Airtable makes $480m a year, but it sold itself for less than what it could make in 3 years. If you’re an agency or solopreneur who runs Meta Ads or writes weekly sales emails, here’s how not to go out of business and grow it instead before 2030.
B2B SaaS used to be the boring but predictable favourite for Venture Capital. Selling a beautifully designed NextJS website, social media marketing or AI automation used to be a real agency business.
But in 2026, you just can't run away from the fact that an intelligent model like Opus, Fable or Kimi K3 really can do it all.
- Ask them to make you an $8,000 to $10,000 website and they can.
- Ask them to create on-brand Instagram posts with perfect typography and they can.
- Ask them to write custom code to replace those 3–5 automations you pay Zapier $250+/year for and they can.
And you can disagree or say that you're better. But the facts remain: good AI models are bloody good at predictable information jobs.
And because they have been trained on ungodly amounts of code, their ability to automate, design or code makes what you do (if you still do) a commodity.
And there is no escaping that.
Which is exactly why Airtable realised that it is a commodity, waiting to be priced as one. And it's exactly why if you come from the world of digital products and services, you're in a crisis.
There are really just two options left: either you do a business or you equip a business with what they will inevitably get to.
Because let's be honest, the guy/gal you want to sell a $10,000 website to could probably do it themselves with Claude Code and a few design skills.
And before you tell me Airtable is one company, look at the whole shelf.
In 2021, a public software company sold for 18 to 19 times its revenue. By 2023 that was 6.7. Then, in the first three months of 2026, it fell again, from 6 to 3.4, the lowest exit multiple in 10 years for Software as a Service.
And it wasn’t the interest rates or the Fed’s decisions.
What moved was that people started opening Claude Code and rebuilding the tools they were paying for.
The investment banks who price these companies for a living wrote it down in plain English: the trigger was AI coding agents and vibe coding.
Now here's the part nobody who sells a digital product or service wants to hear. The thing doing the repricing isn't a chatbot.
It's a black window with a blinking cursor. The terminal. The oldest way of using a computer there is, and in 2026 it's the fastest growing one.
OpenAI's Codex went from three million people a week to five million in eight weeks. Claude Code gets installed 69 million times a month.
And one in five of those people isn't a developer. They're writers, operators, a mum organising voice notes from a stroller walk. People who never touched a terminal in their lives, and now won't put it down.
Heck, I believe in freedom for all. I very much look forward to going back to being human.
If I can operate my life from an iPad Mini with SSH to use my Mac’s terminal, I will take that life over 8 hours of being hunched up in front of a 16 inch+ display anyway.
I call it living 2030 and that's exactly where I want to take you.
Now what makes all of this work is somewhere between Tiago Forte and Tony Stark. Somewhere between ‘Building a Second Brain’ and ‘Jarvis’. And I like to call this thing a Second Brain Operating System.
Now how’s that different?
Well, a Second Brain ends at ‘I have notes, I want them to be related by meaning and for an AI to be able to search + converse with me’.
That’s just one layer of a Second Brain Operating System. And it’s the layer that we start with so that you can become friends with the terminal well before 2030.
More on that in a moment, because clarity first.
I know the word AI Agent gives you nausea, but I have to bring it up. A Second Brain OS is a home for AI Agents that think, act & work like you. Now it’s time for me to bring all the fancy shmancy buzz words and break down each one:
Think like you: ‘Retrieval Augmented Generation’.
The minute you create or update a note, it gets related by meaning to its siblings.
And that’s regardless of whether the sibling notes are from yesterday or from 2 years ago outside the gym. RAG when done right lets an AI model search your knowledge by meaning and match the train of thought going through your head at that moment.
Act like you: ‘APIs exposed as MCP Tools’.
From the button that sends an email, to the one for downloading an image from Canva and to the one that adds a row to Airtable or a Google Sheet, it’s all powered by an API.
- An API for sending an email
- An API for exporting an image
- An API for adding a row to a table
If it can be behind a UI, it can be behind an AI model.
The Model Context Protocol simply ‘standardises’ how an API appears to any AI model, from the big boys like Fable & Kimi K3 to the little guys like Qwen 3.
And it’s as simple as ‘Give me the data you want to add to a sheet and I will confirm if it went through’ or ‘Give me the to, from, subject and body for an Email and I will confirm if it reached’.
Work like you: ‘Skills, Prompts, SOPs, Best Practices or Prompt Chains’.
This is the part that really annoys me. Because it’s caused a mess. And I want to be the first person to tell you that they all mean exactly the same thing.
God knows why, who or when someone decides to rebrand it and call it a new thing. A skill is = 100% a prompt. The only difference (and you tell me if it was worth all the confusion) is that a skill is a “prompt written and stored in advance”. So that someone can use a slash command like ‘/poppins-font-insta-carousel’ to load a skill that (you guessed it) tells the AI model how to make an Instagram carousel with Poppins font.
Phew. 3 mouthfuls.
But much better than the word AI Agent because hey, now you know ‘agency’ happens when something can ‘think, act & work’ like you.
Disclaimer: AI can’t exactly think, so think of ‘think’ as a way to compensate for a lack of a better word. So for now, you can continue to think that AI can think if it makes you think less about how it thinks (if it thinks at all).
Don’t think so much, thinking burns tokens. Just keep reading.
P.S. For F***’s sake, stop using AI on thinking high. Lower is better. The less it thinks, the less it gets confused about what to do. And the less you ‘burn’ in terms of tokens and environmental footprint.
P.P.S. If you’re down to get started with a Second Brain Operating System, here’s how
