The $11 Billion SaaS Reality Check

The $11 Billion SaaS Reality Check

The Throughline Brief · August 9th. 5 min 49 sec.

Cold open

AirTable just took an 88 percent haircut on its valuation. An 11 billion dollar giant dropped down to 1.2 billion in a single deal. It represents a blaring warning siren for the entire software market. Throwing a generative AI wrapper on a decade old code base is a losing bet.

Intro

This is The Throughline Brief for August 9th. I went through every big AI and tech podcast this week so you do not have to. Here is what actually mattered. I am West. Let us get right into the biggest stories of the week.

The 11 Billion Dollar Haircut

Let us start with the 11 billion dollar software reality check. AirTable was the absolute darling of the no code movement. Now, they just took an 88 percent haircut on their valuation. They dropped from a peak of 11 billion dollars down to just over 1.2 billion. Think about that destruction of value. On Twenty V C, Nikash Arora laid out the brutal math behind the deal. The company was doing 485 million dollars in revenue. They were growing at 20 percent year over year. And they ultimately got priced at a 1.285 billion dollar acquisition level. The market is completely shifting its valuation models right now. Throwing generative features at a decade old code base is simply a losing bet against greenfield startups. You are fighting lean companies built natively on AI. And David Sacks hit this exact point over on the All In Podcast. He pointed out that artificial intelligence is actively destroying the perceived value of these older platforms. Specifically, he targeted the no code sector. Sacks argued that no code has to be the most impacted area of software as a service right now. Why? Because Claude is the ultimate no code tool. A frontier model can just write the software you need. You simply speak to an AI instead of using a visual drag and drop interface. That is a massive model shift. Every single software company with a high valuation from 2021 is waking up to a very cold reality this week.

Self-Correcting Models

We also saw a massive shift in how AI handles code. Models are now checking and correcting their own work in real time. On Latent Space, Ali Taha shared something truly wild about Base ten and their G L M five point two model. It acts as a completely self optimizing loop in production environments. It actually maps out exactly where it spends compute energy. It spots its own slow parts. Then it does a forward pass on itself, gets the profile trace, analyzes it, and finds the inefficiencies to correct automatically. No human in the loop. The system just makes itself faster. This autonomy completely changes how we instruct software going forward. We are moving heavily to high level delegation. Over on How I AI, Nick Bowman demonstrated this by letting Open AI Codex handle complex video editing and redaction. He gives it a single high level command. He tells the model to go through the clips first and pull the transcript. Then he just trusts the agent completely. The agent tracks moving elements on the screen. It blurs sensitive data. It verifies its own accuracy. Then it delivers the final cut. You are just giving a goal and walking away. The model handles all the underlying complexity. The era of writing step by step instructions is rapidly closing before our eyes.

Nikash Arora

Nikash Arora had a brilliant moment on Twenty V C while talking about the future of software margins. He was breaking down how AI features actually impact the bottom line for big tech companies. He warned that average intelligence is going to be free in the long term, and the average intelligence will keep getting better, while exceptional intelligence will be paid for. Here is why that matters to you right now. Charging a per seat premium for basic AI writing tools or summarization features is a race to zero margins. It is completely unsustainable. You have to provide exceptional intelligence to command a premium today.

Dino Mavrukas

Over on All In, Dino Mavrukas laid out a staggering industrial gap while discussing global manufacturing power. He dropped a statistic about ship building that puts our actual physical infrastructure into sharp perspective for the audience. Dino stated that the United States can build 100 thousand gross tons of ships every year, while the Chinese can build 23 million gross tons, meaning they can outbuild the US 230 to 1. That is a brutal physical reality check on the state of American industrial might. We talk constantly about leading the world in artificial intelligence and software capability, but our physical production capacity is a totally different story.

Shaan Puri

On My First Million, Shaan Puri was dismantling the standard career advice of following your passion. He was explaining exactly why asking young founders what they want to do is a fundamentally flawed approach to building a company. Shaan said it is like asking his four year old what his favorite food is, and he has only ever eaten cheese pizza and mac and cheese. The advice completely fails when your exposure to potential careers is so incredibly limited. You cannot follow a passion for an industry you do not even know exists. You have to actively expand your inputs before you can decide on an output.

The bottom line

The market is absolutely done paying venture multiples for legacy software with a thin AI wrapper attached. We are seeing a complete demand for ground up reinvention and immediate profitability across the board. You have to build native AI tools or show real cash flow right now. There is literally no middle ground left to hide in. That is your Throughline. See you next Sunday.