Shams retweeted
DEEPSEEK BUILT A FULL 3D SATELLITE TRACKER FOR $1.52
One prompt. One try. Browser app called SATELI.
– Tracks 16,846 active satellites in real time
– Includes 11,157 Starlink sats
– Live 3D Earth with city lights and day/night
– Predicts when ISS and others are visible from your location
– Sky trajectory charts and night vision mode
DeepSeek coded SGP4 from scratch and passed 666 official checks across 33 sats, matching reference math within 8 micrometers.
Built with DeepSeek V4.1 Flash + AgenC Core.
If you just describe what you want, AI may already ship it.
WAIT… So Sam wants to remove PCs from humans so data centers can have more resources to build better AI for humans?
Make it make sense.
How exactly are people supposed to access and use that AI without personal computers, RAM, and local hardware?
You can't eliminate the tools people use to interact with technology and expect them to magically benefit from the technology you're building.
Am I missing something here, or does this logic make absolutely no sense?
Shams retweeted
Hey BetFury share a 1M Battles and $400K Tournaments.
Use promo code F007 , and Let’s go!
Agents, $2 000 000 + #Birthday007 has started😎
Your next assignments ➤ betfury.tv/Birthday_007
✧ $1M Birthday Battles & $300K Agent Missions
✧ $400K Grand Tournament & $200K Secret Bonus Roulette, and more.
Mission accepted? Tap ❤️
SOMEONE JUST DROPPED 192+ WEBSITE DESIGNS FOR CLAUDE.
This could save designers and developers hours of work.
The collection includes:
- 192+ human-designed website templates across different industries
- 32 free interactive website sections
- Real design references Claude can use for better visual direction
- Inspiration for building cleaner, more professional websites
- Different layouts, styles and visual ideas to explore
- A way to give Claude better design inspiration instead of starting from scratch
Instead of relying on Claude to guess what good design looks like, you can give it actual references to work with.
You can use these designs to guide Claude toward the layout, style and visual direction you want.
No more getting the same boring AI-generated layouts every time.
Bookmark it. You might need it for your next project.
Explore the collection 👇
supply.madebyoversight.com
Free website sections 👇
supply.madebyoversight.com/s…
Note: These are design references to guide Claude, not a one-click website builder.
Follow me for more AI tools, insane projects and useful resources
THIS IS ACTUALLY INSANE 💀
ONE SOLO DEVELOPER JUST BUILT A COZY FISHING GAME USING AI.
No hand-modeling.
No traditional 3D workflow.
The setup:
- Claude Opus 5.5
- Blender MCP for 3D models and animations
- Unity 6 for the game environment
A forest river, waterfalls, rapids and a fisherman in a straw hat.
Imagine building an entire game without manually creating the 3D assets.
Would you play this? 👀
THIS COULD BE ONE OF THE CRAZIEST AI LEAKS EVER 💀
AN ALLEGED MYTHOS-CLASS MODEL WITH 14 TRILLION PARAMETERS JUST APPEARED ON HUGGING FACE.
BUT THERE'S ONE PROBLEM...
NOBODY HAS VERIFIED THE WEIGHTS.
HERE'S WHAT THE REPO CLAIMS:
- 14 TRILLION PARAMETERS
- 1,224 MoE EXPERTS
- 7.72 TB OF WEIGHTS
- 772 WEIGHT SHARDS
- HYBRID GDN + MLA ARCHITECTURE
- MODEL CONFIGS, ARCHITECTURE CODE AND TRAINING METADATA
IF THESE CLAIMS ARE REAL, THIS COULD BE MASSIVE.
BUT A REPO CLAIMING TO CONTAIN INTERNAL WEIGHTS DOESN'T PROVE THEY'RE AUTHENTIC.
ANYONE LOOKED INTO THE ACTUAL FILES YET?
LEGIT LEAK OR ELABORATE FAKE? 👀
huggingface.co/burner22/numb…
Readers added context they thought people might want to know
The Hugging Face repo burner22/numbat makes unverified claims of an Anthropic "numbat" Mythos-class leak with ~14T params/7.72TB weights, but contains fabricated weights (inconsistent 6.8T metadata; distributions match synthetic data, not trained models like DeepSeek).
huggingface.co/burner22/numbat
x.com/EricParker/sta…
en.wikipedia.org/wiki/Claude_My…
Shams retweeted
BREAKING: CLAUDE CODE FLAGGED THOUSANDS OF UNEXPECTED FRONT-CAMERA FILES ON AN ANDROID PHONE.
The discovery was shocking.
While checking a full phone backup, Claude Code reportedly found over 14,000 image files that weren’t visible in the Gallery or Recently Deleted folder.
The files were allegedly stored in system directories, including:
/data/system/users/0/faceunlock/cache
And another folder associated with Google Play Services.
According to the report, the files included:
→ Front-camera frames
→ Infrared images used for Face Unlock
→ Timestamps and location information
→ Files dating back approximately 15 months
The owner also claimed to find files created while they were actively inspecting the folder.
That’s a serious privacy concern if verified.
Android manufacturers generally describe Face Unlock as processing facial data to create biometric representations rather than keeping ordinary camera photos.
However, finding files in a system cache doesn’t automatically prove that the phone was secretly taking pictures, that every file was accessible to other apps, or that the data was being sent anywhere.
Still, it’s worth reviewing your device’s privacy settings.
A few things to check:
→ Review which apps have camera permissions
→ Check your device’s privacy dashboard
→ Review system diagnostics and usage settings
→ Keep your Android software updated
→ Investigate unexpected files before deleting system data
The reported findings haven’t been independently verified, so don’t assume every Android phone behaves this way.
But it’s a reminder to pay attention to what your device is storing and which apps have access to its camera.
Would you ever think to check your phone for something like this?
CLAUDE CODE FLAGGED A HIDDEN IPHONE PHOTO CACHE WITH THOUSANDS OF FILES.
While backing up my iPhone to my MacBook, Claude Code spotted a directory I wasn’t even looking for:
PhotoData/Thumbnails
It reportedly contained over 145,000 photos, videos, and screenshots dating back several years.
Some appeared to be thumbnails or cached copies of files I had deleted, including items I thought were long gone.
That got me thinking about what we leave behind on our devices.
Old screenshots.
Personal documents.
Photos we no longer want.
Sensitive information we once saved temporarily.
But here’s the important part: finding files in a thumbnail cache doesn’t prove that every permanently deleted photo remains there forever, that the files are full resolution, or that any app can freely access them. iOS manages caches in different ways, and access depends on the app’s permissions and system restrictions.
Still, reviewing your privacy settings is worth the time.
I checked the apps on my phone and found 23 that I wanted to review. After going through them, I reduced the number I trusted with file access to just 3.
It took about two minutes.
You can do the same:
→ Review which apps have access to your files and folders
→ Remove permissions from apps that don’t need them
→ Uninstall apps you no longer use
→ Keep your iPhone updated
The takeaway isn’t that every deleted photo is secretly available forever. It’s that you should know what data remains on your device and which apps you trust.
Save this checklist and share it with other iPhone users.
I thought this was a game someone spent weeks building.
Turns out DeepSeek V4.1 Flash generated the entire thing from one prompt
A miniature island with:
- Moving boats
- 12 houses with chimney smoke
- A full day/night cycle
- Rain and snow that actually change the environment
- Fireflies, stars and a moon
- A working lighthouse beam sweeping across the sea
- A harbor, beach, cliffs and a pier
- Interactive camera controls
- Procedural sounds of waves, wind and a distant harbor bell
3,100 lines of code.
15 JavaScript files.
Zero external assets.
About $1 in API costs.
The nighttime visuals are actually crazy.
The houses light up at sunset, the boats sail around the island, and the snow slowly covers the rooftops.
And the developer didn't manually edit the generated code.
Imagine what AI could build with a much bigger game idea and a more detailed prompt.
Would you actually play a game built entirely by AI?
Repo
(github.com/tetsuo-ai/agenc-c…)
Here's the prompt 👇
Build "Tiny Harbor", a living miniature island village in three.js, shown like a tilt-shift diorama. It should feel cozy and full of small details people want to rewatch.
TECH
- Plain static site, no build step: index.html, src/*.js, README.md. Serve with: python3 -m http.server 8080
- three.js 0.170.0 through an import map: "three" = cdn.jsdelivr.net/npm/three@0… and "three/addons/" = cdn.jsdelivr.net/npm/three@0…. No other dependencies. No external models or images: build every object from procedural low-poly geometry with flat-shaded, soft pastel materials.
THE ISLAND
- A round island on a calm sea: cliffs on one side, a sandy beach, grass, a small hill.
- A harbor with a wooden pier and three small boats that sail loops around the island, slow down and dock, then leave again.
- A lighthouse on the cliff, a windmill on the hill with turning blades, and about 12 houses with chimneys and lit windows.
- Trees and grass tufts that sway in the wind (InstancedMesh). Smoke rising from chimneys. Seagulls circling. Foam lines where waves meet the beach. A fish jumps now and then.
DAY AND NIGHT
- A full day cycle in 60 seconds by default: dawn mist, bright noon, golden sunset, blue night with stars and a moon.
- Sky gradient and light color follow the sun. Windows and street lamps turn on one by one at dusk.
- At night the lighthouse sweeps a soft beam across the sea, and fireflies appear over the grass.
WEATHER
- Buttons for clear, rain and snow. Rain brings darker clouds, falling drops, ripples on the sea and wet, shiny streets. Snow slowly covers the roofs in white.
LOOK
- Tilt-shift depth of field (BokehPass) so it looks like a real miniature. Soft shadows from the sun and moon, hemisphere fill light, gentle bloom on lights at night, ACESFilmic tone mapping, sRGB output.
- A simple stylized water shader with small waves and soft reflections.
- Target 60 fps on a recent laptop. Cap devicePixelRatio at 2.
INTERACTION
- Drag to orbit, scroll or pinch to zoom, and a slow auto-orbit when idle.
- Click a boat, house, windmill or the lighthouse to glide the camera close to it, and click empty sea to return.
- A small UI: time of day slider, speed (0.5x to 4x), weather buttons, and a sound toggle (procedural WebAudio waves, wind and a distant bell, off by default).
- Works on desktop and phone.
VIDEO MODE (?record=1)
- Fixed 1080x1350 canvas (4:5), UI hidden.
- A scripted 40-second film: dawn wide shot with mist, boats leaving the harbor, close pass by the windmill at noon, sunset as the windows light up, night with the lighthouse beam and fireflies, rain starting, then a slow pull back to the whole island.
- Deterministic, so two recordings match.
CHECKS (run them, then report)
- Open it in a headless browser and confirm there are no console errors.
- Save screenshots to screenshots/ at dawn, noon, sunset, night and in the rain.
- Measure and report fps on this machine.
- README: how to run, the controls and the video mode.
Keep the design original: do not copy any existing game, film or artwork.
NO WAY THIS COST JUST $1.52
SOMEONE JUST GAVE DEEPSEEK ONE PROMPT AND IT BUILT A FULL 3D SATELLITE TRACKING SYSTEM.
Not a basic demo. An actual browser-based satellite tracker called SATELI.
→ Tracks 16,846 active satellites in real time
→ 11,157 Starlink satellites included
→ Live 3D Earth with city lights and day/night simulation
→ Predicts when the ISS and other satellites will be visible from your location
→ Includes sky trajectory charts and night vision mode
BUT THE CRAZIEST PART IS THE ENGINEERING.
DeepSeek implemented SGP4 from scratch and passed 666 official verification tests across 33 satellites.
The result matched reference calculations to within 8 micrometers.
And the entire session cost just $1.52.
ONE PROMPT. ONE TRY. $1.52.
Built using DeepSeek V4.1 Flash with AgenC Core.
AI coding is getting out of hand.
ARE WE ENTERING AN ERA WHERE YOU JUST DESCRIBE WHAT YOU WANT AND AI BUILDS IT? 👀
SOMEONE JUST BUILT THIS ENTIRE GAME WITH AI. 🤯
Someone just spent 13 weeks building an ocean exploration and fishing game using AI on a $200/month plan.
29 ships in bottles.
120 fish planned.
5 zones.
Fishing, exploration, boat upgrades, pirates, storms and fog.
And the crazy part? Almost the entire game was built with Claude Opus.
The total cost so far is around $900–950.
AI game development is getting INSANE.
Are we entering an era where one person can build an entire game studio alone? 👀
I'm nitpicking but it seems every AI company has forgotten what "MAX" actually means.
If I'm paying for MAX, I expect MAX. Not "MAX until you actually need it."
We’re renaming our Max plans:
Max 5x → Max
Max 20x → Ultra
Max is still $100/month, and Ultra is still $200. If you’re a subscriber, there’s nothing you need to do.
We’re also removing the 5-hour limit for verified Ultra accounts, starting today with most existing subscribers. Newer accounts keep a temporary limit until they’re verified, which helps us prevent fraud and abuse. You'll get an email when it's removed.
Weekly limits remain unchanged. View our plans here: claude.com/pricing
THIS IS F**KING INSANE.
Someone's Android app had been broken and abandoned for 4 YEARS.
They asked GrokBot for help.
10 minutes later, they had a working APK.
12 HOURS later, they'd rebuilt the app and added around 50 new features.
No spending hours in Android Studio. No hiring an expensive freelancer.
Just describing what they wanted and letting AI do the heavy lifting.
People are still sleeping on how much AI coding is changing software development.
We're entering a whole new era.
DeepSeek V4 Flash just built a WALKABLE A320 cabin in 51 minutes for $0.80.
Not a static 3D model. An actual airplane cabin you can explore.
You can walk from the cockpit through business class, economy, the galley, and even the lavatory.
WASD movement.
Collision detection. Mouse look. Free-fly mode. Day/night lighting. Openable overhead bins and window shades.
But the crazy part isn't just what it built.
I let the agent run without approvals.
51 minutes.
166 model calls.
98% of input came from prompt cache.
$0.80 total.
And it tested its own build in a real browser before calling it done.
No console errors. ~132 FPS at 1440x900.
Built with AgenC Core, an open-source agent harness using DeepSeek's own API.
GitHub: github.com/tetsuo-ai/agenc-c…
I'm sharing more AI experiments, coding agent builds, and real-world results like this.
Follow me for more AI builds, experiments, and projects worth paying attention to. 👀