An investor's edge comes from personal context built through years of experience, rather than from models and tools that everyone can access.
2
AI should handle reading, data gathering, cleaning, and cross-checking while the human keeps responsibility for the decision.
3
Moshi provides a 24/7 quant analyst that combines market data with an investor's private documents and explains the claims and assumptions behind its analysis.
Summary
Anekha Sokhal argues that finance has a growing gap between the amount of available data and the attention an investor can give it. An investor's edge comes from knowing what is happening early, seeing blind spots, and having enough conviction to act. Since models, prompts, and connectors are broadly available, the harder-to-copy context is the experience held in an investor's head. Sokhal says AI should feed that human reasoning rather than replace it. Machines can gather information, clean it, compare sources, monitor markets, and trace claims. Humans should retain the decision, especially because financial choices require someone to explain why they were made. She presents Moshi as a 24/7 quant analyst built from six deterministic agents. It processes market information and an investor's documents, then reports sources, assumptions, claims, and possible hallucinations. Sokhal is direct that the amount of investor context a machine can capture remains unanswered.
More data widens the gap between information and attention
Sokhal opens with the daily problem facing financial investors: the world produces vastly more data while each person's attention stays limited. She cites about 175 zettabytes of data produced in 2025, alongside information from PDFs, subscriptions, podcasts, X, and LinkedIn. More data does not automatically create more information for an investor. It gives people the same flood to process, while the work of financial synthesis still depends on limited human attention.
Sokhal compares the human brain with the systems built to imitate parts of it. She says the brain runs on about 20 watts, roughly the power of a dim light bulb, while AI systems require a power plant. That comparison leads to a design question: where should machine learning fit in the decision cycle? If the most efficient reasoning engine is already between the investor's ears, Sokhal argues that technology should supply it with better context instead of trying to remove it from the process.
An investor's edge is personal context that cannot be downloaded
Sokhal defines edge through three investor needs: knowing what is happening before others, seeing blind spots and second- or third-order effects, and having enough conviction to size a trade. Better models, prompts, skills files, and MCP connectors are available to everyone, so they do not create a durable advantage by themselves. The harder-to-copy advantage is the context built through decades of work and learning beside experienced traders. When an investor retires or changes firms, that knowledge can leave with them.
Sokhal says an automated agent cannot own a financial decision in the same way a person can. The other side of a trade may be a client, a physical cargo, or a company made up of people. When something goes wrong, someone must be able to explain why the decision was made. An agent that leaks an API key can apologise and suggest rotating it, but that response does not provide the accountability required in finance. Sokhal also warns that a small error can grow as an agent continues to act.
The work before a trade splits into machine work and human judgment
Before making a trade, an investor reads, gathers information, watches the market, checks conflicting sources, and decides what is true. Sokhal calls this the grind, which includes reading, data gathering, and cleaning. She says machines are good at that work. Judgment remains the investor's job. This split lets a system reduce the time spent assembling and checking information while leaving the final choice with the person who must act on it.
Moshi combines market monitoring with an investor's private context
Sokhal describes Moshi as a 24/7 quant analyst for institutional traders. It uses a team of six agents to watch the market and process more information than a person could, together with data from PDFs, emails, and subscriptions belonging to the investor. She says Moshi has read 2.4 billion data points and learned 300 new things in the time needed to read a sentence. The point is to give the trader more useful context for a decision, rather than another model to debate.
Traceability matters because confident financial answers can still be worthless
New context only helps if the investor can judge whether it is trustworthy. Sokhal says Moshi measures hallucinations, measures each claim, tracks where claims came from, and shows the assumptions behind its conclusions. An investor asking for an opinion needs to know how the system reached it, not only what it thinks. This makes traceability part of the product rather than an afterthought to an agent's answer.
A private weekend tool becomes infrastructure when other people must trust it
Sokhal distinguishes a quick personal tool from a system that runs continuously across markets and handles large amounts of data. Once another person depends on it, the system needs reliability, security, storage, and a way to trace decisions back to their sources. She says this is closer to the infrastructure of an institutional quant team than to a dashboard built over a weekend. She also admits that the amount of an investor's real edge a machine can capture is still unanswered.
"Machines are incredible at the grind. They're incredible at cleaning data. They're incredible at gathering data. And then the human is incredible at the decision making."10:42
Who should watch
You are building AI tools for traders or investors and need to decide which parts of the workflow should remain with a person.
Your finance team has useful knowledge trapped in experienced investors' heads, emails, PDFs, and subscriptions.
You are moving from a personal prototype to a system that other people must trust, audit, and use continuously.
