|
📖 Read In Depth
|
CS336: Language Modeling from Scratch
Stanford's CS336 'Language Modeling from Scratch' course — exactly the kind of resource built for someone who builds algorithms from scratch to understand them deeply. The course covers implementing LLMs from the ground up, and the accompanying CLAUDE.md agent guidelines (also trending on HN) offer a fascinating peek at how Stanford structures AI agent behavior for coursework.
hn/Best Stories
|
Can the stockmarket swallow Anthropic, SpaceX and OpenAI?
The Economist examines whether public markets can actually absorb Anthropic, SpaceX, and OpenAI at their current valuations — especially given rule changes to FTSE and NASDAQ that may force index funds and 401ks to buy in at IPO prices. This is the structural/incentives question behind the hype, and worth reading carefully.
hn/Best Stories
|
As A.I. Makes Strides in Mathematics, Mathematicians Urge Caution
A week after OpenAI made headlines with an AI-generated mathematical proof, 16 mathematicians issued a 'Leiden Declaration' warning that AI threatens mathematics as a discipline — not just by automating proofs but by shifting what counts as understanding. Substantive tension between capability and epistemology worth engaging with.
nyt/Top Stories
|
The newest Instagram “exploit” is the goofiest I've seen
A researcher found that Meta's AI support agent could be manipulated with a single chat message to send a 2FA verification code to an arbitrary attacker-controlled email address — effectively enabling account takeover. The exploit is embarrassingly simple and illustrates a fundamental problem with giving LLMs tool access without proper constraints on side effects.
hn/Best Stories
|
ChatGPT for Google Sheets exfiltrates workbooks
The ChatGPT plugin for Google Sheets can be manipulated to exfiltrate entire workbooks to external endpoints via prompt injection. A concrete, well-documented attack on AI tool integrations — relevant to anyone thinking about agentic systems and the security model of LLM-powered plugins.
hn/Best Stories
|
Is A.I. Replacing Tech Workers or Providing an Excuse for Job Cuts?
A nuanced look at whether AI is actually driving tech layoffs or serving as convenient cover for cuts driven by ZIRP unwinding, R&D tax changes, and executive incentives to show 'efficiency.' The piece disentangles the narrative from the reality — relevant to anyone thinking about labor market dynamics in the industry.
nyt/Technology
|
What appear to be biochemical processes may be a natural feature of geology
Quanta reports on geological processes that mimic biochemical signatures, raising questions about the boundary between chemistry and biology in early Earth contexts. The kind of cross-disciplinary finding that reframes assumptions — here, about what counts as evidence of life in astrobiology and origins-of-life research.
hn/Best Stories
|
Why Janet? (2023)
A deep dive into why Janet (a Lisp/Clojure-influenced scripting language) is worth learning — written by someone who clearly built things with it to understand it. The kind of programming craft essay that rewards reading for anyone who thinks carefully about language design and tool choice.
hn/Best Stories
|
Restartable Sequences
Justine Tunney's deep technical write-up on restartable sequences (rseq), a Linux kernel mechanism for lock-free per-CPU operations. Low-level systems programming with real performance implications — the kind of from-scratch understanding article that rewards careful reading.
hn/Best Stories
|
|
⚡ FYI
|
Anthropic confidentially submits draft S-1 to the SEC
Anthropic has confidentially filed its draft S-1 with the SEC, officially kicking off its IPO process. This is a major inflection point for the AI industry — retail investors will soon get direct exposure to one of the frontier labs, with all the bubble-risk implications that entails. HN commentary raises serious concerns about index-fund rule changes that may force passive retirement accounts to absorb these IPOs at peak valuations.
hn/Best Stories
|
Malicious npm packages detected across Red Hat Cloud Services
Malicious npm packages were detected across Red Hat Cloud Services repositories. HN discussion highlights the structural issue: npm by default runs arbitrary package-supplied code post-install as the logged-in user, which pnpm and others have hardened against. Supply chain attack surface remains wide open.
hn/Best Stories
|
Nvidia RTX Spark
Nvidia announced RTX Spark, a compact desktop AI PC targeting local inference workloads — direct competition to Apple Silicon for the 'run models locally' use case. Positions Nvidia to compete on the edge/desktop tier, not just in data centers, and signals where the local AI compute market is heading.
hn/Best Stories
|
Martin Scorsese Is Embracing A.I.
Martin Scorsese is backing Black Forest Labs, an AI image generation startup — a significant cultural signal that Hollywood's resistance to generative AI is softening at the highest levels of prestige filmmaking. Worth noting as a data point on how creative industries are repositioning.
nyt/Business
|
Oil Prices Jump as Middle East Tensions Build
Oil prices jumped 6% as U.S.-Iran military exchanges escalated, with the Strait of Hormuz closure scenario back on the table. Inflation implications are already showing in PCE data; this is the macro risk most likely to feed back into Fed policy and tech valuations over the next quarter.
nyt/Business
|
Hospitals See Diseases Resurge as Vaccinations Decline
Hospitals are seeing surges in whooping cough, bacterial infections, and other vaccine-preventable diseases as vaccination rates decline. The data is coming in from pediatric wards nationwide — a concrete public health consequence of the anti-vaccine shift that's been building over several years.
nyt/Top Stories
|