Tinkering with code, capital and content when not hit by existential questions :)
Bangalore, India
Joined April 2025
- Tweets4.2K
- Following1.2K
- Followers369
- Likes4.3K
Pinned Tweet
I've been doing SEO for about 2 years now, entirely self-taught via YouTube videos, some articles, sub-reddits and a lot of daily practice across multiple blogs I run. Some of them have steady impressions and are slowly climbing. One of them crossed 200+ daily clicks and has held it consistently, which honestly felt like a big milestone after all the grind.
Sharing some of the things I've picked up.
Here is a 49-pointer, nano-course on SEO for beginners:
FOUNDATIONS
-----------------
1. SEO is not about tricking Google . It's about being the best answer. Once I stopped thinking "how do I game this" and started thinking "how do I genuinely help this searcher," my content got better and rankings followed.
2. Crawlability before everything. If Googlebot can't reach your pages, nothing else matters. Check robots.txt regularly. I accidentally blocked a whole section of one of my blogs for weeks before noticing.
3. One page, one primary keyword. Don't try to cram 10 different topics onto one page hoping to rank for all of them.
4. Search intent is the most important concept in SEO. Before writing a word, Google the keyword yourself. Are results blog posts, videos, product pages, or tools? Match the format — Google already knows what users want. Dont waste this insight.
5. Understand the difference between informational, navigational, and transactional queries. You write and optimize very differently for each.
6. Google Search Console is non-negotiable. Set it up on day one. It's free, it shows you exactly what's happening with your site in search, and it flags issues Google actually cares about.
7. Google Analytics alongside GSC gives you the full picture. GSC tells you how you appear in search. GA tells you what users do after they arrive. And where they arrive from ( you also want to track if AI is citing and sending traffic to you apart from traditional Search engines)
8. SEO is slow. Seriously, prepare for that. I used to check rankings every day in the first few months. New content can take 3–6 months to rank. Be patient with SEO results. It takes time.
Ambika Iyer retweeted
I only build 2 types of startups now:
1) AI-proof in 10 years. My startup marketplace (TrustMRR) can be cloned with a prompt, but a network of 20K buyers can’t. Trust compounds over time.
2) Trends. Anything hot on tech Twitter this month. GEO, new AI models, it can be anything, as long as people are actively talking about it.
I also have a 3rd filter: I build products I’m curious about. Or, even better, products I use myself.
At yesterday's meetup I asked Marc why he's never shipped an AI wrapper 😅
His two most profitable products have almost nothing to do with AI. Everyone else is chasing the AI hype. He just doesn't.
His answer had two parts.
1. he builds things he wants to use himself every day.
2. he asks what still won't be replaceable by AI in 5 or 10 years. His example was TrustMRR. Real buyers and real sellers who trust the place. An AI wrapper can die the day OpenAI ships one new feature. But nobody ships a feature that takes your buyers and sellers away.
So you pick the niche, and you don't chase whatever is hot this month. With AI there's a new hype almost every single month. You pick, and then you wait. The compounding only shows up years later. And if you don't actually love the thing, you won't stay long enough to see it.
There's a word going around in China right now, OPC, one person company. Everyone wants to build one, and with AI making everything fast and easy, everyone wants one that prints money fast.
Marc told us he wanted to build a social network early on, right after watching The Social Network. He wanted to get rich. Everyone starts there 🤪
We're both just normal people, zero followers, zero products, no idea how to get users. Honestly we still think about that last part every day.
If you're starting one of these, pick something you like and are already a bit good at. Ship one feature in your MVP. See if anyone wants it. If they do, stay on it and listen to the people using it.
Then go tell the story yourself and build your own flywheel.
Ambika Iyer retweeted
I built a map of every passenger plane in the sky using Claude Opus 4.8.
10,811 planes. 694 airlines. One real moment.
→ Every mark is a real aircraft from live ADS-B data
→ Shape points the way it flies, colour is altitude
→ Click any plane to see where it took off and lands
→ Filter 694 airlines
Data from adsb.lol, adsbdb and OpenFlights.
Drawn with D3.
100% Free and open-source.
Shoutout to @BigBodyCobain for building this and making it open-source for the community.
Geez. We really got to 10k stars! Thanks, guys!
github.com/bigbodycobain/sha…
My mom seems to be building some sort of “plantesteins” in her balacony garden(more of a lab according to me) !!!Some of these leaves are gigantic!!😁😁 #balconygarden
Ambika Iyer retweeted
Driving in one way service road in marathalli on 29th July around 8:00am and drivers audacity is such that he said do complaint and nothing will happen. Registration number of the car is available in the video @blrcitytraffic @DCPTrEastBCP – at Anekal Bangaluru, India
Ambika Iyer retweeted
This year, I’ve productionized three agentic AI systems for clients.
A few things I’ve learned along the way:
1) Tool calling is where the real value starts. Without tools, it’s just a chatbot.
2) Don’t build one giant agent. Split work across smaller agents with clear responsibilities.
3) Put your business systems behind MCP servers or APIs. It keeps the architecture cleaner and easier to maintain.
4) Use your existing identity provider for authentication. Don’t invent a new login system.
5) Enforce permissions at the tool level, not in the prompt.
6) Shared state matters. Agents need context to handle follow-up questions properly.
7) Remove or mask sensitive data before sending it to any external model.
8) Build guardrails into your workflow. Don’t rely on prompts to enforce business rules.
9 Use structured outputs whenever possible. They make integrations more reliable and easier to validate.
10) Keep humans in the loop for high-risk actions. Not every decision should be fully automated.
11) Continuously test and evaluate your agents. Small changes can break real workflows.
12) Add proper logging and tracing. If you can’t see what your agents are doing, you can’t fix them.
13) rack token usage from day one. Costs grow much faster than you expect.
14) Version your prompts, tools, MCP servers, and knowledge base. Debugging is almost impossible without it.
15) AI doesn’t replace security. WAFs, API gateways, rate limits, and access controls are still essential.
One thing I’ve noticed:
Building an AI demo is the easy part.
Taking an AI system to production is a software engineering problem.
That’s where most of the work happens.
Finally seeing budgets reinstated on LLM use in my 9-5 . Unlimited was always risky. It was a wild 4 month ride :)
Ambika Iyer retweeted
Arnab Goswami raises some very important questions in the video below.
The choice for the nation is between sober rational people we may sometimes disagree with (even vehemently on occasion) vs an irrational mob, with clear instigation from forces that want to destroy India, as Arnab outlines.
Let me be clear: there will never be a government that any single person can ever agree 100% with. As an example, I have long had strong philosophical opposition to high stakes standardized exams, which (as countries like South Korea and China have discovered) ultimately result in "teach to the test" and extreme pressure on children. China has banned for-profit tutoring companies (used to be a $100 billion market in China) but continues with their high-stakes Gaokao exams. But that is a separate rational debate and I am a nationalist and patriot first.
youtu.be/vudl-w_3D6Y
Ambika Iyer retweeted
Today, I just narrowly escaped on ORR soon after entering from the Shamsabad Entry. I was negotiating a turn when a RMC truck suddenly came from the wrong side. God just saved me. Please take necessary action.
Ambika Iyer retweeted
LIFT-OFF! 🚀 Vikram-1 has left the pad at Sriharikota. India's first privately developed orbital rocket is flying. History is being made. 🇮🇳
#Vikram1 #JourneyToOrbit #SkyrootAerospace
Ambika Iyer retweeted
Built an AI-powered workflow to classify Zendesk support tickets for my client.
The problem was simple:
Support teams were receiving a steady volume of customer tickets, and each one needed to be reviewed, understood, and classified into the right business category.
Manual classification was slow, repetitive, and inconsistent.
So I built an end-to-end AI-assisted workflow on Azure.
Architecture:
• Zendesk as the ticketing platform
• Azure Functions for the API layer
• Azure AI Foundry for model orchestration
• RAG over the approved classification guide
• Azure workflow automation for processing
• Microsoft Teams channel for validation output
Flow:
New Zendesk ticket created → ticket details sent to the AI workflow → relevant guidance retrieved from the knowledge base → AI predicts the best category → returns top category recommendations with confidence and reasoning → posts the result to a Microsoft Teams channel for human review.
I also added:
• automated ticket processing
• structured category recommendations
• confidence scores
• reasoning for each recommendation
• human-in-the-loop validation
• Teams-based review workflow
• controlled classification against approved categories
This was a practical AI build covering Zendesk integration, Azure architecture, RAG, backend APIs, LLM orchestration, workflow automation, and validation design.
The key point: AI is useful only when it fits into the real workflow, supports human decision-making, and keeps the output grounded in trusted source material.
Ambika Iyer retweeted
India just dropped a semiconductor bombshell , Game on builders.
Folks, today is a massive day for our nation.
The Union Cabinet just approved Semicon 2.0 with a whopping ₹1,27,500 crore push.
This builds straight on ISM 1.0 and takes us deeper into the full chip ecosystem from silicon wafers and fabs to advanced packaging, materials, equipment and homegrown design IP.
Why this fires up international confidence big time, Global players are desperate for trusted, resilient supply chains away from single country risks.
India is now backing this with serious long term money and policy muscle.
That means lower risk, faster setup, and a partner ready to scale.
Investors see a country that’s moving from promising to proven player.
For our young engineers and talents, This is your cue. Focus on chip design, VLSI, compound semis, packaging tech, and materials science with AI assistance.
Hands on skills here will open doors like never before. We are talking lakhs of high quality jobs across R&D, manufacturing, testing, and supply chain not just assembly lines, but real innovation roles that can power millions of careers and transform India’s tech workforce.
This isn’t hype. It’s practical momentum turning India into a global semiconductor force.
The world is watching. Time to build and shine.
Ambika Iyer retweeted
From sleepless nights deep diving into semiconductor concepts to getting direct encouragement from Intel this journey is getting real.
I recently joined the European MentorChip Exchange and Intel’s reply to my thoughts on skill development, packaging just lit up the path ahead.
Their support means everything for someone upskilling from Europe and may greatly impact Odisha’s semiconductor journey , especially as they push forward with advanced semiconductor packaging here.
What started as a small personal initiative after their big announcement has now connected me to Europe’s microelectronics community and real industry feedback.
Grateful for every late night session that’s helping me contribute to Odisha’s growing semiconductor ecosystem.
If you are in semiconductor industry or business , chip design, packaging, or building the next gen supply chain, in semiconductor security, fabrication let’s connect.
therationalinvestor.in/tools…
Check out this free Goal planner -
Most people use 2 modes -
(1) They have a fixed corpus they want to reach say $5M in X years
(2) Others have a fixed SIP they are willing to commit for X years.
Given these 2 goals, find how your money would grow (and value depreciates with inflation) given your money's growth rate .
Understanding money's growth rate and inflation is critical to understand how far or near you are from your financial goals.
#personalfinance #FIRE #moneymatters #Inflation
Google with isomorphic labs is already doing it !
It’s well within Anthropic’s rights to compete in any market they choose.
What’s funny, in this instance, are the number of Pharma companies, who through their unchecked use of Anthropic, are driving revenues into what they think is a model provider but is in fact a competitor lurking in the shadows thereby accelerating their own demise.
I suspect any end market with reasonable ROCE that could be AI accelerated is on the table.
If I were them, I’d probably do the same.
