I don't unpack in a hotel or an Airbnb anymore until Claude Code finishes one sweep.
Bags stay by the door. The laptop opens first.
One prompt, 5 steps, read only the whole way:
Wi-Fi: maps every device on the network and flags the camera brands and the ports cameras use to stream video
Bluetooth: listens for AirTag, Tile and SmartTag beacons that keep showing up after you leave the room
Your phone: runs Amnesty's open source spyware toolkit against known stalkerware indicators
Lenses: walks you through a flashlight sweep of everything that faces the bed and the shower
Report: one ranked table with the next move for each item, from asking for another room to calling the police
It never logs into anything, never touches a device that isn't yours, and tells you straight what it can't see.
I'd rather give it 10 minutes than spend the whole trip wondering.
The full sweep is below, copy paste ↓
Aryaa retweeted
🔴 Bir güvenlik araştırmacısı, Tor ve tornet kullanarak IP adresini her 3 saniyede bir değiştirme yöntemini paylaştı.
WHOEVER LEAKED THIS TRADING SETUP HAS ZERO FEAR
SOMEONE JUST BUILT A 24/7 MARKET-MAKING DESK WITH GROK BOT.
Not one bot.
300 GROK BOTS working together.
-> One watches price action.
-> Another tracks liquidity.
-> Another hunts arbitrage.
-> Another manages inventory.
-> Another monitors risk.
Grok Bot sits at the center, coordinating the swarm and turning all that activity into one trading system.
Market moves → the swarm reacts → risk gets checked → strategies adjust → everything feeds back into the system.
You close your laptop.
The 300 bot trading desk keeps running.
This is what happens when you stop using Grok Bot as a chatbot and start treating it like an entire trading team.
FULL GUIDE ON HOW TO SET IT UP BELOW
my #1 tip for using grok bot: give it an agent passport.
an inbox, a calendar, a funded card with hard limits, a wallet, a signing key, a phone number, a domain, an x account. treat it like you're onboarding a real employee, not configuring a chatbot.
one important correction before you start: don't give it a gmail account. google is actively disabling gmail accounts created by grok bot. use agentmail instead, it's an official plugin already listed in grok bot's browser, oauth, three minutes, 24 email tools ready.
for payments, stripe's mcp server is real and live, letting an agent create customers, issue invoices, and process transactions through natural language, no custom integration required.
then write the actual rules before any of this goes live. what can it read. what can it send. what can it spend. what needs your approval. who can shut it down.
give it the passport. give it the rules. in that order.
Aryaa retweeted
🚨 THIS IS ABSOLUTELY INSANE
a developer just released a free tool that builds a vertical video with a talking presenter.
out of one script and one photo.
it's called lanshu-create-ai-presenter-video.
it landed on GitHub two days ago. 267 stars. MIT.
here's how it works:
1. You give it a script and a presenter photo you have the rights to.
2. The AI writes and voices the narration, generates the presenter, and locks the lips to the audio.
3. The skill edits, burns in subtitles and a cover, checks the sync, and hands back a master with a QA report.
9:16, 1080x1920, 30fps, 45-75 seconds.
all of it happens in one run, on your own machine.
save this before you sit down to cut your next video.
i left the link in the next comment 👇
Aryaa retweeted
THIS FREE AI VIDEO TOOL IS ACTUALLY INSANE.
An open-source AI video generator just crossed nearly 8K GitHub stars.
And the best part?
It runs locally on your own PC and can reportedly work with as little as 6GB of VRAM, even on older GPUs.
It brings multiple AI video and image models into one interface without:
Cloud uploads
Monthly subscriptions
Watermarks
Getting started is simple:
Clone the repo → run the installer → open your browser → start generating.
No expensive AI subscription.
No sending your files to the cloud.
Just a local AI video generation setup running on your own hardware.
This could easily become one of the best free AI video tools of the year.
Save this before everyone discovers it.
Aryaa retweeted
Video editing tools you should ACTUALLY know about 👇🏻
• @runwayml : AI video generation & motion tracking
• @ElevenLabs : Top tier AI voice synthesis & dubbing
• @descript : Text-based video editing & transcript cleanup
• @TopazLabs : Video AI upscaling & frame rate interpolation
• @Frame_io : Seamless client reviews & timestamped feedback
• @envato : Unlimited assets, LUTs, & motion graphics
• @ArtlistIO : Royalty free music & SFX for commercial projects
• @OpusClip : AI powered shorts generator from long content
• @submagicpro : Captions, B roll, & auto zoom generator
• @midjourney : Concept art & custom thumbnail assets
• @Loom : Quick video messaging for client walkthroughs
• @NotionHQ : Project management & client workflow organization
• @audio2edit : Quick audio cleanup & noise reduction
• @epidemicsound : Curated sound design & background audio
• @freepik : Vectors, PSDs, and rapid design templates
• @motionvfx : Professional plugins & cinematic templates
• @BorisFX : VFX, keying, & planar tracking tools
• @pika_labs : Generative AI video effects & animation
Aryaa retweeted
Grok Bot makes clipping YouTube podcasts ridiculously easy.
I just send it a link and say:
“Cut from 16:43 to 18:17.”
A few minutes later, the exact HD clip is ready in the chat.
My Clipper bot has already sent it to my Researcher and Writer bots, which pull the best quotes and prepare the caption.
I didn’t code any of this.
I literally just logged into my Google account on Grok Bot and asked it to clip a YouTube video.
I’ve wanted this for years: tell an AI which part of a podcast you want, and it automatically cuts the HD clip, transcribes it, and starts preparing the post.
Grok Bot makes workflows like this incredibly simple for a non-coder.
Aryaa retweeted
RAG vs. CAG, clearly explained!
RAG is great, but it has a major problem:
every query hits the vector DB. even for static information that hasn't changed in months.
this is expensive, slow, and unnecessary.
Cache-Augmented Generation (CAG) fixes this by letting the model keep static information in its key-value (KV) memory, which is what the model builds internally for every token it reads.
in fact, you can combine RAG and CAG for the best of both worlds.
here's how it works:
RAG + CAG splits your knowledge into two layers.
↳ static data (policies, documentation) gets cached once in the model's KV memory
↳ dynamic data (recent updates, live documents) gets fetched via retrieval
you get faster inference, lower costs, and less repeated work.
the trick is being selective about what you cache.
only cache static, high-value knowledge that rarely changes. cache everything and you'll hit context limits. separating "cold" (cacheable) and "hot" (retrievable) data keeps this system reliable.
you can start today. OpenAI and Anthropic already support prompt caching in their APIs.
one thing to know before you scale it.
prompt caching matches on an exact prefix, byte for byte. your cached layer only gets reused when it sits at the very front of the context in the same order every time.
↳ reorder two cached policy documents and both turn into a miss
↳ cache document A alone and document B alone, then query both, and the second one misses because the model computed its cached state without ever seeing the first
in production this looks like a small fraction of your cached blocks serving almost all the hits. the rest just sits there.
the way out comes from how attention behaves. tokens attend mostly to their own local neighborhood, and only a few reach across document boundaries. CacheBlend recomputes those few and reuses everything else from the separately cached documents.
multi-document queries run two to four times faster, quality holds, and order stops mattering.
it ships in LMCache, which is fully open source.
repo: github.com/LMCache/LMCache
(don't forget to star 🌟)
below, i have quoted my article on KV cache management. it covers where prefix caching stops working and how a proper caching layer fixes it.
give it a read.
cheers! :)
Aryaa retweeted
Joe Rogan uploads 2 to 3 hour episodes. You see 30 second clips. Someone cut them.
Creators pay $29 a month for AI to do that job.
A solo developer in China gave it away for free.
The repo is called AutoClip. 6,391 stars on GitHub. 1,253 forks. MIT licensed.
Every clip you save from a podcast, stream, or interview was cut from something longer. Streamers do not go viral from streams. They go viral from clips of streams. A 2 hour episode holds 15 to 25 viral clips. Finding them by hand takes 4 to 6 hours.
Here is what AutoClip does.
Paste a YouTube link. Or a Bilibili link. Or drop in a local video file. The AI reads the whole transcript, scores every moment for viral potential, cuts the highlights, adds captions, and hands you a folder of vertical shorts ready for TikTok, Reels, and YouTube Shorts.
Here is what makes it different from OpusClip.
- Runs locally on your laptop. Your footage never leaves your machine.
- Works with OpenAI, Google Gemini, Alibaba Qwen, or SiliconFlow. Pick whichever LLM is cheapest for you.
- No credit system. No monthly minutes cap. Process a 4 hour livestream if you want.
- MIT licensed. Fork it. Ship it in a product. Sell it. No permission needed.
- Docker one-line install. Runs in a browser tab at localhost:3000.
The stack: FastAPI backend, React frontend, Celery task queue, Redis, SQLite, yt-dlp for downloads, FFmpeg for cutting. One developer named zhouxiaoka wrote 89 commits and shipped the whole thing.
Here is the interesting part.
OpusClip raised $68 million in venture funding from SoftBank, Samsung Next, and DCM Ventures to build a subscription business around this exact workflow. The whole short form video AI market is priced like a SaaS product because most creators do not know an open source alternative exists.
One person in Hangzhou wrote it in his spare time and put it on GitHub. Six thousand people have starred it. One thousand two hundred have forked it.
One honest note. Last commit was 77 days ago on June 3. The maintainer is one person. Some features on the roadmap are still marked in development. If you want a polished product with a support team, pay OpusClip. If you want to own the tool that clips your videos, download this tonight.
The people who own the software own the workflow.
Your videos. Your AI. Your shorts.
(Link in the comments)
Aryaa retweeted
Introducing Headlong, an open source microharness for persistent agents: self-guided agents that think continuously.
Most agent harnesses are reactive: you send a task, the agent completes it, and then it sits frozen until the next request. Cron jobs and heartbeats wake it up to run a checklist and put it back to sleep.
A Headlong agent is never asleep. It keeps generating thoughts about whatever it decides is interesting, in a self-guided loop inspired by human inner monologue. Your message doesn't start a session. It's one more observation that lands in the agent's thought stream, and the agent decides if and when to reply.
Headlong is built on the idea of persistent agency: continuous inner thought generation between external interactions. The agent sets its own interests and priorities, comes up with its own projects, and sometimes pings you unprompted with progress.
To keep our prototype as simple and small as possible, we implemented Headlong as a microharness: a complete agent harness in under 10K lines of Bash, organized as a handful of small executables. It includes a loop that generates the next thought, shellm (a recursive language model written in Bash), a trajectory stored as a DAG of jsonl files, and context as a projection of that trajectory.
We've been running one Headlong agent internally at Laude for several weeks. The whole team talks to it over Slack and Telegram, and every conversation lands in its single stream of thought. It works in its own fork of Headlong and we've pulled over 50 of its commits into main.
One night, with nobody talking to it, it went back to check whether a recall process it had built was actually wired into its mind, found that it wasn't, diagnosed and fixed the bug, and verified the fix end to end. 48 minutes, no human asked for the fix or was in the loop at any point. Every step is a timestamped line in its log.
Things broke too, and we wrote those up. Background thinking costs us $1 to $2 an hour, our agent stopped its own service three times by accident, and self-delegation died on day one. Details in the post.
One line installs everything and starts an agent. Use a dedicated sandbox and spend-capped API key; it runs real shell commands and thinks around the clock.
Headlong is research software, be careful!
curl -fsSL headlong.ai/install.sh | bash
Launch post: laude.org/updates/headlong-a…
Repo: github.com/laude-institute/h…
Headlong is a @LaudeInstitute / MIT collaboration.
en lugar de ver Netflix toda esta noche, dedica 54 minutos a este vídeo
este tipo te enseña a ganar dinero con youtube shorts en 2026 sin experiencia previa
encuentra tu nicho → graba shorts → sube todos los días → monetiza
🔖 guárdalo, te servirá.
Aryaa retweeted
This Claude trading bot made +$168,236.
Claude Code + TradingView is insane.
In 5 minutes, I'll teach you how to build your own
THIS GUY MAKES COMPLEX AI AGENT CONCEPTS RIDICULOUSLY EASY TO UNDERSTAND.
No jargon wall. No assuming you already know what a reward model or an evaluation harness is.
Just a straight line from "I have no idea how agents actually work" to "oh, that's genuinely simple."
The best explainers do not simplify by cutting corners. They simplify by finding the one analogy that makes the whole thing click.


































