Palantir crashed 25% this week. Nvidia 9%. While the AI bubble may be popping, 75 years of data suggest it’s not time to panic-sell just yet.

Each week, a new company claims they are cracking AGI. This week it’s Google’s Genie 3 turn. Meanwhile, Google is also quietly leapfrogging other tech companies on the AI race with a simple feature “Google Cue”.

As LLM performance plateaus, the real battle is moving up the stack to the software layer. The real question is how do we leverage LLMs to revolutionize the engrained daily habits of billions of users.

Let’s start with a 30,000 feet view on the market… 

The week started with Bubble fears. Meta freezes AI hiring and reorganized its AI department, Sam Altman piling on admitting we’re in a bubble, and MIT saying that 95% of AI pilots are failing. Most Tech/AI Stocks and other high beta assets like Bitcoin fell 10-25% from last week’s highs.

Then J-Pow took the mic from Jackson Hole.

The head of the Fed Jerome Powell was more dovish than expected. While not promising rate cuts, his remarks sent a clear signal to the market that the Fed is preparing to adjust in response to a weakening economy. Futures market are estimating the probability of a rate cut at 75%.

Brief: Seasonality is one of those things that are difficult to explain, but works. Using 75 years of S&P 500 data, I plotted a chart of what the average index price action looks like in the post-election calendar year. The pattern that emerges, and that keep happening in those years is as follow: Early high in the year, followed by weakness in Q1. Then a rebound into the summer high, usually July, before going into a correction early fall, and ending the year with a rally.

The Chart:

So far this year, SP500 Movement is a textbook case of post-election year seasonality.

I overlaid AIQ to see if there are any differences. Here is what the data says:

→ SP500 Spring rally: AIQ vs. SP500: 1.6x amplification
→ Maximum drawdown: AIQ vs. SP500 -10%: 2.6x amplification
→ YTD performance: AIQ +16.7% vs SPY +10.1%: 60% overperformance

AIQ also delivered a textbook post-election year pattern, just amplified.

The Undeniable Pattern:

So far
✅ New Year High in Feb
✅ Spring Low
✅ Summer High

Next?
❓ Fall Low After Correction
❓ Year-End Rally

Why it's important:

As always, do your research. This is not investment advice. Last time I checked, I don’t have a crystal ball.

If you don’t mind me continuing my future predictions, here is what AI would look like in a couple of years…

Everyone's obsessing over who's going to achieve AGI first. In reality, the real AI battle isn't about building the smartest general intelligence. It's about context.

I want to coin a term for what's really happening in AI: Context Acquisition: gathering and understanding personal user data to create an AI that just 'gets it’.

Google just proved this with their Pixel 10 launch.

While everyone debates ChatGPT Codex vs Claude Code, Google quietly built Magic Cue: an AI that knows when you're calling an airline and automatically surfaces your flight details. That's not AGI. That's context.

Current LLMs are already powerful enough to be transformative personal assistants. The limitation isn't their reasoning ability, it's their lack of context about YOU:

What AI companies are really fighting for:

Device Integration: Apple and Google embed AI directly into phones, watches, and home devices. Every interaction feeds the context engine. Your iPhone knows when you wake up, what apps you use, who you text. That's goldmine data for personal AI.

App Omnipresence: Companies like Perplexity are building browser extensions (Comet) that watch everything you do online. Every search, every article, every purchase decision gets fed to your AI assistant.

API Background Learning: AI systems that connect to your existing tools (Gmail, Slack, Notion, calendars) and learn your patterns without you actively prompting/training them.

Apple has superior on-device processing and a very sticky ecosystem, but Google has something more valuable: native integration with products billions use daily.

Magic Cue works because Google owns your digital life:

The transformative AI experience you want isn't waiting for superintelligence or AGI. It's slowly happening with seamless context acquisition. The company that cracks deep personal context + on-device privacy + convenient user interaction, wins the AI race.

Context Acquisition is the key to an AI that just "gets it."

As always, thank you for your readership. Let’s close with our weekly tradition: a number that matters:

That’s how much information you can share when typing. Compare that to 10 Million Bits when seeing, and that’s for humans. Machines can absorb even faster. In fact, eyes process information 49,000 times faster than even the world's fastest fingers can type.

If we stand any chance to provide AI with context as fast as possible, we will need AI Assistants to be looking at our screens and scanning context in the background.

Data Privacy? that’s why Google and Apple are betting on local models that can run on your device, without the need for sending your data to the cloud.

The Trade: Long Google. Long Apple. Long Context Acquisition. Short typing walls of text to prompt a chatbot.

Your move:

This week's poll: What did you think of this week’s edition?

Got intel or feedback? Hit reply. I protect sources :) 

Know someone who may be interested? Forward this.

See you next Sunday.

Playing the odds, Johan

© Expected Value 2025 - Not investment advice. Do your own research, Seriously.