Unsupervised Studying NO. 473 (STANDARD EDITION) – Digital Digest

Hello! I hope your week’s beginning properly,

1. I’m now utilizing the brand new Limitless.ai Pendantand it’s positively the very best AI {hardware} accent I’ve used up to now. Principally real-time synched transcription of textual content, with a full API and a clear cell/internet app. It’s additionally straightforward to show off to have delicate conversations. I’m providing them a really low cost safety evaluation as a result of I’d wish to see them succeed and never get incidented out of existence. No referral hyperlink, however right here’s the product: THE LIMITLESS PENDANT

2. I’ve a query on knives—particularly knife units and knife sharpening. I’ve at all times been a knife man, so now that I’m going to begin cooking I need to have top-tier knives. I presently have a set of Globals as a result of they’re universally thought-about “not unhealthy”, however plenty of individuals are saying to go German. So 1) What do you suggest for high knives by way of design/metal/efficiency, and a pair of) what do you suggest for high instruments/devices for sharpening them? GIVE KNIFE ADVICE

3. I used to be fallacious concerning the Anthropic CEO’s assertion that 99% of code would quickly be written by AI. I heard it as “99% of present builders can be utilizing AI”, and I mentioned he was fallacious. However he didn’t say that. He mentioned 99% of code. Which means, new builders making new issues utilizing instruments like Cursor. I corrected this and gave Richard Stiennon public credit score for correcting me on LinkedIn and on X, and he mentioned probably the most superb and unhappy factor to me: “That is the primary time this has ever occurred to me since being on the web.”

4. I made my first steak and acquired some new Atkins protein shakes that ought to have far much less plastic in them. Thanks once more to everybody who responded to the cooking plea.

5. Wild and Vibe are like each different phrase on the web proper now.

Had one other fantastic dialog with ThreatLocker. On this one we talked concerning the full product suite, the Mac agent, and a bunch of different subjects. ^SPONSORED

Microsoft’s March 2025 Patch Tuesday fastened 57 flaws, together with six actively exploited zero-days and a number of NTFS vulnerabilities involving malicious VHD information. LINK

Apple patched a crucial zero-day WebKit vulnerability that was exploited in a extremely focused assault in opposition to particular people utilizing older iOS variations. LINK

China will power all AI-generated content material to be clearly labeled for customers beginning September 1, 2025, together with textual content, pictures, audio, and video. I typically envy how briskly they will transfer. LINK | CAC ANNOUNCEMENT

Google is trying to buy Wiz for $33B, based on folks conversant in the deal. So first Mandiant, now Wiz. My query is the one my buddy Jeremiah Grossman at all times asks, “Are we safer but?” LINK

Chinese language hackers are utilizing TinyShell-based backdoors to show Juniper Networks routers into long-term espionage platforms that conceal from monitoring. LINK

Juniper Networks has patched a vulnerability utilized by Chinese language hackers to deploy backdoors on routers going again to mid-2024. LINK | CVE DETAILS | SECURITY ADVISORY

Cisco simply patched a crucial DoS vulnerability that may crash BGP on its IOS XR routers with a single message. LINK

Lazarus Group has planted six new malicious NPM packages designed to deceive builders and steal cryptocurrency pockets information. LINK | SOCKET BLOG POST

Trump warned Iran that they are going to be held “totally accountable” for Houthi actions, and he launched new airstrikes in opposition to Yemen. LINK

Paul Millerd thinks AI will quickly allow “vibe writing” the place authors can simply riff on concepts and preserve their type whereas eliminating friction factors that preserve them from writing. I feel that’s 100% true, however quickly that’ll simply be vibe dictation and the AI writes the entire thing. I feel the writing itself is beneficial. LINK

No person (supposedly) is aware of what an agent is, however I feel the hype is exaggerated. Anybody constructing brokers know what one is, even when they will’t outline it. I really feel just like the folks most involved concerning the lack of an ideal definition are on the sidelines watching. LINK

Monte Carlo information predicts that true information + AI breakthroughs will observe the identical sample as earlier tech shifts, which is requiring enterprise-grade observability earlier than widespread adoption occurs. I feel that’s proper, however AI may speed up the observability. LINK

Jiachen Zhu proposes a easy solution to make transformers work with out layernorm, probably making AI fashions quicker and easier. LINK

Douglas Hofstadter says AI-generated content material, like GPT-4’s imitation of his writing, sounds fully pretend and hole. LINK

Anthropic co-founder Mike Krieger says they’re specializing in specialised AI instruments like Claude Code moderately than making an attempt to make Claude as mainstream as ChatGPT. I just like the strategy however I feel they need to be centered on Agent Orchestration. LINK

Notepad is getting AI summaries, and that’s a step too far. LINK

Zoom is getting AI Agent capabilities that may schedule conferences, create docs, and even let you know when your colleagues can be on the workplace. LINK

The pay raises you get from altering jobs has almost disappeared, shrinking from 2.1% in 2023 to only 0.2% final month. LINK

Apple is bettering texting with Android in iOS 19. They’re going to RCS model 3.0, including end-to-end encryption, message enhancing, customized reactions, and the power to recall texts. LINK

Gallup’s newest survey says America’s psychological and bodily well being scores have dropped to 24-year lows, with the pandemic accelerating declines that started round 2013. LINK

Forbes did a brand new research that claims AI will crush tech and finance jobs, however human-centered careers like instructing and nursing will develop considerably. I feel each will develop—the human model and AI-based variations. LINK

The typical period of unemployment within the US has almost doubled from 12 weeks in 1990 to 21.6 weeks in 2024. LINK

Harvard College is increasing its monetary support, making tuition free for households incomes as much as $200K, and protecting all prices for these underneath $100K. They’re clearly making an attempt to counter the elitism narrative, however I simply don’t suppose one of these greater schooling has a lot time left. LINK

Researchers discovered that shingles vaccines may scale back dementia danger by as much as 20%, suggesting viral infections may very well be driving Alzheimer’s illness. LINK

Trump is threatening Europe with a large 200% tariff on wine and champagne if the EU would not take away its 50% tariff on American whiskey. LINK

NASA’s Lunar Reconnaissance Orbiter captured pictures of the tipped-over Athena moon lander sitting inside a small crater close to the lunar south pole. LINK | IMAGES

Gen Z People haven’t got sufficient saved to cowl a single month of spending. LINK

Daniel Kahneman ended his personal life by means of assisted suicide in Switzerland at age 90. Looks like he observed indicators of cognitive decline and determined he didn’t need to undergo that. LINK | DISCUSSION

New information from Our World in Information exhibits that comparatively small donations to efficient charities can dramatically enhance lives within the poorest components of the world. LINK | DISCUSSION

Gradual Disempowerment
Jan Kulveit and colleagues argue that AI may regularly disempower people by means of incremental development with out requiring any sudden functionality jumps or coordinated betrayal. Principally it simply strips away our competence slowly and steadily.

By the best way—that is the good solution to launch a paper! I feel all papers ought to have like a weblog touchdown web page, with a 5-level abstract, an article model, a video abstract, after which a hyperlink to the complete paper. I’m going to make this template. LINK | PAPER

All the things Paywalling
I’m noticing I’m actually uninterested in media websites charging subscriptions. Like ¼ of the stuff I need to hyperlink to requires a subscription now. I don’t thoughts paying people, however there’s no assure these huge magazines are going to outlive or retain the folks I need to hear from. I’m very anxious about equality when all the very best thought and information sources are behind paywalls, and the very best AI can also be costly. So “technically” you are able to do every thing free, however the high quality and expertise is method worse. It is a future to purposely attempt to keep away from.

AI “Nah” Buttons
Anthropic’s CEO, Dario Amodei, threw out an intriguing thought about giving AIs a button to stop duties that they discover disagreeable. However doesn’t that require subjective experiences although? Which means…consciousness? I feel it does. Adherence to coverage, certain, but when one thing has a nasty “vibe” to an AI, I feel it’s aware. LINK

Profession Recommendation in 2025 — Will Larson shares his uncooked ideas on why tech careers really feel quite a bit much less enjoyable in 2025. He talks about valuations, the push to AI, and a bunch of different factors. Actually good. LINK

After Intelligence — A sport that prompts you to think about our future after people aren’t the neatest beings on Earth. I really like how stuff like this may be constructed so shortly. LINK

Muller’s Ratchet on codebases. LINK

Alex Karp on how we have been taught the fallacious issues at school. Actually stunned by this man, and completely beloved his e-book. LINK | THE TECHNOLOGICAL REPUBLIC

Cradle — An open-source platform creates a collaborative house for safety groups to share and analyze menace intelligence information. LINK

Kierkegard on goal. LINK

Train, Don’t Inform — Steve losh says technical documentation ought to assist customers construct psychological fashions as an alternative of simply giving them options to repeat and paste. LINK | HN DISCUSSION

AI Escape Room — Pangea Safety created an escape room problem that checks your skill to jailbreak an AI with immediate injection strategies. Sensible advertising and marketing, too. LINK

The OpenSecrets profile of Tulsi Gabbard. LINK

David Brooks (one in every of my favourite columnists) talks about how Trump is ruining America’s fame, not simply his personal. LINK

Web Velocity Take a look at Extension — This Chrome extension allows you to test your web velocity straight out of your browser tab, with “customary” and “business” modes for reasonable vs. optimized outcomes. LINK