Five Slices #8: Deepseekalypse
IKEA's marketing strategy, boss fights, late bloomers, Jean Mallard's paintings
Welcome to “Five Slices.” Each post has five stories and ideas from history, business, technology, culture, and art. Some will be fun, some will be useful, all will be interesting, and they’ll take you five minutes to read.
You can check out the complete list of posts here.
1. Deepseek rattles Wall Street
Something big just happened in the AI world. Large Language Models (LLMs) have been the focus of AI tech over the last few years. These LLMs can answer questions reasonably well from almost any domain, from natural language reasoning to math. America was leading this space with OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, and Meta’s Llama considered to be the cutting-edge in LLMs. In fact, America is planning to invest $500 Billion into AI over the next four years.
6 days ago, a Chinese company called Deepseek released a model called Deepseek-R1 that beat or matched the performance of all the cutting edge models.
What’s more:
OpenAI has an investment of $13 Billion from Microsoft and spent more than $500 million on training its models. Deepseek spent just $5-6 million on its models. That’s 100 times cheaper.
OpenAI and Google’s models are running on racks of specialized GPUs in data centers in the US. A reasonably powerful version of Deepseek can be run even on an Apple Macbook.
Deepseek open-sourced its code. Anybody can learn from it and modify it.
How did this happen? Are Chinese engineers just built different, like their LLMs? Ironically, the US could have created its own competition. Most LLMs in the world run on high-end GPU chips manufactured by Nvidia. Recently, there have been restrictions on the export of high-end H100 chips to China, and Chinese engineers have been forced to work with cheaper H800 chips. But because they had less powerful chips to work with, they were forced to figure out clever solutions and efficient architectures instead of just spending more money on powerful chips. So Deepseek turned out better.
Now, there’s panic in the US market, with Nvidia’s stock dropping by one-sixth of its value in a single day. Many other tech companies are also crashing.
These fears of “Chinese dominance” are mostly exaggerated – since these models have been open-sourced, it will most probably accelerate AI development globally and demand for chips will increase long-term. But it’s a wake-up call for US tech companies to up their game. If the entire world starts using Deepseek due to its cost and performance advantage, China could influence AI Policy down the line and control the models.
Sources and further reading: Fireship on Deepseek (3 min video), BBC on the AI war and Perplexity CEO interview (40 minute video), why the panic is unjustified (tweet). Credits to Paul Millerd for coining the term Deepseekalypse.
2. The IKEA marketing strategy
IKEA superstores are located on the outskirts of cities. IKEA can choose to build its stores anywhere – then why does it opt for such unfavorable locations?
When people commit to something difficult, they fall prey to the sunk cost fallacy. Because IKEA stores are so out of the way, customers have to plan their entire day around a trip to these stores, and it becomes a day-long trip. After visiting these stores, they are likely spend a lot in a single trip because these trips happen so infrequently, and they end up buying more than they need.
On top of this, IKEA Dubai introduced a scheme called “Buy with your time.”1 People could show their Google Maps Timeline during checkout and pay with their travel time to IKEA. If you had just 5 minutes on your account, you could buy a hot dog, but with 9 hours, you could buy even a table (the time would accumulate on your account). By doing this, IKEA turned customers who lived far away into loyal patrons.
Here’s a fun fact: IKEA is an acronym. It stands for Ingvar Kamprad, Elmtaryd Agunnaryd. Ingvar Kamprad started his business in the village of Agunnaryd which had just 220 inhabitants, and that was his entire address.
3. What makes a fight scene good?
In this scene from “The Hobbit: The Battle of the five armies,” the elf warrior Legolas hangs upside-down from a flying bat’s claws and kills about a hundred enemies in less than ten seconds. Then he spends three minutes fighting just one powerful enemy.
Why does the final fight always take longer? Can any warrior be 1800 times better than the average warrior?
No. As a rule, skills have very little to do with the outcome of battles, and the side with advantage in numbers generally tends to win. If a smaller army has an advantage, like the 300 spartans in the Battle of Thermopylae, it’s because of a geographical advantage or inside knowledge, not because they are exceptional warriors.
But movie scenes are designed like this for drama because of “the conservation of ninjutsu.” It’s more dramatic and satisfying to see well-choreographed one-on-one combat compared to random violence, so the final boss is always more powerful than a group of 100 people. Game designers use this principle in boss fights as well.2
4. Late bloomers
Raymond Chandler started working at an oil company in 1922. He became a highly paid director by 1931. But he wasn’t the ideal employee – alcoholism, absenteeism, and threatened suicides led to him getting fired in 1932, during the Great Depression. He badly needed a job and thought of giving writing a shot. By studying mystery stories in Black Mask magazine, he wrote his first story when he was 45 years old. He published his first novel The Big Sleep at the age of 49.
Chandler is considered one of the greatest detective fiction writers and he wrote 7 novels, 22 short stories, and 7 screenplays before his death. His age and maturity inform the cynicism and bleak imagery of his works. If he had started earlier, he would have probably been a very different writer.
Chandler isn’t the only one with a late start who made it big:
Ang Lee, the director of Life of Pi, directed his first film at the age of 37.
Ray Kroc was a 59 year old milkshake mixer salesman when he franchised McDonald’s (which he eventually seized from the founders)
Edward Jenner invented the smallpox vaccine at 46.
Samuel Johnson is considered one of the greatest essayists. But he had no reputation before the age of 40. All of his popular creative work was done in his late 40s and 50s.3
A lifetime of accumulated knowledge and experience could lead to the depth necessary to create great work. Packy McCormick recently wrote an article called “Burn the Playbooks” which warned that the playbook-ization of success was encouraging young people to create shallow imitations of existing achievements instead of taking the bold risks needed to create something original.
I wonder if a culture that pushes people to achieve a lot in their prime leads to decline in deep thinking.
5. The paintings of Jean Mallard
For a change, here are some paintings by an artist I follow on Instagram. I really like his surreal worlds and soft colors:
You can check out more of his work here.
Read another post here about the millionaire who is trying to live forever. Or read about my experience cloning a billion dollar company. But before that, please like and share this post to support my writing. Thanks!
Thanks to my friend Sooraj for sharing this.
Faceless goons are also more disposable than characters with names. There’s a lot of art to this stuff, and if you’re interested in storytelling, it’s worth exploring.
Writers especially seem to benefit with age if they persist – Bukowski worked as a postman till 49 before he published his first novel “Post Office”. Raymond Carver published his first collection at the age of 38.
You're welcome. Keep writing 😎