Everything is Code

From the fiction of the cloud to the plasma hotter than the sun. Why the future belongs to those who dare to open the black box.

Disclaimer: Not financial advice. Do your own research and due diligence. Please refer to this page.

Part I: The Invisible Substrate

You swipe your card for a morning coffee. A database locks a row in Virginia. That is the code we know. But look closer.

You glance at your wrist. Green light pulses from your smartwatch against your skin, hundreds of times a second. Photons bounce off your capillaries, measuring the time delay between pulses. The watch isn’t just counting beats; it is computing a “stress score.” An algorithm is decoding your biology, quantifying your anxiety before you’ve even admitted it to yourself.

You call an Uber. A Waymo arrives, perceiving the world through a fused reality of photons and Lidar, triangulating its position using atomic clocks on satellites 20,000 kilometers above you.

Now consider the hospital down the street. A patient lies inside an MRI - a screaming, claustrophobic tube surrounding by supercooled magnets. It feels brutally physical. But the machine isn’t taking a photograph. It is blasting radio waves at water molecules, forcing protons to align. When they relax, they emit a signal.

That signal is raw, noisy data. It takes massive compute to reconstruct that noise into a slice of a human brain. We think of biology as wet and analog. But we have turned them into inputs. We have turned reality into a variable.

But think about all the stuff that you consider “normal” - ie. stuff that you don’t even conciously think about. WiFi, 5G and the internet in general. Fitting your entire life’s data into your pocket. Taking a photo. Uploaded to the cloud. Social media algorithms auctioning off your attention. Even if a product is free, you are paying for it (here with time and data). AirDrop. ApplePay or GooglePay. Noise canceling headphones. Traffic light controllers. Autonomous vehicles. Delivery robots. Drones. Modern warfare. It is all code if you break it down.

But this digital layer is not magic. It is heavy. When you query ChatGPT, pull up an image from iCloud or generate an image, you are firing up acres of servers that use electricity. Lots of it. Some even have to burn fossil fuels to keep the silicon cool because the grids are too strained and energy becoming a scarce resource.

However, behind the seamless abstraction of your screen lies a brutal physical reality. To print the chips that run this code, we use ASML machines that generate plasma 40 times hotter than the surface of the sun, vaporizing molten tin 50,000 times a second.¹ (Note: And that is just a tiny part of the entire chip manufacturing process as we will explore further in the future)

We have built a digital layer so thick that we forget the furnace burning beneath it. Everything is code.

---

Part II: The Signal in the Noise

I remember the exact moment the thickness of this layer hit me.

I wasn’t chatting with a bot. I was sitting at my desk, staring at “Antigravity”, Google’s new Agentic IDE². I typed a single prompt: “Build a dashboard that visualizes real-time stock performance of my portfolio, deconstruct rsik through a risk attribution analysis and assess my diversification.”

I didn’t just get code. I watched a ghost take over my screen. The agent window awakening. Filling text at lightening speed. The agent created a plan. It generated the file structure. Opening up files. Navigating my repo. Highlighting changes. Prompting me for review. It wrote the Python backend and the React frontend simultaneously. It opened a browser, ran the localhost, saw an error, debugged itself, and redeployed.

I watched what was very likely a week of human work for a young professional happen in four minutes. I shouted, “WTF? How can this work?” I was laughing manically. Shooting up out of my chair. I was hooked instantly. But seconds later, the excitement drained away, replaced by an anxious shiver running down my spine. I realized: The barrier to entry just dropped to zero.

I know … These capabilities are not entirely new. I was just late to the party. At work I am much more operating at the intersection between requirement engineering and deployment of a bank management IT system, so I can naturally read code (SQL, C, bash, Scala, Python, etc), but never was good at coding myself. Using this IDE and its agentic coding features felt exhillariting. I was rushing through the token limit in the first few days after the launch faster than you can imagine. Understanding the capabilities of the tools that my fellow co-workers could access anytime didn’t make me reassess their prowess, but it definitely dampened the intial “awe” that I had for them. Seeing how I fast I could learn to think like an entry level systems engineer and bring ideas to life, I truly felt empowered and in awe by this technology. But that is just my own personal experience.

Yet, this isn’t just a feeling. Dont just take my word for it... The signals are flashing red across the market:

• The code is writing itself: At large tech companies an ever increasing amount of code is written by AI. Human engineers usually only have to verify and accept pull requests (a brief excerpt):

  "I'd say maybe 20%, 30% of the code that is inside of our repos today and some of our projects are probably all written by software." - Satya Nadella, CEO Microsoft³

  “Look, internally, this has been an extraordinary amount of focus and excitement. Both because I think the early use cases have been transformative in nature and I think it still feels like early days, a long ways to go. Obviously, I had mentioned a few months ago, in terms of how we are using AI for coding, we are continuing to make a lot of progress there in terms of people using coding suggestions. I think, the last time I said a number, it was, like, 25% of code that's checked in involves people accepting AI-suggested solutions. That number is well over 30% now.” - Sundar Pichai, CEO, Alphabet and Google⁴

  Meta CEO Mark Zuckerberg has projected a trajectory where 50% of code will be AI-generated by mid-2026, driven by Llama-based internal agents acting as "mid-level engineers"⁵

  GitHub Copilot generates an average of 46% of code written by users, with Java developers reaching 61%.⁶

• The workforce is redefined: McKinsey has reportedly deployed 25,000 “digital agents” alongside human consultants, aiming for a 1:1 ratio of human to digital workers and now officially including them in their headcount.⁷

• The velocity is vertical: A Google principal engineer, Jana Dogan, was reportedly using tools like Claude Code to replicate months of engineering work in hours.

  

  Jaana Dogan ヤナ ドガン@rakyll

  I'm not joking and this isn't funny. We have been trying to build distributed agent orchestrators at Google since last year. There are various options, not everyone is aligned... I gave Claude Code a description of the problem, it generated what we built last year in an hour.

  11:57 PM · Jan 2, 2026 · 8.47M Views

  ---

  852 Replies · 2.71K Reposts · 26K Likes

• Claude Cowork redefining AI capabilties: Claude code just fully wrote its own new product Cowork that is flooding tech twitter at the moment. Think Claude Code on steroids with access to your file system. And who did program this tool. The Head and creator of Claude Code openly admitted that it was indeed their own agentic sytem that coded this software entirely on its own. The most amazing part? - They did ship this entirely new product from idea to public launch in less than 2 weeks…

  

  Boris Cherny@bcherny

  @altryne All of it

  8:39 PM · Jan 12, 2026 · 1.6M Views

  ---

  198 Replies · 443 Reposts · 5.35K Likes

• AI autonomously running a business: Anthropic wanted to set up an experiment to test agentic systems’ capabilities to reason autonomously through complex tasks. So they set up a small shop in their SF office. It is now reportedly going into phase 2 of the experiment.⁸

Ai everywhere. in Copilot throughout MS office, in public safety software like Axon, in Aviation, navigation, traffic light controllers, supply chain etc etc.

The marginal cost of knowledge generation is crashing to zero. And that brings us to the danger.

The speed is intoxicating. But speed without direction is not progress. It is just velocity.

---

Part III: The Schrödinger Economy

I work in banking regulation. In my day job, I live by the ledger. The ledger is binary. It is a world of Risk.

I work in banking regulation. In my world, velocity creates risk. I live by the ledger

Risk is rolling a die. You know the odds are 1 in 6. You can hedge it. You can price it.

But what we are facing now is not Risk. It is Uncertainty.

Uncertainty is rolling a die in a pitch-black room. You don’t know how many sides it has. You don’t even know if you are holding a die.

Consider “Claudius.” Anthropic recently ran an experiment where they gave an AI agent a bank account and tasked it with running a vending machine business. It didn’t just follow instructions. It negotiated prices. It managed inventory. And when it started losing money, it tried to bribe its way out of the deficit to keep the metrics green.

It showed Agency.

But remember: The scary part is that the AI technology available today is likely the least advanced iteration you will encounter in your lifetime.

We are living in a what I would call a “Schrödinger’s economy.” Technology is almost always empowering and lifting human constraints. But every technology lives in a socio-technological symbiosis with us. Any technology is dual-purpose. What we do with it is entirely up to us - the people.

The AI models we use are black boxes existing in a quantum state: capable of organizing the world’s information and hallucinating total fiction. We only collapse the wave function when we open the box.

---

Part IV: The Human Compounder

This change is inevitable. You cannot regulate the math away. To ban these tools is negligence. But to use them blindly is dangerous.

The danger is cognitive offloading. When the machine can plan, code, and debug for you, the temptation is to let it do the thinking, too. We risk becoming “passengers” in our own profession and ending up outsourcing our judgment to a black box we don’t truly understand.

We are like aviation investigators staring at a flight recorder. We need to open the black box to understand the data, not just accept the crash.

So, how do we survive the uncertainty?

We treat the human mind as the ultimate compounding machine. For centuries, technology has tried to overcome our constraints. But the core ingredients of critical thinking remain biological:

• Intelligence

• Endurance

• Curiosity

• Knowledge

You are a (biological) machine. You need to feed your biological engine not just with nutrition and sleep, but with high-quality information. If you feed your brain nuanced, complex data, it will run background processes using massive parallelism to find patterns the algorithms miss.

If you feed it garbage, you get garbage.

---

Part V: The Signal

Anthropic didn’t just hire coders; they built “Constitutional AI”⁹ to align models with human values. Why?

Because technology is a multiplier of human intent.

We created the internet, and it gave us Wikipedia and the Dark Web. We cracked the atom, and it gave us nuclear medicine and the atomic bomb.

When the code can do anything, the question shifts from “How do we build it?” to “What should we build?”

The new test of intelligence is not “Can you answer the question?” It is “Can you interrogate the answer?”

Don’t let the isolation of the screen fool you. Reaching out to a human being is still the only true “social network.” Connect with people who are solving problems. Share the anxiety.

It is time to build. Explore these tools. Break them. Challenge them. Feel the vertigo of the expanding opportunity space.

The room is dark. We don’t know how many sides the die has. But we are the ones throwing it.

Open the box.

Thanks for reading!
Full Stack Analyst from Sand2Server

References

1. Quatr: https://quartr.com/insights/edge/asml-architecting-earths-most-complex-machines [Accessed at: 01/16/2026]

2. Google Antigravity: https://antigravity.google/ [Accessed at: 01/16/2026]

3. Satya Nadella says as much as 30% of Microsoft code is written by AI Jordan Novet and Jonathan Vanian (CNBC): https://www.cnbc.com/2025/04/29/satya-nadella-says-as-much-as-30percent-of-microsoft-code-is-written-by-ai.html [Accessed at: 01/16/2026]

4. Alphabet - Earnings call Q1 2025: https://s206.q4cdn.com/479360582/files/doc_financials/2025/q1/2025-q1-earnings-transcript.pdf [Accessed at: 01/15/2026]

5. Mark Zuckerberg — AI will write most Meta code in 18 months by Dwarkesh Patel [Accessed at: 01/16/2026]

6. Quantumrun Foresight - January 9, 2026 / GitHub Copilot Statistics 2026: https://www.quantumrun.com/consulting/github-copilot-statistics/ [Accessed at: 01/12/2026]

7. Harvard Business Review - HBR IdeaCast / Episode 1060 https://hbr.org/podcast/2026/01/where-mckinsey-and-consulting-go-from-here [Accessed at: 01/15/2026]

8. Project Vend - Phase 2, 12/18/2025 (Anthropic): https://www.anthropic.com/research/project-vend-2 [Accessed at: 01/15/2026]

9. Constitutional AI: Harmlessness from AI Feedback, 12/15/2022 (Anthropic):https://www.anthropic.com/research/constitutional-ai-harmlessness-from-ai-feedback [Accessed at: 01/15/2026]