Kanika
@KanikaBK • 37,339 subscribers
I test AI tools and agents daily. Building practical workflows for creators and builders who want real results. DM/[email protected]
Shorts
Videos
🚨 7 HOURS AGO ANDREJ KARPATHY just dropped something UNCOMFORTABLE TO WATCH. He gave a score to every job in America on AI replacement risk. Software developers: 8-9 Paralegals: 8-9 Medical transcriptionists: 10 Think about YOUR job for 3 seconds. Here's what you must know👇
Kanika518,736 views • 7 months ago
Grok Bot just got a LOT more interesting. Someone figured out how to run Claude Code, Codex, and Cursor inside Grok Bot with a surprisingly simple setup. The setup is basically: 1. Copy the `setup-grok-prompt` 2. Give it to your agent 3. Done. Four minutes. That changes the conversation completely. Because Grok Bot isn’t just sitting there answering questions anymore. You can start thinking about it as a layer that orchestrates other AI coding tools and agents while you stay focused on the outcome. And this is exactly the direction I’ve been exploring. I wrote a deep dive on how I built a Second Brain using Grok Bot that works 24×7, 365 days. Not another “here’s what Grok can do” article. It’s about turning Grok Bot into something that can remember, organize, execute, and keep working in the background. If you’re curious about what happens when you stop treating AI like a chatbot and start treating it like a digital worker… My article is a must-read.
Kanika87,178 views • 1 month ago
🚨 JUST IN: MICROSOFT just open sourced a VOICE AI THAT TRANSCRIBES 60 MINUTES OF AUDIO in a single pass. 100% FREE. It knows who spoke. It knows when they spoke. It knows exactly what they said. All in one shot. No chunking. No context loss. It's called VibeVoice. Not a transcription tool. Not a basic speech to text wrapper. A frontier voice AI family with ASR, TTS, and real time streaming. All open source. All free. Here's what it actually does 👇 VibeVoice ASR - Speech Recognition: → Processes 60 minutes of continuous audio in a single pass → Never slices audio into chunks so global context is never lost → Identifies WHO spoke, WHEN they spoke and WHAT they said simultaneously → Supports customized hotwords for domain specific accuracy → Works in 50+ languages natively → Already adopted by Hugging Face Transformers library → Already being built on by the open source community BY PEOPLE WHO HAD NO IDEA THIS LEVEL OF ACCURACY WAS ALREADY FREE. VibeVoice TTS - Text to Speech: → Generates up to 90 minutes of speech in a single pass → Supports up to 4 distinct speakers in one conversation → Natural turn taking and speaker consistency throughout → Expressive speech that captures emotional nuances → Supports English, Chinese and multiple other languages VibeVoice Realtime - Streaming TTS: → Only 300 millisecond first audible latency → Streams text input in real time → 0.5B parameters so it actually deploys anywhere → Robust long form generation up to 10 minutes → Lightweight enough for production use today The core innovation nobody is talking about: Most voice AI models slice long audio into short chunks. Every time they slice, they lose context. Speaker tracking breaks. Semantic coherence breaks. Accuracy drops. VibeVoice uses continuous speech tokenizers running at an ultra low frame rate of 7.5 Hz. This preserves audio fidelity while dramatically boosting computational efficiency. The entire 60 minutes stays in context. Nothing gets lost. Nobody gets misidentified. The numbers: → VibeVoice ASR 7B - available now on Hugging Face → VibeVoice Realtime 0.5B - try it on Colab right now → 50+ supported languages → 11 distinct English voice styles → 9 multilingual speaker voices → Already integrated into Hugging Face Transformers → Finetuning code now available The wildest part? A voice powered input method called Vibing just built itself on top of VibeVoice ASR. Available on macOS and Windows right now. The open source community is already shipping products on top of this. 100% Open Source. Free to use. Free to fine tune. Free to build on. 🔖 Save this before your competitors find it first. 👇
Kanika222,078 views • 6 months ago
Everyone is building an Obsidian second brain. mikenevermiss found a German dev who built something that actually runs his business. Instead of organizing notes, he wired up 7 AI workflows that do real work every day. They qualify leads. Chase invoices. Research competitors. The second brain era is over. Say hello to the second employee.
Kanika39,352 views • 1 month ago
🚨Sam Altman just said: "There will be a middle layer that becomes really important. A whole new set of startups that tune existing large models. And those companies will create a lot of enduring value." Not foundation models or generic apps but PayBuddy is built for the middle. Most builders are either too high or too low in the stack to capture it. My article covers the 25 AI concepts you need to understand to find your place in it. Full guide below.
Kanika52,636 views • 2 months ago
🚨 THIS 100% FREE GITHUB REPO JUST MADE EVERY AI NOTE-TAKING APP POINTLESS. Drop any source in article, PDF, YouTube transcript, or Journal entry. Claude reads it, links it to everything in your vault, and files it into a connected knowledge graph on your own machine. Most tools save your notes. This one connects them. Every source gets linked to everything you already know. Karpathy described the idea - 15 skills. Four vault methods. Works with Claude Code, Codex, and Gemini. Most people have years of knowledge buried in folders they open twice a year. This turns all of it into a second brain that works while you sleep. MY ARTICLE COVERS THE EXACT METHOD I BUILT TO DO THE SAME THING. Guide below!
Kanika38,795 views • 2 months ago
🚨 This guy just leaked the GITHUB's 1 command that removes Claude's watermark and it is spreading everywhere right now. Most people using Claude have zero idea the text carries invisible signals that identify it as AI-generated even after copying and pasting across multiple apps. Hidden Unicode characters. Statistical patterns in sentence structure. Markers that trained systems can detect with scary accuracy. He found the exact command that strips it all. This is blowing up fast because everyone thought these watermarks were permanent. My article covers 25 hidden Claude features most people completely miss and this watermark removal technique is the one going viral right now. Full guide below.
Kanika29,967 views • 2 months ago
This guy took notes in Obsidian every single day for 1 full year. 365 days. No breaks or skipped weeks. He started with 0 notes and 0 connections. Just an empty vault and a habit he was not sure would stick. His second brain remembered everything. This is exactly the problem I was trying to solve when I built my Web Clipper templates. Most people clip articles and books into Obsidian and never touch them again. I spent 4 months testing 6 templates on everything from dense research papers to 500 page books until I found what actually turns captured information into connected thinking. Full breakdown in my article👇
Kanika48,976 views • 4 months ago
0:29
Sensitive content
This media may contain sensitive content.
Hong Kong University just released a FREE AI TOOL that turns a one-line idea into a full short film. No editing required. 100% OPEN SOURCE. It's called ViMax You type a prompt. Or paste a novel. It does the rest. Director. Screenwriter. Producer. Video generator. One system. One input.
Kanika38,986 views • 5 months ago
ANOTHER FREE OPEN SOURCE GROK BOT ALTERNATIVE JUST DROPPED AND THIS ONE HAS VOICE MODE, MOBILE APPS, AND BOTS THAT DELEGATE TO OTHER BOTS. It is called Rakazo and it is the most complete open source AI teammate platform I have seen. Persistent bots with memory, conversation history, and routines. Full access to browser, terminal, files, and desktop. Voice mode and a mobile app. Bots can delegate to peers or auto-create short-lived subagents. Bring your own model through any provider. Run the whole stack locally. One Docker command to get started. The part that makes this relevant to building a night shift is the persistence layer. These are not one-shot agents that run a task and disappear. They are teammates that stay on, remember what happened last time, run on a schedule you set, and keep working whether your computer is on or not when deployed on a server. Six bots running overnight on a schedule is not a pipe dream with Rakazo. It is a Docker command and an afternoon of setup. My article covers how to build a 6-bot night shift in Grok Bot that runs while you sleep and Rakazo is the open source infrastructure that makes that possible without a paid subscription. Save this. Both links in comments.
Kanika12,256 views • 1 month ago
This guy opened Obsidian 90 days ago with 200 isolated notes rotting in folders he had not opened in 6 months. He just started linking. One note at a time. One connection at a time. Every single day. Today he has 847 notes and over 2,300 connections between them. The graph does not look like folders. It looks like a brain. And it shows him what he actually thinks about, not what he assumed he cared about. This is exactly why I built my Web Clipper workflow around connection and not storage. If you want to see how I turn heavy books and dense data into actual insights inside Obsidian, I wrote about my 6 best Web Clipper templates that I use daily. Link in the post. He is not smarter than he was 90 days ago. He just stopped filing and started linking. His notes are not stored. They are alive. If you have been using Obsidian as a glorified Google Drive folder, like and bookmark this. 90 days is all it takes.
Kanika33,755 views • 4 months ago
🚨 CHINA IS ON FIRE! Tencent open-sourced TencentDB Agent Memory. Long-term memory for AI agents that runs 100% locally. ↳ No Pinecone or cloud APIs ↳ You don't need to repeat yourself every session ↳ 61% fewer tokens ↳ PersonaMem accuracy: 48% → 76% ↳ Runs on plain SQLite ↳ FREE. Open source. 5.1k stars in 48 hours. Most people paying $70/month for vector databases will miss this.
Kanika23,775 views • 3 months ago
Boris Cherny, Head of Claude Code at Anthropic: "I do not prompt Claude anymore. I have loops running that prompt Claude. My job is to write loops." The person who built the most used AI coding tool stopped prompting entirely. Most people using it every day are still typing one request at a time. This Github repo changes that. ↳ Seven production loop patterns with real token cost estimates ↳ CLI that scaffolds your entire loop in five minutes ↳ Starters for Grok, Claude Code, Codex, and Opencode The repo gives you the framework. My article covers the mental shift that makes it actually work.
Kanika20,693 views • 2 months ago