APPLE WANTS $129 FOR A TRACKPAD. THE PHONE IN YOUR DRAWER CAN BE ONE FOR $0.
it has a touchscreen, a gyroscope and Bluetooth. since Android 9 it can pair with your computer as a normal Bluetooth mouse. no root, no server, no app on the computer
I turned my old Android into a trackpad with Claude Code. here’s the $0 version:
→ Claude Code writes a small Android app
→ it uses BluetoothHidDevice, the API Google added in Android 9
→ phone flat on the desk: the screen becomes a trackpad
→ phone in your hand: the gyroscope moves the cursor where you point
→ a small dead zone keeps the cursor still while your hand rests
→ tap = click, hold = drag
→ Mac or Windows pairs with it like any Bluetooth mouse
paste this into Claude Code:
“Build an Android app (min SDK 28) that registers as a Bluetooth HID mouse with BluetoothHidDevice. Flat on a table it works as a trackpad, held up it maps the rotation vector sensor to the pointer with a small dead zone. Tap = left click, long press = drag. Use Google’s wearmouse sample as a reference. Show me the plan first.”
the catch: some phone makers block this Bluetooth mode, so test yours before you sell the trackpad
$129 trackpad or $0 drawer phone. save this ↓
13
19
356
Claude just f*cked the Magic Mouse... again
This is f*cking crazy.
My Android phone (Solana Seeker) is now a Wii remote and a trackpad for Mac and Windows.
> 2 apps: desktop (Mac menu bar / Windows tray) + Android
> 0 servers
> Bluetooth only, nothing goes online
> Cursor updates 240 times a second
> Built-in aim trainer
Point the phone at the screen and the cursor lands where you aim. Tap to click, hold to drag.
Lay it flat on the desk and it's a trackpad.
One aim trainer mode is tennis against AI versions of Elon Musk, Sam Altman, Jensen Huang and a few other CEOs (stylized bots, not the real guys). Train your aim, roast imaginary billionaires.
AirPods case, Apple Watch, iPhone, now Android.
Coffee maker or toaster next?
3
32
YOUR AI HAS AMNESIA AND A CONFIDENT VOICE. THAT’S THE WORST COMBO.
Stop trusting Claude’s memory. Give it a second brain that shows its sources.
Most people treat Claude Code like a chat window. The ones who get it treat a folder like an atom: plain-text files orbiting one map. Claude loads the map first, then answers only from what’s written. It remembers what you decided and why, and it refuses to invent a reason you never wrote down.
No apps. No plugins. No context rot.
What circles the map:
→ notes/: one fact per file, named after the content
→ people/: one card per person (role, responsibilities, update cadence)
→ projects/: current status + next hard deadline
→ root: long-term memory, lessons learned, and a numbered decision log that always includes the reason
→ CLAUDE.md: the single source of truth that sits in the middle and is read before anything else
Every answer comes back in exactly one of three ways:
→ Cited: it names the file it used
→ Not written down: it says so instead of guessing
→ New: it creates the file and updates the map in the same reply
facts · people · projects · decisions · map
The prompt to paste into Claude Code is in the reply. Save this and set it up tonight ↓
11
19
516
Make this folder my second brain:
1.Create notes/, people/, and projects/.
2. Add root files for memory.md, lessons.md, and decisions.md (numbered log; every entry must include the reason).
3. Write CLAUDE.md as the map that explains where each type of information lives. Pin memory and decisions at the top.
4. One topic per file. Filenames are lowercase with hyphens, named after the content.
5. Answer only from the file the map points to and name the file under the answer. If nothing is written down, say so. Never invent.
Show me the full layout first.
1
41
YOUR CLAUDE.MD IS MAKING CLAUDE DUMBER.
Anthropic’s own docs say it: keep it under 200 lines because longer files eat context and Claude follows your rules worse
Claude reads the entire CLAUDE.md at the start of every session, before you type a word
the fix is to turn CLAUDE.md into a short map and move everything else into files Claude opens only when it needs them:
→ CLAUDE.md: what lives where, plus your rules. that’s all
→ notes/: one topic per file, named after what’s inside
→ people/: one file per person
→ projects/: status and where you left off
→ decisions.md and lessons.md: what you never want to explain twice
two traps people fall into:
→ @-importing everything doesn’t help. imported files load together with CLAUDE.md
→ a CLAUDE.md inside a subfolder loads only when Claude works in that folder, so folder rules go there
and the rule that changes every answer: if it’s not in a file, Claude says “not written down” instead of guessing
paste this into Claude Code in an empty folder:
“Set up my notes here. Keep CLAUDE.md short: a map of what lives where and my rules. Create notes, people and projects folders plus decisions.md and lessons.md. One topic per file, lowercase with hyphens. Answer only from the file the map points to and name it. If it isn’t written down, say so. Show me the plan first.”
save this before your CLAUDE.md hits line 201 ↓
20
25
682
DON’T BUILD YOUR STARTUP ON THE MODEL EVERYONE IS HYPING THIS WEEK.
TwIL-8B just dropped. 8.79B parameters, 4.98 GB, runs on a laptop, 93% of the best frontier model on LiveCodeBench. 577K people saw the launch post. the catch sits in the model card
Read this before you ship a single line on it:
→ the license is webAI Non-Commercial. test it all you want, but a paying customer can’t touch it
→ it burns tokens. on AIME it uses about 19.5k per problem. GPT-5.5 uses about 5k
→ the 93% coding score comes from the “verbose” variant, not the balanced one
→ every score is webAI’s own run, against competitor numbers from different test setups
→ hard science is still far off: GPQA Diamond 70.2 vs 93.6
→ tool use, agent tasks and SWE-bench weren’t tested at all
→ no extra safety tuning beyond the base model, the card says so itself
imagine building 3 months of product on it and finding the license on launch day
How to use it the smart way:
→ prototype locally for $0 in API calls
→ run it on 10 of your real tasks before you trust any benchmark
→ need a commercial 8B: it’s built on IBM’s Granite 4.2 8B, which is Apache 2.0
small models are closing the gap fast. the license decides what you can sell
model card in the reply ↓
Modest intelligence is arriving faster than most people realize.
Today, we're open-sourcing TwIL-8B.
Just 8.8 billion parameters, yet approaching frontier-model scores on several difficult benchmarks:
• 93% of frontier performance on coding
• 92% on instruction following
• 90% on contest mathematics
All from a model small enough to run on your own hardware.
But the benchmarks aren't the most important part of this announcement.
For years, the AI industry has focused on building the most capable model, regardless of the cost required to run it.
I believe that equation is changing.
The next era of AI will be defined by the cost of completing work. Not simply the intelligence of the model, but how efficiently that intelligence can accomplish the task.
If a model with 8 billion parameters can reliably complete the same work as one with hundreds of billions, why pay for the larger model?
As smaller models become more capable, intelligence becomes something we can distribute, specialize, and run across the hardware we already own.
That's the future we're building at @thewebAI
TwIL-8B is available today. Open weights.
And this is just the first of several announcements we're making today.
HF Link: huggingface.co/webAI-Officia…
BREAKING: Binance says the $6.34B hack started with a fake job interview and a poisoned AI coding tool
their official report dropped yesterday. the attack took 5 months to set up:
→ June: an engineer searches for a popular AI coding tool and installs a tampered Homebrew package. security swaps his laptop, but his role on the wallet team is now exposed
→ September: another engineer gets a job offer on a real job platform. fake HR video call, fake CEO, then an “assignment” repo to run on his work laptop. the malware inside was a 0-day and nothing flagged it
→ Oct 20: from that laptop the attacker goes through an internal AI tool, reaches two wallet engineers and finds a deployment token
→ Oct 29, 17:15 to 19:56 UTC: a container gets deployed and a script hits the wallet signing service
→ 20:05: XRP starts leaving. then SOL, OP, 18,270 BTC, and ETH through an EIP-7702 delegation
→ 23:56: the last transfer
about $6.34B was hit and $2.85B is frozen or intercepted. Binance says it covers the loss itself, and 98% of the tokens were replenished by Oct 31
the FBI attributes it to North Korea’s TraderTraitor group. Binance says it stopped 57 North Korea-linked attempts this year before this one got through
if you have a work laptop: install dev tools only from the official site, and never run an interview “assignment” on the machine you work on
full report linked in the reply ↓
STOP PAYING A FRONTIER MODEL TO READ YOUR EMAIL.
Superlinked’s tests: a small open model does the same job in 13 ms instead of 1.48 seconds, at 0.7% of the cost
most of what an agent does all day is sorting, extracting and searching. over and over
what they measured against GPT-6.1 Sol at list prices:
→ classify text: 90% of the quality at 0.7% of the cost
→ detect objects: 104% of the quality at 0.4% of the cost
→ redact personal data: 95% at 2.9% of the cost
→ OCR: 91% at 3.8% of the cost
→ PDF to Markdown: 87% at 9.4% of the cost
→ extract names and dates: 80% at 0.06% of the cost
→ chat: 99% at 27% of the cost
→ median across 14 tasks: 98% of the quality at 3% of the cost
the speed gap hurts more than the price. an agent pays it on every call, so a 10-call task pays it 10 times
they just launched SIE Serverless, a hosted version of their open-source engine (Apache 2.0, 3.5k GitHub stars). new accounts get $20 in credits, and teams can apply for $500 more
the move: send the repetitive calls to small models and keep the frontier model for the work that needs real reasoning
these are the vendor’s own benchmarks, so run a few tasks on your own data before you switch anything
links to the benchmark and the engine in the reply ↓
🚨 Huge news 🚨 today we are launching SIE Serverless - a managed inference product focused on small models. It’s useful for spiky workloads on 20+ task specific workloads across Search, OCR and Agentic execution.
Sign up for a $500 credits launch grant 👇
It’s powered by our open source inference engine that has received thousands of github stars in the last few months, battle-tested across finance, legal & tech. We believe that self-hosting is going to be critical given all the trust & security issues in AI, but practically there are always workloads where you’d like somebody else to deal with the GPU quota like synthetic data generation, evals, public research and beyond.
The team at Superlinked has been burning the midnight oil for months and I’m so proud of every single one of you - what a privilege to call you my team mates y’all!
We built it, now help us make some noise 🙌
A like, a comment or a repost would mean a lot! ❤️
Stop letting Claude invent your life.
Most people keep feeding it the same context over and over and still get hallucinations.
This is the opposite: one empty folder, four plain-text files, and a rule against guessing.
No apps. No plugins. No memory tools. Just folders and one map that loads first.
It works like this:
→ notes: one file per fact or topic
→ people: one card per person (who they are, what you agreed, last contact)
→ projects: exact status and where you left off
→ root: long-term memory, lessons, and every decision you’ve made
→ CLAUDE.md: the map Claude reads before anything else. It tells it exactly where every piece of information lives
Claude is only allowed three answers:
→ It opens the right file and cites it
→ It says “not found” instead of making it up
→ It creates a new file and updates the map
Paste this into Claude Code inside an empty folder:
Build me a second brain here:
1.Create three folders: notes, people, projects.
2. Create root files for long-term memory, lessons learned and decisions.
3. Write CLAUDE.md as the map: where each kind of info lives, with memory and decisions pinned so they are always in focus.
4. Rules: one topic per file, name each file after what’s inside, lowercase with hyphens.
5. When I ask something, open only the file the map points to and cite it. If it isn’t there, say not found. When I give you something new, create a file and update the map.
Show me the structure before you create anything.
Set it up once. Use it every day. You stop repeating yourself.
22
26
568
Apple and Google are holding your photos hostage for $9.99 a month.
That’s nearly $1,200 over 10 years.
Stop paying — and new photos stop backing up.
Your memories are now on a subscription.
Two 4TB drives cost about $380 once. that’s 3.2 years of rent, and then the monthly bill is gone.
One prompt in Claude Code does the boring parts:
→ checks your disk and tells you if your library fits
→ installs Immich, an open-source Google Photos on your own computer, plus Tailscale so your phone reaches it privately
→ pulls your iCloud originals with icloudpd. you type the password yourself, never in the chat
→ imports your Google library from Takeout with immich-go, so dates and albums survive
→ compares file counts until it can show N of N on your disk
→ turns on auto backup in the Immich phone app
→ copies everything to the second drive
→ stops there. cancel the plan yourself, and only after those counts match
the catch: your computer has to be on for new photos to back up, and every drive dies eventually. that’s why you buy 2
before: $9.99 a month for as long as you keep taking photos
after: about $380 once, on 2 drives you actually own
prices change by country and by month, so run the math on your own plan first
stop renting your own memories. the full prompt is in the reply ↓
full prompt is in the reply ↓
Help me move my photo library off iCloud and Google Photos onto this computer. Work one step at a time and ask before you install anything.
1. Check my free disk space and tell me if my library fits. Ask me for my library size if you can’t see it.
2. Install Immich with Docker, then Tailscale, and show me how to connect the Immich phone app.
3. For iCloud, set up icloudpd to download originals. I will type my password and 2FA code into the terminal myself. Never ask me to paste them in this chat.
4. For Google Photos, walk me through Google Takeout, then import the archives with immich-go.
5. Compare file counts between the source and Immich and show me any missing files.
6. Copy everything to my second drive and verify that copy too. Do not delete anything from any cloud, and do not cancel any subscription.
1
2
94
BREAKING: $3.5 billion left Binance hot wallets on Ethereum in 40 minutes. estimates now go up to $5.8 billion, and CZ says “it does look like a hack”
what’s known so far:
→ the first transfer at 20:24 UTC sent 167,464 ETH and 63 tokens, including 1.615 billion USDT, to a brand-new address
→ about 3 minutes earlier, an address funded through Tornado Cash submitted an EIP-7702 delegation for a Binance hot wallet
→ estimates range from $3.3 billion (Blockaid) to $5.57 billion across Ethereum, Bitcoin and Tron (Decrypt) and up to $5.8 billion (Coinbase’s Conor Grogan)
→ Tether says it froze more than 1.567 billion USDT. Circle froze 823 million USDC
→ Tether also blacklisted 7 wallets further down the chain
→ part of the money had already moved on, and some USDT was swapped into ETH within minutes
Binance says it is investigating a “potential security issue” and that “user funds are completely safe”. it hasn’t called it a hack
CZ no longer runs Binance, but as a shareholder he expects “all user funds will be fully covered”
BNB dropped 5.5 to 7% to around $699. bitcoin slipped under $79,000
the numbers are still moving, so treat every figure here as a first estimate. follow for updates as Binance confirms what happened ↓
🚨BREAKING: Bitcoin just dropped below $79,000 for the first time in the last 40 days.
$450M longs liquidated in just 30 minutes
15
28
1,274
8,700 signals. 28 trades. this bot says that’s how $100 became $4,768
the setup, as the creator describes it: $100, a hard spending cap, no manual trades. it threw away 99.7% of everything it saw
5 agents inside the GROKIUS terminal:
→ Scout tracks 512 smart wallets
→ Filter kills anything that breaks a hard rule
→ Analyst runs the deep checks
→ Sniper copies only what survives
→ Guard sets a stop before every entry and tightens it on pullbacks
the biggest reasons a coin got killed:
→ 2,519 had a dev wallet over 5%
→ 1,931 had liquidity under $25k
→ 1,763 had top 10 holders over 35%
→ 1,259 had bundled buys
→ 923 were already up 100% in the last hour
→ 150 had a smart wallet that sold early
the reported result: 28 trades, 64% win rate, 10 stops hit, max drawdown 19.7%. worst trade SIB at -$313.70, best JEFF at +$1,080.80
the idea worth stealing even if you never touch meme coins: a filter that says no to almost everything, and a stop that’s set before the trade
the part to keep in mind: these are the creator’s numbers, with no wallet or transaction link attached, so treat them as a claim until you see one
not financial advice. bookmark this for the next “my bot did 47x” post ↓
BREAKING NEWS: My iPhone has 319 screenshots that can wipe out my bank account, crypto, and identity. 27 apps can already open all of them.
I plugged my phone into my Mac for a backup and ran Claude Code through 16,400 photos. I expected nothing.
It found 319 screenshots I’d completely forgotten:
My debit card, front and back, full number and CVV. Taken in 2023 “so I wouldn’t forget it.”
The 12-word seed phrase of my crypto wallet. Saved “just for a minute” two years ago.
My passport photo page.
Email 2FA backup codes.
A bank transfer with my full account number.
My mom’s Wi-Fi and router passwords.
I forgot every single one. My phone didn’t.
Then I checked who could see them. Settings → Privacy & Security → Photos. 27 apps had Full Access — including a flashlight, two QR scanners, a photo editor from 2021, and a weather app.
Full Access means exactly that. Every photo, every screenshot, every day those apps stay installed. One compromised app and those 319 screenshots are gone.
The fix takes 3 minutes:
Plug your iPhone into your Mac and run the prompt (in the first reply) through Claude Code. It only reads your local backup and shows the risky screenshots blurred.
Settings → Privacy & Security → Photos → switch every app you don’t fully trust to Limited Access or None.
Delete what it found, then empty Recently Deleted. Otherwise it stays for 30 days.
I went from 27 apps to 2 in three minutes.
Check your parents’ phones too. They screenshot everything.
How many apps on your phone have Full Access to your photos right now?
BREAKING: Claude Code found 12,847 photos of my face on my iPhone that I never took.
It looked like the phone was taking them every time I picked it up. From the club. From home. At a red light while I was driving. Almost every time I took the phone in my hands there was a frame.
They were sitting in a system cache with timestamps and location data. None of them were in Photos.
I plugged the phone into my MacBook and asked Claude Code to scan for any front-camera files that weren’t in my Photos library or Recently Deleted. It flagged Private/var/mobile/Library/Caches/com.apple.faceid. Over 12,000 files. Mix of infrared frames and regular front-camera images. Each one had a timestamp and GPS coordinates attached.
I started scrolling. Most were from the last 14 months. Then I saw one from four seconds ago. I was still looking at the folder.
Apple says the TrueDepth system turns the capture into a mathematical model inside the Secure Enclave and discards the image. On this phone the raw frames were still in the cache. Several of the processes that wrote to the folder belonged to system diagnostics. A couple of old apps with camera permission also had access to the same directory.
I turned off diagnostics sharing and revoked camera access from every app that didn’t need it.
If you want to check your own phone, plug it in, run a backup, and ask Claude Code to look for front-camera files outside the Photos library. The folder path above is where mine were.
Yeah, so this is pretty messed up.
I plugged my iPhone into the MacBook and just asked Claude Code to look for any front-camera stuff that wasn’t already in my Photos or Recently Deleted. It found this folder:
Private/var/mobile/Library/Caches/com.apple.faceid
Over 12k files. Mix of infrared TrueDepth frames and regular front-camera shots. Every single one had a timestamp and location.
I started scrolling, and most of them were from the last year or so. Then I saw one from like 4 seconds ago. I was literally still looking at the folder.
Apple says the TrueDepth camera turns it into some math model inside the Secure Enclave and throws the image away. On my phone, the raw frames were just sitting there. Some of the processes writing to that folder were system diagnostics. A couple of old apps that still had camera permission could also touch it.
I turned diagnostics off and revoked camera access from everything that doesn’t actually need it.
If you wanna check yours, plug it in, make a backup, and tell Claude Code to scan for front-camera files outside the Photos library. That’s where mine were.
BREAKING: Claude Code found 12,847 photos of my face on my iPhone that I never took.
It looked like the phone was taking them every time I picked it up. From the club. From home. At a red light while I was driving. Almost every time I took the phone in my hands there was a frame.
They were sitting in a system cache with timestamps and location data. None of them were in Photos.
I plugged the phone into my MacBook and asked Claude Code to scan for any front-camera files that weren’t in my Photos library or Recently Deleted. It flagged Private/var/mobile/Library/Caches/com.apple.faceid. Over 12,000 files. Mix of infrared frames and regular front-camera images. Each one had a timestamp and GPS coordinates attached.
I started scrolling. Most were from the last 14 months. Then I saw one from four seconds ago. I was still looking at the folder.
Apple says the TrueDepth system turns the capture into a mathematical model inside the Secure Enclave and discards the image. On this phone the raw frames were still in the cache. Several of the processes that wrote to the folder belonged to system diagnostics. A couple of old apps with camera permission also had access to the same directory.
I turned off diagnostics sharing and revoked camera access from every app that didn’t need it.
If you want to check your own phone, plug it in, run a backup, and ask Claude Code to look for front-camera files outside the Photos library. The folder path above is where mine were.
15
35
910
A bot turned $2,061 into $15,986 overnight on an old Android in a drawer. The only number in that post you can check is the referral discount.
The story this week: a dusty phone, a meme coin app, Grok bots that tap buy while you sleep. The same account says it took $100 to $75,796 in 11 days.
What the text actually gives you:
→ A coin 24 minutes old, bought at $159K market cap at 11:20pm, sold at $1.44M at 3:46am for +$13,924.
→ No wallet address and no transaction link anywhere in the text.
→ The trading app sits in the first reply, with “your fees there drop 10% for life”.
→ The full guide “drops soon”, so you follow now and wait.
→ The last line is a question, so the replies fill up and the post travels further.
How to read these in 10 seconds:
→ Look for a wallet or a transaction link. Meme coin trades usually leave an on-chain record, and a real win takes 1 link to show.
→ Open the first reply. If it’s a referral link, you’re reading an ad.
→ Count what you can verify against what you have to take on trust. Here it’s 1 discount against 9 dollar figures.
→ Stop wiring real money into “bots tap buy” setups.
A coin that’s 24 minutes old can lose most of its value faster than you can read this.
Before: “my bots printed while i slept”
After: an affiliate link with a big number on top.
Bookmark this for the next “$100 into $75K” post in your feed ↓
ChatGPT found 98,473 photos and videos I permanently deleted from my Android.
Three years of deleted photos and videos still sitting in a hidden cache. Any app with storage access can read them.
After the iPhone post I plugged my Pixel into the MacBook and told ChatGPT to look for leftover deleted files. It pointed straight at a folder I had never opened.
DCIM/.thumbnails. Almost 100,000 files. Photos, videos, screenshots. Every one of them was something I had deleted. ChatGPT noted that most were missing from the gallery trash and from MediaStore.
Photos from 2022, 2023, 2024 and 2025. The ones I deleted right after taking them. The ones I cleared from Trash. The ones I deleted three separate times to make sure they were gone. All still there.
Passport photo I took for a visa application in 2024. Screenshots of my bank app with full account numbers visible. Private photos I sent someone and deleted right after a breakup. Screenshot of my crypto seed phrase I took temporarily in 2023. Photo of my ID and credit card I took to remember the numbers and deleted seconds later. Screenshots of my ex’s private stories I saved and permanently deleted in 2024.
Most were compressed previews. A few dozen were close to full size. All sorted by date going back to 2021.
When you delete a photo on Android, the gallery takes it out of your camera roll and Trash keeps it for 30 days. A compressed copy stays in the thumbnail cache and MediaStore so the phone can show previews when you search or scroll. The system almost never cleans this folder. Files from 2022 were still sitting there.
They told you it was permanently deleted. It wasn’t.
I checked which apps on my phone had storage access. 27 apps. Weather apps. QR code scanners. PDF readers. A flashlight app I downloaded in 2021 that I had forgotten about. A calculator. Two of them still asked for full file access every time I opened them. Any of them could have been reading every photo I ever deleted.
Went from 27 apps to 5 in about two minutes.
On iPhone the cache was harder to reach. On Android any app with storage permission can open this folder directly.
Full step by step below. Takes 2 minutes. Save it so you don’t lose it and send it to anyone with an Android phone.
Claude Code found 145,763 photos I permanently deleted from my iPhone.
Three years of deleted photos and videos still sitting in a hidden cache. Any app with file access can read them.
I plugged my iPhone into my MacBook to run a backup. Claude Code was doing the security audit from my last post and flagged a directory I wasn't even looking for.
PhotoData/Thumbnails. Over 146,000 files. Photos, videos, screenshots. Every single one was something I deleted.
Photos from 2023, 2024 and 2025. The ones I deleted the second after taking them. The ones I cleared from Recently Deleted. The ones I deleted three separate times because I wanted to be sure they were gone. All still there.
Passport photo I took for a visa application in 2023.
Screenshots of my bank app with full account numbers visible.
Private photos I sent someone and immediately deleted after a breakup.
Screenshot of my crypto seed phrase I took temporarily in 2022.
Photo of my ID and credit card I took to remember the numbers and deleted 5 seconds later.
Screenshots of my ex's private Instagram stories I saved and permanently deleted in 2023.
Some were full resolution. All organized by date going back to 2021.
When you delete a photo on iPhone, iOS removes it from your camera roll and Recently Deleted. It keeps a compressed copy in a Spotlight search index cache so your phone can find things when you search. Apple never clears it. Not after 30 days. Not after a year. Never.
They told you it was permanently deleted. It wasn't.
I checked which apps on my phone had Files and Folders access. 23 apps. Weather apps. QR code scanners. PDF readers. A flashlight app I downloaded in 2022. Any one of them could have been reading every photo I ever deleted this entire time.
Went from 23 apps to 3 in about two minutes.
Full step by step below. Takes 2 minutes. Save it so you don't lose it and send it to anyone with an iPhone.
Readers added context they thought people might want to know
The author is posting AI-generated engagement bait text with various random claims of deleted photos or videos on a daily basis.
1. x.com/coolerwave/sta… - "BREAKING: Claude Code found 12,847 photos of my face on my iPhone
2. x.com/coolerwave/sta… - "Claude Code found 214,856 photos and videos I permanently deleted
This 800-year-old elephant is a clock.
Al-Jazari built it in 1206. A bowl with a tiny hole sinks inside the elephant for half an hour. When it hits bottom, a ball drops into a serpent's mouth, the serpent tips, pulls the bowl back up, and the driver strikes a drum. Then it starts over.
I asked Opus 5.5 to rebuild it in 3D:
→ a cutaway belly, so you see the bowl fill and sink
→ the ball, the serpent and the drum fire in the right order
→ two full half-hour cycles, squeezed into 15 seconds
→ one html file, three.js, no 3D models
Ancient machines, part 5 ↓
This 800-year-old wheel in India is a clock.
The Konark Sun Temple wheel tells time with its shadow.
I asked 3 AI models to rebuild it in 3D.
Gemini 4 Argon → Most beautiful carvings 🔥
Opus 5.5 → Best at telling the time
Kimi K3.1 → Good design, but simple wheel
Gemini 4 Argon has the richest carvings
the prompt:
"Rebuild al-Jazari's elephant clock (1206) in 3D as a single html file with three.js. Cut a window into the elephant's side to show a water tank with a floating bowl that slowly fills and sinks. When it hits bottom: a ball rolls off the tower roof into a serpent's mouth, the serpent tips forward and pulls the bowl back up by a string, a figure raises its arm, the driver strikes a drum, the ball falls into a vase. Loop it. Add render(t) so every frame is deterministic."
3
91