2025-06-01 08:00:00
Every social product needs a screenshot moment that can turn the user’s progress into shareable status symbols.
I was listening to a podcast with the CTO of Duolingo yesterday. He mentioned how obsessed people are with streaks on the app.
It was the same with Snapchat Streaks. I was never a Snapchat user, but I still remember how much people used to talk about them. Kids wouldn’t just send snaps, they were obsessed over maintaining streaks. People would even give friends their login info before going on vacation, just to make sure the streak didn’t break.
When there was this power outage in Portugal, Duolingo had to let people there continue their streaks because a lot of them were upset about losing their streak due to something not in their control.
These streaks, basically anything you can screenshot and be flexed online, are the ultimate growth loop. And smart products like Duolingo know it. That’s why they double down on shareable elements: streaks, XP badges, leaderboard rankings, and now, a fluency score you can post on LinkedIn to tell the world how good your German is. Yes, the CTO mentioned that they are coming with a score for language learners.
Nobody knows what “Intermediate Spanish” or “Limited Working Proficiency” means anyway. A 90/100 score in German is easier to brag about.
All the stuff that keeps popping up on my timeline —
All of these are the same thing.
It’s signalling.
If I were building any consumer product in fitness, health, education, or productivity, I’d work backwards from this question: What’s the most impressive thing my users could screenshot and post on social media?
Think distribution first.
That’s how someone finds out there’s an app that tells you your biological age. That’s how people get curious about their own fluency score or fasting streak or mental fitness level. Nobody sits around thinking “what do my lifts compare to,” but if their friend posts a screenshot saying “yours is like lifting a truck”, people get curious to track their lifts too.
These things aren’t just fun, they’re retention mechanics disguised as social proof:
They’re free marketing too.
All of these show up on your feed without the company spending a dime. And each one makes you wonder: should I also be doing this?
If you’re building a social product, don’t just focus on features. Focus on status objects. The screenshot. The thing that people will share on x dot com.
And if you do it right, people will do your marketing for you.
2025-05-31 08:00:00
I am getting smarter and dumber at the same time thanks to LLMs.
I’ve been forcing myself to spend at least an hour every day chatting with LLMs for the last few months.
Not the way most people do it though. I don’t treat ChatGPT like a therapist. I don’t use it to process my personal life. I definitely don’t feed it prompts like “analyse my chat history and tell me who I am.” I have zero interest in being psychoanalysed by a chatbot.
And I also know that someday OpenAI is going to run ads. And the last thing I want is to start seeing targeted offers based on my deepest vulnerabilities.
Where LLMs truly shine for me is in high level thinking. Today I was analysing the business model of some late stage Indian startups, and the back and forth with ChatGPT genuinely sharpened my own frameworks. It’s like having a sparring partner who never gets tired and always shows up ready to challenge your assumptions.
Same with technical projects. I’ve used it to understand complex workflows, even set it up to question me until my thinking got clearer. That kind of Socratic prompting has been huge for improving how I reason through architecture and implementation. I’m probably getting dumber at the basics though.
I barely type anymore. Everything is a thought dump through voice to text. I rely on LLMs to rewrite things so much that even this reflection you’re reading is rewritten by a Editor GPT and then passed through Claude Opus for another round of editing. My default writing muscles are slowly decaying. Sometimes I wonder if I can even write a simple email without AI assistance anymore.
There’s this weird polarity happening. At the high end, I’m getting sharper than ever. My ability to synthesise complex information, spot patterns across industries, and build mental models has improved dramatically. But on the low end, I’m delegating so much that I might forget how to write a basic text like a normal human. Even my basic whatsapp messages are now being rewritten by ChatGPT.
I think a lot of power users of LLMs are going through this cognitive bifurcation, even if they don’t say it out loud. We’re getting sharper at high level synthesis while atrophying at routine tasks. In my case, it’s showing up as better systems thinking but declining baseline communication hygiene.
2025-05-30 08:00:00
No one knows business which model is better long term. And that’s the most honest thing I can tell you about business strategy.
Take Zepto. When they started with their opex-heavy model: local dark stores, pre cooked food, delivery workers almost working like full time employees, always on standby so that they can do 10 min delivery, everyone was like “This is it! Absolute control over the delivery experience. 10 min delivery. This is how you win food delivery!”
They launched Zepto Cafe. Scaled it super fast. Launched it as a separate app. Mostly frozen food in 10 minutes. Zomato followed. Swiggy followed. Standalone companies popped up everywhere trying to copy the model.
Fast forward to today. Zepto’s prepping for IPO and guess what? They’re scaling down ZeptoCafe. Suddenly the “future of food delivery” doesn’t look so futuristic when you need to show profitability to public market investors.
So was Zepto Cafe a good model or not? Nobody knows. And that’s the point.
This same thing will playing out with Snabbit vs Urban Company.
Snabbit’s going full opex heavy - positioning workers (they call them experts) close to high demand areas, treating them almost like full time employees with insurance, providing e-bikes for female staff. They’re promising 10-minute service for daily chores.
Thanks to this model they will have high worker utilisation, low travel dead time, defensible density economics. A stricter ETA forces operational excellence; builds a habit loop around speed and reliability. Though I would argue Urban Company is pretty reliable too.
This is again a case of classic counter positioning. Urban Company operates the light touch marketplace model. Mostly freelancers, mostly low-frequency but high-ticket services like beauty, furniture and electronics repair. They can’t suddenly position their freelancers to fulfil 10 min delivery. For that you need even stricter SOPs, more control over the service workers, and the ability to have more control over their work hours and availability during those hours. Snabbit is going mostly after the home services market.
When Homejoy started (more than a decade back ago), they went after the services market and got destroyed by platform leakage. If someone’s coming to your house every day and the marketplace is taking a cut, what happens is that the supplier builds a relationship with the customer and the transaction moves offline. There are actually 2 cases when this happens:
Why would you pay platform fees for someone you see daily? You wouldn’t. You’d just get their number and pay them directly.
That’s why Uber works so well. You can’t build a relationship with your driver. It stays transactional.
(A friend wrote that Namma lets you favourite drivers now, so that you can get matched again with the same driver, which is possibly the worst feature you can implement as a marketplace.)
I have discussed this concept in detail in my ‘Marketplace for therapists’ post.
Yes, Snabbit experts will cover several adjacent buildings, not one home. And you can ensure that a customer rarely gets the same person two days in a row, so bilateral trust never deepens enough to go off app. But if service quality is similiar and you are happy with someone who you have got allocated 3/4 times, you might want to take that relationship offline.
Yes, these workers will stick to the platform till they keep earning ₹40k per month, but we have all seen how gig economy workers’ earnings drop over time once you stop subsidising the supply cost and cut your incentives.
There are other assumptions too:
Is Snabbit model better than Urban Company’s? Who knows.
What I do know is that when you counter position against an incumbent, timing is everything. You want to hit them when they can’t respond - like when they’re about to IPO.
Think about it. Urban Company has spent years building a lightweight marketplace. Their entire P&L, their processes, their margins - everything is optimised for that model.
Up front costs (relocation subsidies, insurance, e-bikes) are fine for a young startup because they’re baked into the model. For Urban Company, they just can’t increase opex when they have to IPO.
If they suddenly try to match Snabbit’s 10-minute promise, they’d have to:
Ride hailing and food delivery has tried zone based supply in the past. It helps in higher utilisation when there is predictable high demand. It also costs you more because unless you have a differentiated fleet who is paid more supply is not going to sign up for the same price, without any added incentives.
By formalising workers with eKYC, insurance, and rapid response safety, Snabbit gets ahead of upcoming gig worker regulation. If laws tighten, Urban Company might face issues while Snabbit will be already compliant.
And if Urban Company is thinking about IPO this year, no way they’re making that change now. Yes, they have launched Insta Help, but I assume it is still more of an experiment than changing their entire company. Similiar to how Swiggy and Zomato launched separate 10 min food delivery apps instead of reinventing their entire company to optimise for that delivery SLA.
Counter positioning is about adopting a model that would destroy the incumbent’s economics if they copied it. So they don’t copy it.
But just because you can counter position doesn’t mean your model is sustainable.
Remember Zepto Cafe? Everyone thought 10 min food delivery is the future until Zepto needed to show better numbers for IPO.
Zomato parent Eternal has already ended its Quick 10 min delivery pilot
The same thing could happen to Snabbit. Sure, they’re treating workers well, providing insurance, keeping them in specific zones. But is this sustainable? Can they maintain high utilisation? If there is high utilisation between the same supply and demand, will there be platform leakage?
What really matters isn’t the model itself - it’s where the company is in its funding cycle and what narrative they need to sell.
A challenger who has the capability to counterposition and win marketshare? Sure, why not take an opex heavy approach. A pre IPO company like Urban Company and Zepto? Maybe not.
And this is why I am assuming there will be a lot of entrants in this 10 min services space, like 10 min food delivery a year back.
tl;dr Perception of business model success = Where you are in the funding cycle × Whether incumbents can afford to copy you.
2025-05-24 08:00:00
I saw someone tweet we should build the TBPN for India. As someone who’s followed TBPN from day one, let me be clear: you can’t build TBPN. It’s an N-of-1 product.
This reminds me of when the ‘Derek Guy’ was blowing up on Twitter. People said: “If you want to grow fast today, just do what this menswear guy is doing.” However, the problem is: you’re not him. You didn’t grow up obsessed with men’s fashion since you were a teenager. You didn’t spend a decade watching, learning, critiquing, and fighting with people online. You will never have his deep knowledge of menswear, his biting tone, and years of immersion in fashion.
Yes, you can mimic the TBPN format. You can recreate the delivery. You can copy the lighting, the sound effects, you can even create vibe reels. But you can’t copy the taste of John Coogan and Jordi Hays. You can’t fake their love of startups. Showing up every day to do 3 hour live streams. You can’t manufacture the charisma or the humour or the edge.
Similarly, when the All in podcast was blowing up during the pandemic, a bunch of people tried to replicate the format. Logan Bartlett, Nikita Bier, and Zak Kukoff gave it a shot. They created the 3 Cartoon Avatars podcast. I remember half a dozen other copycats too. Unfortunately, it didn’t land for any of them.
Nikita’s humor, which works brilliantly on Twitter, just didn’t translate to the podcast arena. Their dynamic felt forced. The chemistry wasn’t there. Consequently, the whole thing came off like a weak copy of the All in crew.
Ironically, Logan actually did become a good podcaster — but only after the others left. He leaned into what he’s uniquely good at: his takes as a VC, his ability to ask smart questions, and his access to top tier guests willing to come and share real alpha. That’s when his podcast clicked and it became his thing.
The truth is, just because you’re good at shitposting on Twitter doesn’t mean you’ll be good on a podcast. Every platform amplifies different strengths. Moreover, if you’re just trying to copy a format without knowing what your strengths are, you’ll end up sounding like a knockoff.
It’s a common trap: seeing someone else win and thinking you can just “do that.” However, most formats aren’t plug and play. They’re built around people’s lived experience, tone, chemistry, and voice.
And when that’s missing, the audience feels it instantly.
I would not recommend copying TBPN. It just wouldn’t work.
2025-05-20 08:00:00
I was reading about Lindy and noticed they chose to build on top of n8n instead of Zapier.
Here is why they did it (based on Tegus transcripts, summary by NotebookLLM):
While Zapier was acknowledged as having significantly more integrations (around 1600 compared to n8n’s approximately 600 at the time), both were considered “champions” among the vendors they researched, having the most horizontal integrations.
It is worth noting that the initial technical integration with n8n was considered a “con” compared to Zapier, requiring about two weeks of engineering time to build a specific adaptation layer. Additionally, n8n had less support for federated OAuth compared to Zapier at the time, which was another downside Lindy experienced later. However, despite these points, the advantages in terms of scaling cost, performance, and open-source benefits led Lindy to choose n8n.
That got me curious. I’ve been seeing more and more n8n tutorials pop up on Twitter and YouTube lately. A lot of people already know about n8n, but I think this is a good opportunity to look at how n8n carved out a niche against a dominant incumbent like Zapier.
(I know Zapier has since changed their pricing model, they also want to grow in the age of AI, but this post is not about the current state, it’s about how n8n was able to counter position and grow over the years.)
Here is how n8n counter positioned its way into Zapier’s market.
Zapier was the unquestioned king in workflow automation. For over a decade, it defined the category of no code automation, enabling millions of marketers, founders, and small teams to automate workflows between thousands of apps.
It was slick, simple, and built for non-technical users.
So how did n8n, an open source, developer first tool launched years later, manage to carve out a fast growing niche in Zapier’s backyard?
It’s a classic case of counter positioning. They offered a fundamentally different product for a different audience, and in doing so, changed the rules of the game.
Zapier was built in the early 2010s for the exploding world of SaaS tools. Its pitch was elegant: “Connect your apps and automate workflows.” Whether you wanted to add Gmail leads to a spreadsheet or send a Slack message every time a new Stripe payment came through, Zapier had you covered. No coding required.
But as the years went by, Zapier grew into a closed ecosystem. Its monetization was usage-based, charging per task or per integration. While this worked well for lightweight use cases, it became increasingly expensive and restrictive for complex or high-volume automations.
At the same time, developers were left behind. Zapier was never built for teams who wanted to write custom code, build loops, or integrate with complex APIs. It was simple, and that simplicity became a constraint.
Enter n8n (node automation), a tool that looked similar at first glance with drag-and-drop workflows, but was fundamentally different under the hood.
n8n was open-source. Anyone could self-host it, inspect the logic, and extend it.
Unlike Zapier, which charged per integration or task, n8n adopted a workflow based model, focusing on orchestration rather than atomic triggers. This unlocked more ambitious use cases without punishing users for high volume usage.
You could write JavaScript directly inside nodes. This meant developers weren’t boxed into pre built connectors. They could manipulate data, build conditionals, and integrate with APIs flexibly.
For companies dealing with sensitive data, being able to run their automation on their own infrastructure was key.
This wasn’t just an alternative to Zapier. It was a reframing of what automation could be.
n8n didn’t try to outdo Zapier at being Zapier. Instead, it positioned itself for a different worldview. Zapier targeted non technical users with a closed, cloud only, task based billing model with very limited customization. n8n went after developers, data and infrastructure teams with an open source, self hostable, workflow-based model where you could code if you wanted to.
Rather than attacking Zapier head on, n8n flanked it by targeting the technical community who wanted power and customization, not just convenience.
And it worked. Today, n8n powers internal automation at companies from startups to enterprises. It’s used to orchestrate entire internal data flows, build data lakes, and even replace lightweight ETL pipelines. From a user experience that looks simple on the surface to what’s under the hood, it’s all about developer level control.
The n8n vs Zapier story is a blueprint for any startup looking to enter a mature category.
Don’t copy the incumbent, change the frame. n8n didn’t say “we’re like Zapier but cheaper.” It said “you can own your automation stack, customize everything, and run it wherever you like.”
Start with underserved users. Zapier wasn’t bad, it just wasn’t built for developers. n8n won by focusing on the power users who needed more control.
Monetize differently. Zapier’s per task model made large scale workflows expensive. n8n’s workflow based and open core model let users scale affordably.
Open source isn’t just a feature, it’s a distribution strategy. By being open source, n8n invited contributions, drove word of mouth, and lowered barriers to adoption. It spread like infrastructure tools often do: organically, from dev to dev.
Zapier is still a great product. But n8n showed that even in a mature market, there’s room to challenge the default if you’re willing to think differently about pricing, audience, and product philosophy.
Most people try to build a better version of what already exists. The smart move is to build something fundamentally different for people who have been ignored by the market leader.
Is this real counter positioning?
It’s a fair question. Everyone loves throwing around strategy terms, but most of the time they’re using them wrong.
What is Counter-Positioning?
Hamilton Helmer defines it pretty clearly in 7 Powers: A newcomer adopts a business model or operating model that the incumbent could technically adopt, but doesn’t because doing so would mess up their existing business.
The key insight is that the incumbent can’t easily fight back without shooting themselves in the foot. The newcomer gets a temporary advantage because they’re operating under a better model for their target segment, while the big player is stuck.
So does n8n fit this?
Let’s look at the three ways n8n positioned itself differently:
Zapier charges per task. This gets expensive fast if you’re running high-volume automations. In contrast, n8n offers workflow-based pricing and self-hosting options.
Could Zapier ship this? Sure, technically. But think about what would happen if they offered flat pricing or self-hosted plans tomorrow:
That’s exactly the kind of trap counter-positioning creates. Zapier can see what n8n is doing, but copying it would hurt their existing business.
Zapier built for marketers and ops people who want simple drag-and-drop automation. Meanwhile, n8n went after developers who want custom JavaScript, CLI tools, and extensibility.
If Zapier suddenly pivoted to serve developers with scripting capabilities and complex APIs:
Again, the cost of following n8n is too high.
This is the big one. Zapier is a closed platform with proprietary integrations. N8N is open source, enabling self-hosting, community contributions, and custom nodes.
For Zapier to go open source would mean:
Never going to happen.
But is it perfect counter-positioning?
Not quite. n8n didn’t flip Zapier’s profit formula as dramatically as, say, Netflix did to Blockbuster (rental fees to subscriptions) or Tesla did to traditional automakers (dealer network to direct sales).
n8n is still fundamentally an automation tool. They repackaged the value proposition around developer control and self-hosting, but it’s more of a technical re-segmentation than a complete business model revolution.
They flipped three strategic levers that Zapier can’t easily follow without significant pain.
The real test is this: Can Zapier easily neutralize n8n’s advantage? The answer is no, not without messing up their core business. And that’s what makes it counter-positioning.
2025-05-19 08:00:00
While transitioning from my last job, I took a small vacation in Sri Lanka and that time I was reading two books and thinking very deeply about both. Two books that couldn’t be more different.
One was Michael Singer’s The Surrender Experiment. The other was Robert Caro’s biography of Lyndon Johnson, The Path to Power.
I realised I was basically reading two opposite instruction manuals for life.
Singer’s story is wild. In the 1970s, he was living in a shed in Florida, meditating, building a small yoga temple. He came across a microcomputer. Taught himself to code, wrote some billing software for a local doctor, and before he knew it, he’s co-founded Medical Manager, a company that ended up dominating physician offices across America, IPO’ed in 1997, and got acquired by WebMD for ~$5 billion in 2002.
Michael Singer swore he never had a plan. He wrote two books about it.
I read his book The Untethered Soul a while back, and both The Surrender Experiment and Untethered Soul serve as play books on how to say yes and surrender to the flow of life. He just kept saying yes to whatever showed up.
While he was busy surrendering, some executives were cooking the books. The DOJ came knocking in 2009. He settled for $2.5 million before charges were dropped. Turns out saying yes to everything has its downsides.
LBJ is the complete opposite. This guy had his path to the White House mapped out while he was a kid.
Teacher’s college, congressional aide, House seat by 28, Senate next, presidency eventually. Every move was calculated. Relentlessly executed.
Caro describes LBJ holding meetings while sitting on the toilet because he couldn’t waste a single minute. The man was pathological about control.
In 1941, he made a rare mistake. During the final count of a special Senate election, he relaxed for once, went to his ranch instead of watching the vote count. Lost by 1,311 votes when his opponent’s team stuffed the last ballot boxes.
This never happened again. In 1948, he won his Senate seat after running his own ballot operation with military precision. Kept climbing until he became the President in 1963.
Reading these books in parallel messed with my head. Two guys who couldn’t be more different, and both reached the top of their fields.
Singer’s openness, and the willingness to say Yes without overthinking, created billion dollar opportunities he never could have planned for. But it also let fraud happen right under his nose.
LBJ’s iron grip got him to the presidency.
Steve Jobs said you can’t connect the dots looking forward, you can only connect them looking backward, once you’ve reached some end point in life and can see how the pieces fit together. But I think LBJ would disagree. You can connect the dots forward. You can obsess over them, and you can will that life into existence. But not everyone is LBJ.
Or even Singer.
We can build our life like Singer, one unexpected opportunity at a time. But some of us also crave LBJ’s certainty, his step by step master plan.
I went from tracking each day of my life, doing yearly OKRs, to now just following my gut. Saying Yes to things I am interested in. No concrete plans.
Yes, there is some vague direction, but no real sense of control. Easy to do now after a decade of saving, building a niche network, and developing a skill set, but Michael Singer did it from day one. For me there is this constant struggle between trying to control everything and letting go.
Maybe there’s a time for surrender and a time for control. The trick is knowing which moment calls for which approach.
Both men, Michael Singer and LBJ had opposite approaches, both led somewhere interesting. For now I am in my Singer mode: say yes, fuck around, see what happens.
I’ll run this playbook for a year and report back.