Unsupervised Studying NO. 484 – Model Slux

Hey, hope you’re doing effectively! Loopy week already.

• I believe this is perhaps considered one of my favourite newsletters ever. I’m simply actually feeling the brand new hyperlink type and breakout type, and story choice appears higher? Anybody agree?

• Ordered some noise cancelling earbuds for sleeping. Will let you understand how they do. They’ll be paired with my eye masks that I already use.

• Wrote a new essay on why I’m scared for the primary time for Apple, and what I considered right now’s WWDC keynote. READ IT (3 minute learn)

• I’m going to do a profile on all of the CCP members and construction. I’m certain there are a bunch on the market already, however I’m out of the blue excited by how the federal government truly works. Like, who’re these individuals? And what are their politics?

• One other essay on why it doesn’t make any sense to say, “It’s simply subsequent token prediction.” READ IT (5 minute learn)

• I’m additionally going to do a brand new future pattern/funding evaluation train like I did years in the past with my buddy Ty Sbano, the place we picked shares based mostly on potential traits. This time I’ll use AI although (clearly). Will most likely flip the evaluation and my feedback and proposals from it right into a PDF that I promote for a couple of bucks or one thing.

Trump Overhauls Biden’s Cybersecurity Insurance policies With New Govt Order
Trump considerably reversed Obama and Biden cybersecurity insurance policies and centered on new issues. Right here’s a abstract of what it added and faraway from current insurance policies:

• Defeating international threats

• Safe software program practices

• Border gateway safety

• Publish-quantum cryptography readiness

• Fashionable encryption protocols

• AI for vulnerabilities

• IoT safety requirements

• Limiting sanctions’ scope

Bellingcat Exams Whether or not AI Can Really Geolocate Photographs Now
Bellingcat examined 20 AI fashions on 500 geolocation duties and solely OpenAI’s newest ChatGPT fashions beat Google Lens at figuring out photograph places. THE ARTICLE | BELLINGCAT’S 2023 AI GEOLOCATION STUDY

Bishop Fox 2025 Purple Group Instruments Checklist
Bishop Fox put collectively their favourite crimson group instruments for 2025, masking C2 frameworks and Energetic Listing exploitation stuff. THEIR LIST

OpenAI Printed Their Annual Report on How Unhealthy Actors Are Utilizing AI Maliciously
I really like that they put these stories out. My important takeaways have been:

• 4 out of 10 main abuse instances seemingly originated from China, from social engineering to cyber threats

• They’re seeing misleading employment schemes, activity scams from Cambodia, and remark spamming from Philippines

• Covert affect operations doubtlessly linked to Russia and Iran are utilizing AI as power multipliers

Anthropic Launches Customized AI Fashions For Nationwide Safety
Anthropic constructed particular “Claude Gov” fashions particularly for U.S. protection and intelligence companies that work higher with labeled materials and refuse to assist much less usually. THE ARTICLE | ANTHROPIC’S ANNOUNCEMENT 

Britain Will Ship 100,000 Drones To Ukraine By 2026
Britain simply dedicated to a tenfold enhance in drone deliveries to Ukraine, sending 100,000 by April 2026. THE ARTICLE

 AI Lastly Finds One thing Educated Scientists Have Missed for A long time
This must be one of the vital unimaginable and slept-on items of AI information I’ve ever heard.

• Quite a lot of skilled scientists have been making an attempt to determine how a selected form of bacteriophages (viruses that infect micro organism) turn into cellular and do what they do

• This has been a thriller for a really very long time, and this group of scientists are the world consultants on the subject

• They gave a bunch of knowledge to a brand new Google mannequin educated to do novel analysis and create novel speculation

• It created a novel speculation that the human researchers had missed, which they are saying was due to their very own bias

• Upon testing it, it turned out the AI was right

In the event you care about these things you must take heed to your entire episode. It’s extraordinary. Mainly they’d been making an attempt to determine how this explicit virus was in a position to do what it did. They knew X, they knew Y, they knew Z. And so they assumed that as a result of XYZ that _____ couldn’t have been potential (take heed to get the complete detailed clarification).

However what they didn’t understand is that they have been making an assumption. A defective one! They have been blinded by bias (their phrases), which didn’t enable them to see the answer. And because of this they’re so blown away by what the AI did. As soon as they noticed the speculation they weren’t blown away. It was apparent. What they have been blown away by was the truth that it discovered the answer after they had not, due to their bias. Consider: they’re the world f-king consultants on this!

The implications listed below are unbelievable. Consider what number of analysis papers are on the market. Consider how a lot knowledge is mendacity round ready to be explored. These are all dots ready to be related. And there aren’t practically sufficient researchers to try this work.

That is how we get new, actual profit from AI. Particularly in well being, the place we desperately want to take a look at how molecules work together with cells and such. So unbelievably hyped about this! Go test it out. THE PODCAST | THE COGNITIVE REVOLUTION PODCAST

Apple Releases Controversial Paper on AI
Apple launched a paper that individuals are decoding as saying AI’s can’t motive. I believe the paper is form of lacking the purpose, and likewise individuals are misinterpreting what they really mentioned. They by no means claimed no AI can motive.

What they mentioned is that with sure math issues you may confuse the AI should you change the parameters, which feels extra like memorization than deep understanding. I believe it’s a good level, and a good weak spot of the AI they examined. However, acknowledging I’m biased right here, I believe we’ll look again on this as a “variety of r’s in strawberry” second.

In the meantime, Stanford simply discovered that AI alone scored 90% efficacy towards human medical doctors’ 75% in one other story this week. The sort of stuff is simply clickbait for AI doomers and skeptics. Now we have to verify we’re watching what AI is doing in the true world. On actual issues. THE PAPER | HN DISCUSSION

OpenAI Massively Dropped o3 Costs
They decreased the price of o3 by 80%, and are releasing o3-pro right now. THREAD

OpenAI Should Preserve All ChatGPT Conversations Indefinitely On account of Authorized Maintain
OpenAI supposedly(?) cannot delete any ChatGPT logs proper now as a result of they’re beneath a courtroom order to protect all the pieces for ongoing litigation. “We’re required to retain all knowledge and can’t course of deletion requests throughout this era” – OpenAI. Tremendous tousled given the truth that they’ve choices the place individuals particularly paid for the alternative. THE ARTICLE | HN DISCUSSION

OpenAI Makes ChatGPT’s Voice Mode Sound WAY Extra Human
ChatGPT’s voice mode now has higher intonation, practical pauses, and even does practical sarcasm. I additionally confirmed it will possibly sing. It’s actually, actually good. THE ARTICLE | OPENAI RELEASE NOTES

Stanford Examine Exhibits Medical doctors Plus AI Beat Conventional Diagnostic Instruments
Medical doctors utilizing AI as a collaborative companion acquired 85% accuracy versus 75% with conventional instruments, however the true story is that AI alone scored 90%. THE PAPER

Microsoft Reshuffles Management to Give attention to AI Brokers
Microsoft is reorganizing its high executives overseeing Workplace 365 and Dynamics to prioritize promoting AI brokers that may automate white-collar work. THE ARTICLE

My New Favourite Description of a Enterprise Moat

Why Bell Labs Really Labored So Properly
Bell Labs succeeded as a result of they gave good individuals full freedom to discover no matter them, then—solely later—related their discoveries to actual enterprise issues. I need to implement this on the nationwide scale. THE ARTICLE | HN DISCUSSION

BYD’s 5-Minute Charging Places China within the Lead for EVs
BYD simply demonstrated 1,000-kilowatt chargers that add 250 miles in 5 minutes—which is mainly fuel station velocity for electrical automobiles. This scares the crap out of me. THE ARTICLE | SHANGHAI DEMO VIDEO

The Chinese language Tech Behind Amazon’s Humanoid Robots
Amazon is testing humanoid robots at their San Francisco workplace for bundle supply, constructed on Chinese language tech. They’re establishing an indoor impediment course to see if these AI-powered bots can deal with real-world deliveries. THE ARTICLE | AMAZON’S HUMANOID ROBOT TESTING

YouTube Loosens Content material Guidelines Utilizing “Public Curiosity” Customary
YouTube now lets movies keep up in the event that they violate group pointers however are deemed within the “public curiosity,” bumping the violation threshold from 25% to 50% of content material. THE ARTICLE | NY TIMES ORIGINAL REPORT

Rents Are Dropping in Most Main U.S. Cities for the First Time Since 2023
28 out of 44 main metropolitan areas noticed year-over-year lease decreases in Could 2025. REDFIN’S RENTAL MARKET REPORT

Caffeine Retains Your Mind Awake Even Whereas You Sleep
New analysis exhibits caffeine would not simply hold you awake—it truly prevents your mind from correctly sleeping even once you suppose you are getting sleeping effectively. THE ARTICLE | HN DISCUSSION

Las Vegas Fights Report Warmth With Huge Tree Planting Initiative
Las Vegas hit 120 levels final yr and warmth killed over 500 individuals, so now they’re planting 60,000 bushes by 2050 to chill the most well liked neighborhoods. THE ARTICLE 

Mushrooms Could Talk Utilizing Up To 50 Phrases
A scientist analyzed electrical alerts between fungi and located patterns that look remarkably much like human language construction. THE ARTICLE | ROYAL SOCIETY STUDY

Forests Offset International Warming Extra Than Scientists Beforehand Thought
A brand new UC Riverside examine exhibits replanting all of the bushes we have misplaced for the reason that 1800s may cool the planet by 0.34 levels—a couple of quarter of present warming. The key sauce is that bushes do not simply suck up carbon, additionally they launch compounds that replicate daylight and make clouds. THE ARTICLE | THE NATURE STUDY

Somebody Constructed An MCP Server That Really Runs On Cloudflare Staff
This boilerplate by Fatih Kadir Akın helps you to deploy MCP servers to Cloudflare Staff with OAuth and PostgreSQL assist in-built. THE PROJECT

Information Visualization Reveals Patterns in D&D Monster Designs
Somebody created a extremely cool knowledge visualization of Dungeons and Dragons monsters as a part of Tidy Tuesday, and the patterns they discovered are fairly good. THE REDDIT POST | FULL ANALYSIS 

How Anthropic’s Groups Use Claude Code
Insanely good content material right here, and I really like the truth that it’s their precise inner software they usually’re exhibiting how they use it. THREAD | FULL TUTORIAL PDF

We Are No Longer a Critical Nation
Paul Krugman argues that markets are beginning to deal with America like an unreliable rising market fairly than a protected haven. A key level he raises: US rates of interest and greenback are actually transferring in reverse instructions, one thing sometimes seen solely in rising market drama. THE ARTICLE 

Nice Clarification of How Mannequin Context Protocol is Totally different from Conventional APIs
Largest distinction is APIs are for builders, and MCPs are for AI’s (brokers). THE COMPARISON

APIs Turn out to be the Basis for AI-Prepared Companies
Most corporations aren’t prepared for AI not as a result of they lack fashions, however as a result of their techniques cannot speak to one another by APIs. THE ARTICLE

Bigfoot Veo3 Movies VIDEO

Mysterious Object Fires Indicators at Earth Each 44 Minutes
Astronomers discovered this bizarre house object that blasts radio waves and X-rays at us for 2 minutes straight, then goes quiet for 44 minutes, they usually don’t have any clue what it’s. What I don’t get is the way it retains pointing at us given how briskly it’s transferring and how briskly we’re transferring. THE ARTICLE | NATURE RESEARCH PAPER 

Mapping Latitude and Longitude to Nation, State, or Metropolis
Austin Henley breaks down the surprisingly complicated problem of reverse geocoding—turning coordinates into location names. I do know a couple of individuals into these physical-related puzzles and it looks as if an excellent mixture of bodily exercise and mental stuff. THE ARTICLE | HN DISCUSSION

Calculus in 30 Seconds from a e book in 1910 SNIP

Software program Is About Guarantees
Bram Adams argues that software program success comes from making clear, testable guarantees to customers—what precisely you may ship given your constraints and assets. THE ARTICLE 

This Is How They Inform Me Bug Bounty Ends
My buddy Joseph Thacker thinks AI brokers will finally discover all vulnerabilities routinely, however that there’s nonetheless numerous room for creativity. THE ARTICLE | JOSEPH’S BLOG 

New OSINT Instruments Listing Organizes 100+ Scattered Sources
The creator of R00M 101 (tremendous cool too) constructed a filterable listing of 100+ OSINT instruments as a result of they have been uninterested in searching by GitHub repos and random Discord servers to seek out what they wanted. THE PROJECT | HN DISCUSSION | ROOM 101 OSINT TOOL