ESD
Hey,
Komron here, and I'm back with another weekly edition.
Here's what I've got for this week:
🐧 How Tencent went from a copycat chat app to China's operating system
📈 ESD Trending: 5 startup stories of the week
💡 The 11 ideas YC wants to fund in Winter 2027
Let's get started!
🐧 The Copycat That Became China's Operating System
There's no company like Tencent. Facebook, PayPal, Nintendo, Spotify, Netflix, and Shopify, all rolled into one. Plus a 21st century Berkshire Hathaway.
But most people outside China have never heard of it. This is how a copycat chat app became the infrastructure a billion people live on.
The copy
It started in 1998. Pony Ma and four friends borrowed $120,000 during the Asian financial crisis.
Their first product was a copy: a Chinese version of ICQ called OICQ. Millions signed up. Then AOL threatened to sue. They renamed it QQ.
The near-death moment
By 2000, QQ had millions of users. And Tencent was 3 months from running out of money. They tried ads, subscriptions, premium features. Nothing worked. Chinese internet users didn't pay for online services.
Pony Ma kept this secret from most employees. He stayed calm while looking for anyone who would invest. Finally, IDG Capital and PCCW put in $2.2 million for 40% of the company. Tencent survived with weeks to spare.
The penguin
But they still had no business model. Then they noticed something strange: users were obsessed with QQ's penguin mascot. They wanted to dress it up. Make it theirs. Western companies would have ignored this. Pony Ma didn't.
In 2002, Tencent started selling virtual clothes for penguin avatars. Chinese users wouldn't pay for software. But they'd pay for status. For identity. Digital items that cost nothing to make, with profit margins above 90%. By 2003, virtual goods were making millions while Western tech was still recovering from the dot-com crash.
The games
Next came gaming. Tencent licensed Korean games instead of building from scratch. Every QQ user was already a potential gamer. Marketing cost: zero.
In 2008, they invested in Riot Games, and in 2011 they bought a majority stake for a reported $400 million. League of Legends now has 100 million monthly players.
The bet on mobile
By 2010, Tencent seemed unstoppable. Then smartphones arrived. Users were abandoning desktops. If Tencent couldn't move to mobile, everything they built could disappear.
In 2011, they launched WeChat. It had a simple feature: voice messaging. Hold a button, speak, send. For Chinese users, this was huge. Typing thousands of characters on a tiny keyboard was painful. Within about 14 months, WeChat had 100 million users.
In 2014, WeChat introduced digital red envelopes: the Chinese New Year tradition of giving money, brought online. Over the holiday, 8 million users sent 40 million red envelopes. Millions linked their bank accounts. WeChat Pay took off.
Today, WeChat is how a billion people in China live. Taxis. Doctor appointments. Bills. Food. Investments. One app for everything. It's not an app. It's infrastructure.
The radar
And Tencent never stopped investing. Stakes in over 800 companies: Epic Games, Spotify, Tesla, Snap. Every investment gives them a window into what's coming next. A global tech radar no one else has.
The lesson
For years, Tencent was dismissed as a copycat. Now Western tech giants keep trying to copy WeChat's super-app model. The imitator became the innovator.
Pony Ma built all of this without becoming famous. He rarely talks to media. He avoids the spotlight. While Western founders became celebrities, he just shipped products. The founder of one of the most valuable companies on earth, and most people have never heard of him.
ESD Trending: 5 hot startup stories of the week
Aarush Yadav and Arnav Kumar's glasses read the tiny muscle movements in your face when you silently mouth words. Radio waves plus AI turn them into text at 200+ words a minute. They work with ChatGPT, Claude and Slack, so you can message, ask AI or code without typing or speaking. They raised $1.3M from YC, Afore Capital and angels.
Lukas Gruber posted on X that Instinct is raising at $30B, just a month after its $1B round at $10B. The post passed 1M views. Nobody has confirmed it yet. But the $10B round also started as an unconfirmed report in mid-September, and Instinct announced it 12 days later.
Restaurants and groceries, one company. DoorDash will pay $65 a share in cash and stock, about 26% above Instacart's price before rumors leaked. Instacart keeps its brand and app. Regulators still have to sign off, so it won't close before late 2027. Delivery in the US just got a lot less crowded.
Thrive Holdings buys old-school businesses and rebuilds them with AI. It already owns 75+ firms in accounting, IT services and the built environment. Now it's taking ADT, the 150-year-old home security company, private for $10 a share in cash, a 46% premium. The AI roll-up playbook just jumped from small firms to a public company.
Spawn founder Jacob Sansbury says a "fleet" of named AI agents runs most of his game startup, from the engine to finance. They shipped 77,000 changes in 30 days, around the clock, with no pull requests. Token spend is 30x payroll. The numbers are self-reported, but his post on X went viral.
💡 YC Just Told You What It Wants to Fund in Winter 2027
YCombinator published its Requests for Startups for Winter 2027: 11 ideas its partners want founders to work on. The theme: AI's bottlenecks are moving past the models, into hardware, infrastructure, finance and specialized software. YC makes clear that you don't need to pick one to apply.
The 11 requests:
Abundant space access: businesses that only work once launch costs fall 100x.
The $15-an-hour expert: AI that lets non-specialists handle benefits, taxes, claims, medical billing and immigration.
Robots made in America: robots and their supply chain, from motors to sensors.
LLMs that write well: a model people actually enjoy reading.
Plaid for government benefits: verify income and residency once, apply everywhere.
Wall Street for the physical buildout: loans, leasing and hedges for robots, GPUs and data centers.
Vertical foundation models: small specialist models for manufacturing, energy and biology.
AI for City Hall: starting with building permits.
Data center supply chain: transformers, switchgear and cooling.
AI for the air-gapped economy: frontier AI that runs inside banks, hospitals and defense contractors.
Own your intelligence: helping companies train and run their own models.
Our take:
Follow the deadlines. New Medicaid re-checks and FCC rules on US-made robots both start in 2027. A rule with a date on it is a sales calendar.
Boring buyers are back. Cities, agencies, banks, hospitals. Slow to sell to, but big and sticky budgets.
Start as a service. "The $15-an-hour expert" and "AI for City Hall" can begin as a service with AI behind it, before you write much code.
Test one this week: pick the idea closest to a job you've done, call 10 people who live the problem, and do the job by hand once with AI tools. If you can't make it faster manually, software won't save you.
"The full list, with each partner's reasoning, is worth 15 minutes of your weekend. Read it on YC's site,then reply and tell us which one you'd build."
That’s all from me today,
See you next week 🫡
Komron & the ESD Team
Still testing the format, one deep story, five quick hits, and one practical thing per edition. Too much? Too little? Reply and tell us what you’d change. I read every single one. 🙏🏻