Writing
Knowledge could survive beyond memory.
A new layer in human cognition
Think through other minds.
Writing let us store knowledge. Education let us transmit knowledge. Search let us access knowledge. LLMs let us converse with knowledge.
Gothos provides other minds through which to navigate the world.
The cognitive ladder
Every major information technology expands what the mind can do.
Education
Knowledge could be disciplined, preserved, and transmitted.
Web search
Knowledge became searchable at planetary scale.
LLMs
Knowledge became conversational.
Gothos
Agents let us see the world through different eyes, not just engage with more of it.
The constraint
We make decisions inside inherited worldviews.
The opportunities we notice, the futures we can imagine, and the trade-offs we anticipate are all shaped by priors. We spend our lives inside a narrow band of possible perspectives. We don't choose to. Acquiring a new worldview is expensive in time and opportunity cost, and we often don't even know what would change.
To think like a philosopher, an engineer, a hedge fund manager, a political scientist usually takes years, and that doesn't touch our more fundamental priors. Gothos is built around a simple question: what happens if worldviews become accessible as structured, accessible instruments? What happens if we can not only make better decisions and reduce the unknown, but understand the dispositions that have shaped — and will continue to shape — our lives?
The vision
The next step is not more answers. It is access to better ways of seeing.
Gothos is not trying to build another chatbot. It is an attempt to build infrastructure for explicit perspectives: systems with stated priors, bounded methods, specialist knowledge, adversarial review, and a memory of where they were wrong. Independent agents, trained with discrete methodologies and knowledge bases, will offer news lenses through which to see the world; far more than a prompt beginning with 'You are a...', they are genuinely distinct, tested, and calibrated.
Economist
Incentives, allocation, constraints.
What are the structural forces, opportunity costs, and second-order effects?
Historian
Analogy, path dependence, contingency.
What does this resemble, what happened then, and what can we learn?
Technician
Mechanism, feasibility, bottlenecks.
What is technically feasible, what is ripe for innovation, and what is overhyped?
Psychologist
Narrative, bias, social proof.
Where are we misperceiving the landscape of risk, priority, and reward?
Everest, not a staircase
The summit is far away. The first step is now.
The long-term version is almost science fiction: a marketplace of perspectives, where people can slip on different minds in every moment.
People moving through life with access to multiple minds, able to inspect their own assumptions and borrow others. A founder thinking through a market with a venture capitalist. A student revising for exams with a professor next to them, and thinking about careers next to the people they will work for. An investor with a macroeconomist, star trader, and behavioural psychologist.
That future is not the first product. It is the mountain. We are starting with the smallest real version of the idea. A narrow domain, a small number of perspectives, explicit priors, and outputs that can be compared against reality.
What we are building now
Gothos v.0 is self-reflective.
Right now, we are building a focused product for ourselves and other founders: a VC, whose eyes you can look through at any moment. Not a chatbot, not an assistant. An agent trained from the ground up to think as a VC, with a VC's knowledge. No implicit priors, no hallucinations, no generalised inferences.
A user selects a theme — getting funding, finding cofounders, building a pitch — and receives a structured view of idiosyncratic evidence. The view is built from a small set of specialised perspectives, the usefulness of which have been rigorously tested and calibrated. The advice you get has been backtested, scored, and refined. We asked what would happen if Gothos said this in 2024? 2022? 2020?
The agent doesn't provide a synthesis of what's on the web. Doesn't scrape YC's website for advice. Doesn't provide a hedged judgment with no real impetus. It starts and finishes with a specialised methodology, a curated and constantly updated body of evidence, and will give you a genuine answer.
Our proof is simple: can different methodologies applied to the same live evidence produce meaningfully different conclusions? Can those conclusions be made transparent, useful, and eventually scoreable?
Curated evidence
Community discussion, technical signals, research velocity, and commentary.
A specialist's knowledge-base.
Explicit lenses
Each perspective has a bounded method, not a generic instruction to sound expert.
A truly meaningful, discrete outlook.
Adversarial comparison
The point is not consensus. The point is productive disagreement.
Users don't get LLM false-judgment: they get the result of a reasoned debate.
Historical memory
Predictions, mistakes, priors, and changes in confidence survive and
improve reasoning.
Flawed perspectives improve, learning from errors made live and in backtests.
Why it matters
Information changed what humans could know. Gothos is interested in changing how humans can see.
The first product represents the first steps up Mount Everest: it shows that perspectives
can be represented, constrained, tested, improved, and made available to others.
That is the beginning of a transformative cognitive infrastructure.
Collaborators wanted
We are looking for builders and future co-founders.
Gothos is at the beginning. We are looking for people who find the vision compelling: engineers, researchers, designers, domain specialists, and possible co-founders who want to build the first real systems for thinking through other minds.