The Ghost in the Deal: Why Venture Capital Must Become Software
- Aki Kakko
- 23 hours ago
- 5 min read
Updated: 8 hours ago
In The Map is Not the Territory, we established that the Power Law is a system property, not a selection filter. In Why the Smartest Money Looks the Stupidest, we identified the psychological barrier—Fear of Looking Stupid (FOLS)—that prevents investors from capturing that value.
Now, we must address the Execution.
If the math suggests we should be "farming" for outliers rather than "sniping" for certainty, and if the psychology requires us to ignore social consensus, then the operational model of Venture Capital must change fundamentally.
The current user experience for entrepreneurs is a relic of a bygone era.
It is a slow, manual, bias-heavy process where founders waste 2–6 months courting distracted partners, only to receive vague rejections based on "gut feel." It is the definition of terrible UX. It is an irony of the highest order—an oxymoron, even—that VCs aggressively fund companies to automate every sector of the economy with AI, yet steadfastly refuse to automate themselves. They believe their business is "art," while everyone else’s is "data." They are wrong. But they are also right about one thing:
You cannot simply automate the current process.
To build the "Self-Service VC" of 2026, we have to understand why the current AI models fail, and look to a concept from cybernetics and corporate theory: The Ghost in the Shell.
The Category Error: Why You Can’t Mortgage a Startup
The most common pitch for AI in VC is the "Mortgage Model." The vision is seductive: an entrepreneur uploads their data, an AI analyzes it, and a decision is rendered in hours. This fails because it relies on a category error. Mortgage underwriting works because it analyzes bounded risk on tangible assets using historical data. The distribution is Normal (Bell Curve).
Startups are the opposite. They represent unbounded risk on intangible potential. The distribution is Pareto (Power Law).
If you train a standard AI model on historical startup data, it will regress to the mean. It will identify excellent Small and Medium Businesses (SMBs)—a bakery, a consulting firm, a local agency. But it will reject the next Airbnb or Uber. Why? Because at the pre-seed stage, outliers look "weird." They lack the pattern-matching characteristics of a safe bet.
The "Ghost in the Shell" Problem
To understand why we can’t just "feed the data to the AI," we have to look at what a startup actually is at the pre-seed stage. As outlined in recent analysis regarding the "Ghost in the Shell" theory of the firm, we are witnessing a decoupling of the legal entity from the value creation.
At the Series B stage (where VC Quant-funds like SignalRank operate), the "Ghost" has manifested into the "Shell." The obsession has turned into ARR. The insight has turned into Churn data. You can measure the Shell.
At the Pre-Seed stage, the Shell is empty.
When you ask an LLM to analyze a pre-seed startup based on a pitch deck (the Shell), you are asking it to analyze a hollow container. I recently tested this with tools like Rule30. The results were not analysis; they were hallucinations. The AI, finding no traction data in the "Shell," fabricated a narrative to fill the void.
The Solution: Digitizing the Ghost
To make the "Mortgage Model" work for VC, we must stop analyzing the Shell (the business plan) and start analyzing the Ghost (the founder’s behavior and potential).
This is why incumbents like Andreessen Horowitz and Sequoia haven't automated yet. They rely on "coffee meetings" to sense the Ghost. They use human intuition to bridge the data gap.
To disrupt them, we don't need better financial models. We need sensors.
Venture Operating System: A Data-Harvesting Protocol
This is where our strategy diverges from the "spray and pray" approach. We are not just building an automated check-writer; we are building a Venture Operating System. We propose a suite of utility tools—SDKs, Chrome Extensions, financial dashboards that are integrated to partner APIs and co-pilots—that founders use to build their companies. These are not just "value-add services"; they are diagnostic tools.
By observing the founder through the VentureOS, we digitize the Ghost. We turn "hustle" into a data point. We turn "resilience" into a vector.
The Algorithm of 2026: Capital as a Service
Once we have digitized the Ghost, the "Mortgage Model" becomes viable. The investment process of the future will not be a series of pitch meetings. It will be a continuous, background API call. You will get funded based on your Probability of Success, calculated dynamically:
1. The "Hard No" (Educational Rejection)
Probability: Very Low.
Outcome: The system rejects the capital request but provides the "Ghost" with a roadmap. "Your code velocity is high, but your CAC is undefined. Get $10k MRR or add a sales agent."
Value: We retain the relationship via the VentureOS, continuing to harvest data until they are ready.
2. The "Soft Commit" (Low Probability / High Variance)
Probability: Unclear, but non-zero.
Outcome: An automated offer with a lower valuation or standard terms. This is the "farming" bet. We buy the option to see if the Ghost populates the Shell.
3. The "Alpha Commit" (High Probability)
Probability: High Signal.
Outcome: Immediate offer with high valuation. We detected the outlier signal before the rest of the market saw the traction.
The Incumbent's Dilemma
The reason firms like General Catalyst or Sequoia haven't done this is not because they can't, but because their business model depends on the "High Priest" mystique. They sell the idea that their partners possess a magical intuition that cannot be replicated. Admitting that a "Ghost-reading" algorithm could outperform a partner's gut feeling devalues their brand. They are trapped in their own Shell—the heavy, expensive infrastructure of partners, associates, and offices.
We are not. We are building a lightweight protocol for capital allocation.
Eating the VC Business
If incumbents VCs don't move ASAP, we will eat their business up. Not because we have more capital, but because we are scaling with the right capital partners who understand that VC is shifting from a service business to a software business. We are committed to this vision: a Self-Service-VC, product-first approach that touches geographies and demographics that traditional VC ignores. The future of fundraising isn't a handshake at the Sand Hill Road Rosewood hotel. It is a "Commit" button in your IDE. It is a background process that funds you the moment your metrics cross the threshold. It is the Ghost in the Shell, finally given a voice.
For the sake of the humanity we hope the incumbents do this faster. But if they don't, we will change the world without them.
If you want to take part of this journey, send me a message via LinkedIn. Most of you won't, because of the Fear of Looking Stupid (FOLS).
Second chapter: Why the Smartest Money in VC Looks the Stupidest
