Valuations, Argon, and Bitcoin: what Asia's AI sell-off, Google's cheaper frontier AI, and Bitcoin dipping to 80K mean for people starting from zero

Markets are asking one question this week: is AI actually worth what everyone is paying for it?

Asian markets sold off hard on fears that artificial intelligence stocks are overvalued. One market report put it plainly: the selling came from concerns over stretched AI valuations plus higher than expected US inflation. At the same time, India’s Sensex jumped more than 1500 points to 74315 as money rotated out of expensive tech and into cheaper markets.

What is happening is a repricing. Investors are no longer buying every company with the word AI in its name. They want proof. Earnings. Revenue. Real customers. The days of hype money are ending, and the days of proof money are starting.

For a person starting from zero, this is good news. Hype markets reward insiders. Proof markets reward builders. When the crowd stops throwing money at buzzwords, the people who can actually build things get their opening.

What would you rather do: chase a hype wave, or build something real while everyone else panics?

Bitcoin feels the same cold wind

Bitcoin sits at 80501 dollars this morning, down about 2.6 percent in the last 24 hours. Ethereum fell harder, down about 5.4 percent to 2346 dollars. The total crypto market dropped about 5.5 percent to 2.71 trillion dollars, and Bitcoin now makes up almost 60 percent of the whole market.

Read that last number again. When Bitcoin’s share of the market rises during a sell-off, it means people are dumping the risky small coins and running to Bitcoin. Fear makes people pick the safest boat.

Here is the builder lesson in those numbers. Every market has seasons. Money runs toward hype, then it runs away from hype, then it comes back to things that work. Bitcoin has survived every one of those cycles. Your job is not to time the cycle. Your job is to keep building through it. The people who kept building through the last crash are the ones with products and skills right now.

When the tide goes out, you find out who was swimming naked and who was building a boat.

Do you want to be the person who reads the fear, or the person who builds through it?

Google’s new model is cheaper, longer thinking, and honest about its limits

On October 13, Google opened its newest frontier model, Gemini 4 Argon, to paying subscribers. This is Google’s most capable model, built for long, complex work like software engineering, financial research, and cybersecurity defense.

Here is the technology in plain terms. Older Gemini models could write at most about 64000 tokens of output at a time. Argon works with a shared pool of about 1 million tokens split between your instructions, its internal thinking, and the answer it writes back. That means it can chew on a much bigger problem in one sitting: a whole codebase, a long document, a full business plan.

It also has three thinking levels: LOW, MEDIUM, and HIGH. More thinking means better answers on hard problems, but it uses up more of the token pool. Google’s own tests put Argon at the top on software engineering (77.9 percent on DeepSWE) and finance tasks. Independent testing placed it roughly even with OpenAI’s GPT-6 Astra, behind Anthropic’s Claude models, with one standout: a 15 percent hallucination rate, the lowest of any top model. And here is the honest tradeoff: it invents less, but its answer accuracy was 50 percent against Astra’s 63 percent. No hype. Every tool has strengths and limits.

Here is why this matters for you. The price of serious AI keeps falling. Google is charging 2 dollars per million input tokens and 10 dollars per million output tokens through the end of 2026, and the price doubles in 2027. One independent analysis put the cost per task at about 1 dollar and 99 cents, roughly 60 percent of OpenAI’s flagship cost.

And here is the part nobody is saying out loud. Google says Argon helped its Chrome team finish a major refactor of about 6 million lines of legacy code in weeks instead of months. The same class of model that reworks a browser is available to you for 19 dollars and 99 cents a month on Google AI Pro, with a code building mode called Canvas.

One concrete thing to do today:

• Open the Gemini app and paste in a real project, your notes, your business idea, a messy spreadsheet.

• Ask it to do one complete piece of work in a single session: a full project plan, a documented rewrite of a script, a financial model with the math shown.

• Do not ask it 20 small questions. Give it one big job and watch how far one sustained session goes.

That is the skill the market is about to pay for: knowing how to aim a long thinking model at a real problem.

So while investors panic about AI valuations, the actual cost of building with AI keeps dropping. The hype is getting cheaper and the tools are getting better. That gap is where beginners win.

The uncomfortable truth

The AI valuation scare and the Bitcoin dip are the same story. When money was cheap and hype was free, everyone looked like a genius. Now the tide is going out. The people left standing will be the ones who built real skills and real products while the tourists went home.

So here is the push. Pick one thing this week. One model, one project, one skill. Give it 60 minutes a day. In 90 days you will know more about building with AI than 95 percent of the people arguing about AI stock prices on the internet. The sell-off is not your problem. Being unprepared is.

Join my free Skool community: https://www.skool.com/henry-jenkins-5224

My website:

https://henryljenkins.com

LightofLife.global

Twitter @Nevaquit

YouTube @HenryJenkins3

Subscribe: https://jenkinsh.substack.com/subscribe

Thanks for reading Henry’s Substack! Subscribe for free to receive new posts and support my work.

Henry Jenkins