Social Capital

01Approach

What we do.

Everything we do sits on one of two sides. Both run on our own capital.

Proprietary investments and machine learning

Our own capital deployed through quantitative, systematic, and algorithmic trading strategies. We build these ourselves, from signal research through execution, rather than licensing a model or renting a platform.

The same machine learning stack drives our research into industrial and energy systems, and runs inside the stealth company built on it, where the constraints are physical rather than financial.

Active investments

Operating companies we own and grow. We set strategy, build the back office, and stay for the long term rather than for a holding period. Today that is Midsommer Health, a behavioral health organization across four states.

The two are not unrelated bets. They share people, standards, and one question: what does a system do under real constraints, and how do you make it do better.

02About

Who we are.

Social Capital was founded by Spencer R. Potesta, a licensed psychotherapist and researcher whose work runs across clinical practice, organizational psychology, and machine learning. The firm was built to hold both sides of that at once rather than choose between them.

How a firm raises money determines what the firm becomes. Outside capital arrives with a clock attached. The clock sets the time horizon, the time horizon sets the ethics, and the ethics eventually set the org chart. Every decision inside every company bends toward that schedule whether or not it is right for the business or the people in it.

We would rather own the clock. Investing only our own capital means we can hold a company for as long as holding it is the correct answer, pay people properly before it is convenient, and make decisions that look expensive this year and obvious in ten.

It also means the firm can point its returns somewhere other than a shareholder. Social Capital is built to fund a healthcare endowment held by the Social Capital Foundation, a separate 501(c)(3) described further down, and to take care of the people who work here. We do not think this costs us growth. We think it is the condition for it.

Time horizon

No fund life, no deployment window, no forced exit. Nothing in our structure requires us to sell something we would rather keep.

Ethics

There is no outside return to defend. The right call and the fundable call are the same call, which removes the pressure that quietly corrupts good operators.

Structure

The organization is built for the work in front of it, not for a reporting requirement designed around someone else's quarter.

03Holdings

The active side. Companies we own and operate.

Midsommer Health

Active · Behavioral health

A clinician-led behavioral health organization and practice network across Illinois, Ohio, Pennsylvania, and New York. It is our pride and joy, and where the thesis gets tested in public.

The standard private equity and venture playbook in this sector takes a fee, compresses cost, and sells. We funded Midsommer entirely with internal capital, so there is no fee clock and no exit clock. Clinicians are paid at the top of the market, management fees stay near zero until a practice can comfortably carry them, and clinical decisions stay with clinicians. We measure the company on how many people it reaches and how well it treats the people who work inside it, not on dollars generated.

midsommer.org

Industrial and energy research

In stealth

Machine learning applied to the optimization of very large compute and power loads, from data centers on the ground to systems designed to operate in orbit. The company itself is not named publicly yet, though the research behind it is described below. We would rather be judged on the work than on the announcement.

The modeling stack is shared with our systematic trading work. The same architecture that prices and positions in markets is pointed at physical constraints instead of financial ones, which is the reason the two sit inside one firm.

04Research

The proprietary side. Our own capital, our own models.

We build our models in house rather than licensing them from a vendor, because the interesting part of this work is the part that does not generalize. Research sits in the parent firm rather than inside any one company, so what we learn in one place is expected to show up in another.

On the investment side, that means quantitative, systematic, and algorithmic trading models that we write ourselves, from signal research through backtesting, risk, and execution, and then run on our own capital. Nothing is outsourced to a vendor model or a third-party platform, because the edge lives in the parts of the pipeline that cannot be bought.

The same class of model, retrained against physical constraints instead of financial ones, is what runs inside our stealth industrial and energy company. The work we are furthest along on there is energy optimization under physical constraint. Compute-heavy facilities spend most of their power on thermal load and idle capacity, and the physics governing that is knowable. We build physics-constrained models that treat conservation laws, thermal limits, and hardware failure modes as hard boundaries the model is not allowed to violate, then optimize energy usage inside them. A model that can suggest an impossible operating point is not useful to anyone running real equipment.

We apply this in three settings: industrial facilities, terrestrial data centers, and compute designed to operate in orbit, where the power and cooling budget is effectively the whole design problem.

Quantitative and systematic tradingAlgorithmic strategies we build end to end, from signal research through risk and execution, run on our own capital.
Physics-constrained optimizationThermal limits, conservation laws, and failure modes enter the model as hard boundaries rather than penalties that can be traded away.
Data centers, terrestrial and orbitalEnergy, thermal, and load optimization for compute-heavy facilities on the ground and for systems designed to run in orbit.
Industrial energy systemsProcess, load, and power optimization inside industrial plants, at multiple scales of the same system.

Licensing and collaboration

We do license our research and our models, selectively. We are not raising capital, but this is a different conversation and we are open to it.

Build

We build models against your systems, your constraints, and your instrumentation.

Advise

We work alongside your team on approach, architecture, and how to validate what comes out.

Deploy

We plug into a compute-heavy operation already running at scale and optimize against live load.

Reach out

05Mandate

What we are building.

What we are building is an institution meant to outlast any single company inside it, and to be worth working at for an entire career rather than a cycle.

That comes down to two commitments. The first is to the people here: compensation, benefits, and enough runway to do work that takes years to come good. The second is a healthcare endowment, funded by the firm and built to keep giving long after any of us are running it.

The endowment sits with the Social Capital Foundation, a 501(c)(3) with no corporate relationship to Social Capital. It is governed separately, and we intend to be a source of its funding rather than an owner of it. Funding it well over a long enough period is the reason the rest of this exists.

We would rather be measured in decades than in quarters, and in people reached than in dollars raised.

socialcapitalfoundation.com

06Contact

Talk to us.

We are always open to conversations with operators, researchers, clinicians, and organizations interested in licensing the research. We are not raising capital and do not accept outside investment.

Thank you. Your message is in, and someone will come back to you directly.