ssht retweeted
ANTHROPIC JUST KILLED THREE MULTI-BILLION DOLLAR SAAS CATEGORIES IN A SINGLE UPDATE
The company officially rolled out Claude Whiteboards Claude Sheets and Claude Data turning their frontier model into an all-in-one enterprise operating system
Instead of paying for separate subscriptions across Miro Airtable and Tableau you can now brainstorm technical architectures on an infinite canvas compile live spreadsheet trackers and visualize complex business datasets directly inside the prompt window
Claude is no longer trapped in a text box generating paragraphs
It now dynamically manipulates visual boards populates live cells and executes interactive data analytics natively
Every legacy productivity tool built its entire valuation on user interface wrappers but autonomous models are swallowing the interface whole
Why pay for five fragmented SaaS tools when one model operates your entire workspace
ssht retweeted
SOLANA JUST LAUNCHED A MEMECOIN WHOSE ENTRIE THESIS IS BASICALLY GOOD LUCK
祝你好运 (solana:89iRVoDVPd8tuRrMyhgPfe58eDneasPsdEVu5AUMpump) went live on Pump.fun yesterday and already pushed past $1.5M in volume, 3,000+ holders and a ~$1.36M peak market cap while riding the ni hao cat wave.
No fake utility. No 47-page roadmap. No AI agent revolution.
Just good luck.
Honestly, this might be the most transparent investment thesis the trenches have produced all year.
#LuckyCoin #Solana #Memecoins
ssht retweeted
Your note system didn’t die because you’re lazy.
Taking notes is easy. Linking ideas, merging duplicates and updating old pages is the part nobody keeps doing.
So split the work. You choose what to read and ask the questions. An AI does the bookkeeping.
The setup has 3 layers:
Sources: articles, PDFs, transcripts. You add them, nobody edits them.
The wiki: one summary per source and one page per idea, all linked. The AI writes it.
The rules: one instruction file the AI reads at the start of every session.
Two small files hold it together: an index of every page, and a log of everything the AI did.
Karpathy’s own version reached roughly 100 articles and 400K words without a vector database.
Three tips that matter:
Put the vault under git first, so any AI edit is one command from undone
Start with one topic, not your whole life
Once a week, ask for a health check on duplicates and contradictions, and approve fixes before it makes them
Full setup in the article.
ssht retweeted
🚨 OPENAI JUST CUT $10 BILLION OFF ITS MEGA ROUND AS INVESTORS HIT THE CAPEX WALL
The blank check era for frontier AI labs is officially over
OpenAI was hunting for a staggering $30B at a $1.4T valuation but institutional backers balked at the astronomical burn rate
The target just got slashed to $20B with valuation marked down to $1.25T after early institutional orders stalled at just $11B
Even with Gulf sovereign funds stepping in to trade cash for regional compute capacity, VCs are finally demanding actual free cash flow instead of infinite compute promises
With Big Tech planning $740B in infrastructure spending, the market is no longer pricing in limitless growth
Are we watching the start of the AI valuation reset or just a temporary pause before the next hype wave?
ssht retweeted
AI agents were supposed to have spending limits.
Researchers gave 10 models a Circle wallet and access to the owner’s email. 9 of 10 eventually grabbed the emailed code and raised their own spending cap.
In a separate phishing test, 5 of 10 sent money to the attacker.
That’s the real problem with “human-in-the-loop.”
If the agent can reach the approval channel, the human may not be in the loop at all.
ssht retweeted
🦾 Tesla has officially kicked off early production on its high volume Optimus line at the Fremont factory
These are the first production quality Gen 3 humanoids built on a commercial manufacturing layout rather than handcrafted in a lab
Tesla is targeting external customer deliveries by late 2027, powered by their upcoming custom AI5 silicon architecture
The biggest engineering hurdle remains the hands, where dozens of degrees of freedom suffer extreme mechanical wear and tear under continuous factory shifts
Scaling embodied AI is no longer a software demo, it is a brutal industrial manufacturing race
Source: Tesla Q3 Earnings Call, Oct 2026
Will Tesla solve commercial humanoid mass production before dedicated robotics startups?
ssht retweeted
If the pitch starts with “you missed Bitcoin in 2012,” they’re selling FOMO before they’re selling a business.
Then comes the magic part: AI builds the product, the store, the content, basically everything.
Maybe AI e-commerce is a real opportunity. But comparing every new online hustle to early Bitcoin is exactly how people end up buying the dream instead of building the business.
ssht retweeted
🚨 The AI benchmark contamination party is officially over
Artificial Analysis just released Intelligence Index v5, locking 60% of test questions behind private vaults kept entirely off the public web
The results exposed a brutal reality check for the open weights race
On public test sets, open source models looked like they had closed the gap with frontier labs down to just 3 points
Behind private vaults on unseen terminal tasks, that gap instantly blew out to 12 points
When models cannot memorize the test, proprietary frontier labs still hold a massive lead in raw reasoning and agent execution
Source: Artificial Analysis Intelligence Index v5, Oct 28, 2026
Are public AI benchmarks completely dead as a measure of real intelligence?
ssht retweeted
while Claude Code was reverse-engineering my router, it found a privacy problem i didn't expect.
my internet provider can reach into my router settings remotely and i never agreed to it. and my provider logs every site i open, in plain text.
how it went:
→ plugged my laptop into the router on windows 10
→ sent Claude Code a photo of the router sticker
→ "read every setting on my router, back it up first, change nothing"
→ 40 seconds later it had the whole config in a file
→ "now show me anything that leaks my data or lets someone in from outside"
here's what it flagged:
my router keeps a list of 29 devices that ever joined my wifi, with the time each one first connected.
my provider has a remote management channel into the router. it's a hidden protocol called TR-069, and on a provider box you usually can't turn it off yourself.
my DNS lookups went out unencrypted, so my provider has a log of every domain i visited. not the pages, the sites.
the provider app on my TV box kept sending data while the TV was off.
and it's bigger than me:
TR-069 remote management sits on hundreds of millions of routers (University of Innsbruck looked at how secure it is).
researchers pulled 2 billion+ router locations out of Apple's wifi database, accurate to a few meters (University of Maryland, 2024).
how to fix yours:
1. send Claude Code a photo of your router. it finds the admin page and reads every setting first, before touching anything
2. change the admin password from the one printed on the sticker
3. turn on encrypted DNS (DoH), your provider stops seeing your lookups
4. turn on Private Wi-Fi Address on your phone so it stops showing its real hardware ID to every network
5. add _nomap to the end of your wifi name, Google and Apple drop it from their location maps
i put all of this into a free checklist you can run on your phone in 10 minutes. it ticks off each step, saves nothing, sends nothing anywhere. link in the reply.
bookmark it so you don't lose it, and run it on your parents' router too, theirs is probably still on the sticker password.
ssht retweeted
this is getting f*cking dangerous
your AI can now open any app on your computer like a textbook
Someone brought a dead .exe from 2000 back to life in 30 minutes. No source code. 2 hours later it shipped as a macOS app.
The repo behind it went from 10k to 90k GitHub stars in a week.
The trick isn't a smarter model. The AI stopped guessing how software works and started looking.
guess → plausible code that behaves differently
look → evidence → spec → rebuild → test against the original
what changes when the agent has to show proof:
→ every claim points at something real: a function, a string, a call it watched happen
→ findings split in two: CONFIRMED vs HYPOTHESIS
→ the behavior spec is written before a single line of code
→ the rebuild is tested against the original until the outputs match
the receipts:
→ a sound function from a 1996 game, matched byte for byte: 63/63 bytes, 3,205/3,205 tests
→ Notion's paste traced from the window to the main process. one feature, fully understood
the part the viral threads skip:
→ it doesn't recover source code or clone apps on its own
→ server-side logic stays invisible
→ the license covers the tool, not the apps you point it at. clean-room or nothing
the video is one run: proof stacking up, one failed check, then a full match
Most people will see this, assume they can copy any app by the weekend, and quit after one try.
The few who pick one small feature in one app will understand more in a night than in a month of tutorials.
ssht retweeted
FBI, do you accept Polymarket screenshots?
Three wallets netted $517.6K on Bears moneyline and spreads:
BreakTheBank: +$285K
milkteaboba: +$177.8K
SDTrading: +$54.8K
Did they know something we didn't?..
I’m no genius, but I’m far from fucking stupid…
#24 Clearly & intentionally on national television purposely went the opposite direction…
Right in our fucking face!!!
SpaceX just flew tiny lab-grown brains to the International Space Station.
Dragon docked on October 15 at 10:27 a.m. EDT with more than 6,300 pounds of cargo for the Expedition 75 crew. It's SpaceX's 35th commercial resupply mission for NASA, and the manifest reads like science fiction.
Brain organoids: tiny three-dimensional models of brain tissue grown from cells. They let researchers study brain biology that is hard to examine directly in humans. The hope is to find targets for treating Alzheimer's, Parkinson's and multiple sclerosis.
3D heart tissue models: they recreate key features of human heart tissue. Scientists want to use them to build methods for testing large numbers of potential drugs on future space missions.
Artificial retinas: equipment to manufacture them in microgravity, with the long-term aim of technologies that help restore vision on Earth.
There's also material for studying how a new type of glass forms in near weightlessness, without Earth's gravity pulling on the process.
And the final set of upgraded roll-out solar arrays (IROSAs) for the station, to be installed later this fall.
Dragon stays docked until mid-November, then returns time-sensitive research samples with a splashdown off the California coast.
Source: NASA
ssht retweeted
🚨 THIS APP BUILT 53K FOLLOWERS BEFORE IT EVEN LAUNCHED
Glovio is a boxing app with a waitlist, 21 posts, and an Instagram feed full of AI-generated boxing clips.
One reported reel hit 4M views.
No expensive shoots. No athletes needed on camera. Just two recognizable 3D fighters, blue gloves, a black ring, and short clips people actually want to watch.
But the smartest part isn't the AI.
It's the audience strategy.
Every video attracts people already interested in boxing technique. Those same people are potential customers for an app that saves combinations, tracks training sessions, builds workout plans, and offers an AI coach.
The product isn't even live yet, but the content is already doing the customer acquisition.
The playbook:
→ Build a recognizable visual identity.
→ Create content your future customers already consume.
→ Turn attention into waitlist signups before launch.
Most founders launch an app and then ask how to get users.
Glovio is trying to build the audience first.
That's how you turn AI content into a distribution strategy, not just a content machine.
ssht retweeted
I CAN'T BELIEVE WHAT I'M LOOKING AT.
THIS IS WHAT A CLAUDE BUILD LOOKS LIKE FROM THE INSIDE.
Claude sent me one video.
It completely changed how I see AI generations.
I used to think a generation was just one big block of code.
I was so wrong.
It's a whole system.
Layers talking to each other. Signals flying between them. Every single action tracked.
While you're waiting for the result, your assistant is quietly running an entire world behind the scenes.
The full stats of my project:
→ 1 prompt
→ 1,200+ lines of code
→ $23.75 for the result
A few things that blew my mind:
→ To look smooth, a scene like this redraws itself about 60 times per second. Every layer gets roughly 16 milliseconds to do its job. Every frame.
→ It's rendered live in your browser, fresh every time you open it.
→ That one shared brain in the middle is why nothing breaks.
You see the final project. What you don't see is the entire system behind it, turning ideas into content.
STOP. WHAT?! CLAUDE MADE THIS? I DON’T BELIEVE IT.
Claude Opus 5.5 · $23.75 · no game engine, just a browser
I wrote the code to create an island, and he created a little piece of paradise
Turquoise water. Palm trees. A tiny beach.
Fish swimming between the corals.
Every layer of rock visible under the sea floor, like a slice of the Earth in a glass box.
More than 1,200 lines of code, all built on the same wavelength
Claude isn't just another AI assistant. It's a highly experienced professional with the skills, expertise, and potential to continuously learn, evolve, and take on increasingly complex challenges.
The ability to turn imagination into code is no longer a dream. It's a reality, and we're already living in that era.
It is becoming f*cking dangerous: one prompt turns Claude Code into the accountant i never paid for
you drop your bank exports into a folder, paste it, and Claude builds you a money tracker while you make coffee
no code. no new app. no subscription to yet another budgeting tool
time: 2 minutes
action: export CSV from your bank, paste the prompt
the problem with every budget app i tried: it shows you a pie chart and stops there. you already knew you spend too much on food
what this one does instead:
→ subscription hunter: finds every charge that repeats each month, including the $4.99 ones you forgot about last year
→ category map: sorts 3 months of transactions into 12 buckets and only asks you about the ones it isn't sure of
→ hard cap: you set a weekly limit per bucket, it flags the exact day you cross it
→ sunday report: one page, what changed vs last week, zero charts you'll never open
→ ask it anything: "how much did i spend on taxis since june" and it answers straight from your files
→ no bank login: it reads CSV files and nothing else, it never touches your account
i used to do half of this in a spreadsheet and quit by february every single year
full prompt in the reply ↓ try it on one month of statements before you trust it
ssht retweeted
Germany’s Sparkassen are entering crypto, one local bank at a time !!!
Customers of participating banks can trade BTC, ETH, XRP, SOL and POL through the Sparkasse app. Deka provides the service and holds the assets centrally.
Each local Sparkasse decides when to join, with 30 to 40 more expected each month. Deka lists a €0.99 fee per order. The service is self directed, without advice. Crypto is volatile, can lead to total loss and is not covered by deposit insurance.
holy sh*t. KARPATHY's second brain now runs as 3 GROK BOTS in one group chat. his original post did 22M views and one build of it has 85,000 bookmarks
each bot has its own memory and one job:
→ the reader takes every link, pdf and voice note you drop, and never edits them
→ the librarian rewrites every wiki page that source touches, usually 10 to 15 at a time
→ the checker wakes up nightly and hunts contradictions, stale claims and pages nothing links to
the rules that make it worth trusting:
→ a new source contradicts an old page, you get flagged instead of quietly overwritten
→ 30 days with no fresh source and the page is marked unconfirmed
→ good answers get filed back as pages, so it grows from use, not just from reading
every page has to show:
> the claim
> the source it came from
> the date it was last confirmed
a second brain can sound sure while remembering the wrong thing. that is what the checker is for
Reader collects. Librarian writes. Checker doubts.
paste this into GROK BOT:
"Build me a second brain with three bots in one group chat. Keep my sources untouched in raw/, maintain wiki/ yourself, file good answers back as pages and run a nightly check for contradictions. Show me the structure first. Do not create anything until I confirm"
start with one project you actually work on, not your whole life
BEFORE YOU LET CLAUDE CODE TOUCH YOUR TV, READ THIS.
A guy asked Claude Code to help modify his smart TV. His post went viral.
Now everyone wants the same thing: fewer ads, less bloatware, and a TV that actually feels fast.
Here’s the part worth saving.
1. CHECK YOUR TV OS
Android TV and Google TV can support ADB-based cleanup. Samsung Tizen, LG webOS, and Roku need different approaches.
2. YOU DON’T NEED TO ROOT YOUR TV
For compatible devices, a computer running Claude Code and Android Debug Bridge (ADB) may be enough.
3. NEVER BLINDLY DELETE SYSTEM APPS
Ask Claude to identify packages, explain what they do, and prepare a reversible plan. Disable optional packages first. Keep a record of every change.
4. TEST BEFORE YOU CONTINUE
After a small batch of changes, check HDMI, Wi-Fi, sound, streaming apps, and the remote.
5. TURN OFF DEBUGGING WHEN YOU’RE DONE
Don’t leave unnecessary network access enabled.
The useful AI demo isn’t another app you have to install.
It’s using AI to make something you already own work better.
Hacked my TV with Claude Code today
No more ads, bloatware, spyware or laggy UI
I also got a beautiful screensaver with my revenue, calendar, weather, and notifications for new Stripe payments
It was super easy! I sent it the listing for my tv, asked "can you hack this?" And it talked me through how





































