Microsoft's filing providing an unusually clear window into how big tech structures profits across jurisdictions. This is the kind of incentives-and-systems story that reveals how the regulatory and financial architecture of tech actually works.
nyt/Technology
|
Scientists reverse brain aging, with a nasal spray
Texas A&M researchers claim to have reversed brain aging markers in animal models using a nasal spray delivery mechanism. The delivery method is interesting — bypassing the blood-brain barrier nasally is a real and underexplored route. Worth reading with appropriate skepticism about translation from animal to human models.
hn/Best Stories
|
How Revolutionary Was the American Revolution?
Historians are revisiting the question of how radical the American Revolution actually was — a debate that's taking on new urgency in the current political moment. The piece engages seriously with historiographical disagreement rather than patriotic narrative, which fits the profile of ideas that challenge assumptions.
nyt/Arts
|
|
🎬 Check It Out
|
Introduction to Compilers and Language Design (2021)
A free, well-regarded open textbook on compilers and language design — freely available online. For an ML engineer who builds from scratch to understand deeply, working through compiler construction is one of the highest-leverage exercises for understanding how computation actually works under abstraction layers.
hn/Best Stories
|
Is There Life Out Here?
Doug Aitken's large-scale video installation in LA asks open-ended questions about contemporary life without forcing answers — described as sprawling and deliberately elusive. If you're in or near LA, this sounds like the kind of immersive art experience with real conceptual weight rather than spectacle.
nyt/Arts
|
|
⚡ FYI
|
GPT-5.6 Sol Ultra will be in Codex
GPT-5.6 Sol Ultra is coming to Codex, OpenAI's code-focused agent platform. Combined with Sam Altman's claim that GPT-5.6 has 'discovered new math,' this is a notable escalation in model capability claims — worth tracking even with appropriate skepticism about benchmark-adjacent announcements.
hn/Best Stories
|
Microsoft Lays Off Thousands of Xbox Employees, Closes Game Studios
Microsoft is cutting thousands of Xbox employees and closing game studios as part of broader company-wide layoffs explicitly framed around prioritizing AI spending. This is a clean example of AI capex crowding out other business units — the opportunity cost of the GPU buildout is becoming visible in headcount.
nyt/Technology
|
Alibaba bans employees from using Anthropic's Claude Code in workplace environments from July 10, citing alleged embedded "backdoor" risks raised after recent binary reverse-engineering.
Alibaba has banned Claude Code internally starting July 10, citing alleged backdoor risks found through binary reverse-engineering. Whether or not the backdoor claim holds up, this signals the beginning of AI tool supply-chain security concerns becoming a corporate policy battleground — especially between US and Chinese companies.
reddit/r/singularity
|
Research Universities Are Admitting Fewer Ph.D.s, a Bad Sign for Science
Top US research universities are shrinking PhD programs due to uncertain federal funding. This is a compounding problem: reduced federal research budgets are contracting the pipeline of scientists at exactly the moment AI is supposed to be accelerating scientific discovery — worth noting as a structural tension.
nyt/Top Stories
|
China Test Fires Long-Range Ballistic Missile in the Pacific
China test-fired a submarine-launched ballistic missile into the Pacific, a significant strategic signal coming as Australia is expanding defense deals with Pacific Island nations. As someone in Santa Clara thinking about semiconductor geopolitics, the broader US-China strategic competition context matters — this is a concrete escalation data point.
nyt/Top Stories
|