Before we start, quick poll
Would you want to clone yourself if it reduces your work by hours? |
Context: last weekend, I got 200+ DMs asking about AI tools after Sharan’s podcast with me!
I spent 6+ hours answering the same 15 questions repeatedly, stuff I'd already covered in past newsletters.
Light bulb moment: I needed an AI that thinks like me but works 24/7.
Not some generic chatbot, but something that gets my perspective on AI tools, my skepticism about hype, and my focus on ROI for businesses.
So I built one using Gemini Gems in 15 minutes.
I'll show you how to build your own AI clone.
Before that, let’s look at what’s new in the AI world.
1. Apple Crawls to Google for Siri Rescue (After Years of Trash-Talking AI)
Apple is in talks with Google to use Gemini AI as the backbone for its "revamped" Siri launching in 2026. They approached Google to build a custom AI model that would run on Apple's Private Cloud Compute servers.
My take: This is Apple admitting defeat in the most expensive way possible. After years of claiming privacy superiority and mocking cloud-dependent AI, they're banking on Google to save Siri. The irony? They'll pay billions to Google while their own AI team gets sidelined. Meanwhile, Google gets Apple's data AND their money.
Read more
2. Harvard Dropouts Launch ₹21,000 Surveillance Glasses That Record Everything
Two former Harvard students raised $1 million to build "always-on" AI smart glasses that listen to and record every conversation without external indicators. The glasses transcribe everything you hear, then display relevant information in real-time.
The co-founder says they "give you infinite memory" and help you "cheat on everything." Price: ₹21,000 with pre-orders starting this week.
My take: This is Black Mirror disguised as productivity tech. Their logic: the "trust users to get consent" approach is old school. Places like India, where privacy laws are still evolving, this could be a nightmare scenario.
3. Elon Musk Launches "Macrohard" to Clone Microsoft Using Only AI
Musk announced xAI is building "a purely AI software company called Macrohard" designed to simulate Microsoft's entire operation using artificial intelligence.
He claims since software companies don't manufacture physical hardware, they can be "simulated entirely with AI." xAI filed the trademark on August 1st but he tweeted about it back in 2021.
My take: Classic Musk - a cheeky name with a serious (or delusional?) mission. Building an AI company is one thing, but challenging to replace Microsoft's entire ecosystem? This sounds like another "we'll have a million robotaxis by 2020" promise. But if he pulls this off even partially, it would prove every software job is replaceable. Sleep well, developers.
1. Madespace.ai - Interior Designer That Actually Has Taste
Madespace uses AI trained by actual interior designers in California to redesign your rooms without text prompts. You upload a photo, choose your style, and get professionally curated designs with instant shopping links.
My take: It’s an AI tool that understands aesthetics isn't just about following rules but about taste. Perfect for people who know their space looks terrible but can't articulate why.
2. Grok 2.5 Goes Open Source - Elon's Gift to Developers (With Strings Attached)
xAI just released Grok 2.5 model weights on Hugging Face for free download. It's a 500GB package that requires 8 GPUs with 40GB+ memory each to run. The catch? It comes with a "custom community license" that has "anti-competitive terms" according to AI engineers. Grok 3 will follow in 6 months.
My take: Free weights sound amazing until you realize you need ₹50+ lakh worth of GPUs to run it locally. The restrictive license means this isn't truly "open source" - it's more like "look but don't really touch."
3. Omnara - Claude Code Finally Escapes the Terminal Prison
Y Combinator-backed Omnara lets you run Claude Code sessions and switch seamlessly between terminal, web, and mobile. Start coding on your laptop, get a push notification when Claude needs approval, and respond from your phone while grabbing coffee.
My take: This solves a big problem with Claude Code - being chained to your desk. The "human-in-the-loop" approach means you're not babysitting terminals anymore. The real genius? They built the infrastructure layer that every AI agent tool will eventually need.
Coming back to me drowning in repetitive questions…
Google Gemini Gems helped me clone my expertise, not just my responses. Basically a custom AI assistant that thinks like me.
Step 1: Setting Up Your First Gem
The Simple Setup Process:
Visit http://gemini.google.com
Click "Gem Manager"
Hit "New Gem"
Step 2: Write your instruction prompt
Here’s my exact prompt for the AI Tool Consultant Gem
You can write your own prompt based on what you wish to create, in the instructions window.
You can name your Gem (I used "Vaibhav’s AI Tool Consultant")
Step 3: Train it on your own data
You can upload your own files (research, reports, statistics, pdf guides etc) to train your gems.
I uploaded a pdf document which had the details of my most recommended AI Tools.
Step 4: Save your Gem and you are ready!
Click on the save button on the top right and that’s it. Done!
You would be taken to the chat window where you can talk to your newly made Gem.
Here’s the response:
You see how it asks(in Hindi) if I want a free tool or a paid one?
I asked it to respond in English and…
Yeah I know he has a bit of attitude (I don’t btw)...but still does a good job.
It goes a step further to give paid options as well.
Result?
Before Gems:
6 hours/week spent on DMs
24-28 hrs average response time
Constant guilt about delayed responses
Repetitive strain from typing similar answers
After Gems:
45 minutes/week spent on complex queries only
2-hour average response time
90% of questions handled automatically
More time for actual content creation
My DM response rate went from 60% to 95%.
I gained 2,847 new followers just from word-of-mouth about "Vaibhav's AI assistant that actually gives good advice.
The 3 Mistakes To Avoid while using Gems
Mistake #1: Generic Prompts Don't write "help people with business questions." Be specific about your methodology, biases, and approach. Generic prompts create generic responses.
Mistake #2: No Personality Training If your Gem sounds like ChatGPT, people won't trust it. Include your communication style, common phrases, and perspective on the industry.
Mistake #3: Trying to Handle Everything I initially tried to build one "super Gem" that did everything. It was terrible at everything. Three focused Gems work infinitely better than one confused generalist.
Reply to this newsletter with how you are using AI at your work. I’ll share the AI tools pdf I used to train my gems with the first 30 people who respond.
Waiting to read your responses!
Until next time,
Vaibhav 🤝
If you read till here, you might find this interesting
#AD 1
In 2010, a Grammy-winning artist passed on investing $200K in an emerging real estate disruptor. That stake could be worth $100+ million today.
One year later, another real estate disruptor, Zillow, went public. This time, everyday investors had regrets, missing pre-IPO gains.
Now, a new real estate innovator, Pacaso – founded by a former Zillow exec – is disrupting a $1.3T market. And unlike the others, you can invest in Pacaso as a private company.
Pacaso’s co-ownership model has generated $1B+ in luxury home sales and service fees, earned $110M+ in gross profits to date, and received backing from the same VCs behind Uber, Venmo, and eBay. They even reserved the Nasdaq ticker PCSO.
Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.
#AD 2
Every handles the operational grind — from incorporation to banking, payroll, and compliance — so you can focus on what matters most: building your company.
Start by incorporating your company for free, then access tools that grow with you. No busywork. No guesswork. Just a clear path forward.