Welcome to Arithmos. Here is how it works, in plain English.

Arithmos turns a plain-English sentence (“companies building AI picks-and-shovels”, “UK dividend aristocrats”, “European defence names under £50bn”) into a transparent, rule-based basket of stocks called an index.

You describe the exposure you want. The agent picks the companies, explains why each one made the cut, assigns a weight to every holding, and runs a backtest so you can see how that basket would have behaved historically. You can then export the holdings to any broker and buy them yourself.

Most “AI investing” tools are thin skins over a general-purpose chatbot, asking it nicely to return a list of tickers. Arithmos is built the other way round. Our in-house portfolio-construction agent runs inside an agentic research stack we built end-to-end and wired to real market-data infrastructure.

Our research team comes from quantitative equity, systematic trading, and market-data engineering backgrounds. Years of that institutional experience are baked into the agent, the tool schema, the weighting policies, and the validation layer. Those are the parts that decide what the agent is allowed to do and which decisions it has to justify with data. None of that shows up in a competitor’s screenshot, and none of it can be reproduced by stitching together off-the-shelf chatbot APIs.

The stack behind every generated index:

• In-house portfolio-construction agent. Calibrated on thousands of evaluated runs, with guardrails that force it to screen, verify, and ground every holding in tool-returned data. It cannot invent tickers, fabricate fundamentals, or skip the validation step.
• Proprietary multi-provider data fabric. A resilient fan-out across Yahoo, FMP, Tiingo, Finnhub, Twelve Data, Alpha Vantage, Polygon, Stooq, and TradingView. Each query hits the primary provider and automatically falls through on auth errors, rate limits, empty responses, or upstream wobble. Survivorship-bias controls, currency normalisation, and corporate-action handling sit on top.
• Deterministic schema enforcement. Every index emits a machine-readable rulebook (universe, filters, weighting method, rebalance cadence, per-holding cap). Weights are normalised, capped, and summed to 1.0 in code, not left to the model.
• Tamper-evident verification pipeline. Our admin verification harness independently recomputes every backtest from raw prices using an isolated math library, compares against the cached result, and writes a hash-chained record so any drift between what you see and what the portfolio actually did is detectable.
• Institutional-grade metrics.CAGR, annualised volatility, Sharpe & Sortino (both user-configurable against the prevailing risk-free rate), max drawdown, best/worst year, rolling alpha vs any chosen benchmark, survivorship-adjusted total return.
• Platform-agnostic export. CSV, JSON, and broker-native formats (Trading 212 today; IBKR, Schwab, Fidelity, and Vanguard on the roadmap) so the file you take to your own broker matches what you see on screen.

The agent itself is the easy part. The hard part is the research methodology, the data plumbing, the verification layer, and the guardrails that make the output reliable enough to put real money behind. That is what our team has spent the last several years building, and it is why surface-level wrappers produce markedly different results.

An index is just a list of stocks put together by a rule. The S&P 500, for example, is a list of roughly 500 of the biggest public companies in the United States. The rule decides who’s in, who’s out, and how much of the list each one represents.

An index fund is a pooled investment (a mutual fund or ETF) that buys every stock in an index in the same proportions. Instead of picking one company and hoping, you own a slice of the whole list. If the list goes up, your slice goes up.

Why people like index funds:

• Diversification. You own many companies, so no single blow-up wipes you out.
• Low cost.There’s no star manager to pay; the rule runs itself.
• Transparency. You can see exactly what you own.
• It works. Most active stock-pickers fail to beat a simple index over long periods, after fees.

Historically, “the index” meant whatever Standard & Poor’s or FTSE published. This tool lets you define your own index, built around a theme you actually care about, with the same kind of rules professional index providers use.

“Weighting” is how the rule decides what percentage of the index goes into each stock. The same list of companies can behave very differently depending on the method.

1. Open the home page and type a description into the prompt box. Be specific if you have opinions: “US only”, “max 20 holdings”, “exclude tobacco”, “cap any one name at 8%”.
2. Watch the agent build it. You’ll see it choose the universe, score candidates, apply caps, and finalise weights. Each step is visible so you can see why a name was picked.
3. Land on the index page. You get: holdings table with sectors, weights, and rationale; a backtest chart vs the S&P 500; metrics (CAGR, vol, Sharpe, Sortino, drawdown); and the excluded list.
4. Hover the little i buttons anywhere on that page for plain-English explanations of each term.
5. Use the export menu to download the holdings in a format your broker can read, or re-run to build the same prompt again with fresh data.

Index

A list of stocks (or other assets) assembled by a rule. 'The index' on its own usually means the S&P 500.

Index fund / ETF

A fund you can buy in one trade that holds every stock in an index in the right proportions.

Ticker

The stock exchange's short code for a company. AAPL = Apple; MSFT = Microsoft; AZN.L = AstraZeneca in London.

Holding

A single stock inside the index, plus its target weight.

Weight

The % of the whole index that one stock makes up. Weights across all holdings sum to 100%.

Allocation

Plain English for the same thing: how the 100% is split up.

Market cap

Share price × number of shares. A company's total stock-market value. 'Large cap' ≈ >$10bn, 'small cap' ≈ <$2bn.

Sector

A bucket of companies in the same type of business: Technology, Healthcare, Financials, Energy, etc.

Universe

The pool of candidates the rule picks from (e.g. 'all US-listed companies over $500m market cap').

Benchmark

The index you compare yours to, usually the S&P 500. Helps you see if you beat 'the market' or not.

Backtest

A simulation of 'if I'd held this basket over the last N years, what would have happened?' Uses historical prices.

CAGR

Compound Annual Growth Rate. The single smoothed % per year that would turn your starting value into your ending value. Lets you compare periods of different length fairly.

Volatility

How much the price bounces around year to year. Measured as the standard deviation of returns, expressed in %. Higher = bumpier ride.

Sharpe ratio

Excess return above the risk-free rate per unit of total volatility. Above 1 is decent; above 2 is great; below 0 means cash (at the selected rate) would have done better.

Sortino ratio

Like Sharpe, but only penalises downside volatility; upward swings don't count against you. Better suited to long-only equity strategies where big up-moves are good news, not a problem.

Risk-free rate

The return you can earn for certain, typically a short-term government bond yield such as the US 3-month T-bill. Sharpe and Sortino measure the excess return above this rate. At 0% the ratios show raw return-per-risk; at 4.5% they reflect the real opportunity cost of investing rather than holding cash.

Downside deviation

The standard deviation of negative returns only (below a target). Used in the Sortino ratio in place of total volatility.

Drawdown

How far the index has fallen from its most recent peak. Max drawdown = the worst such fall in the whole period.

Rebalance

Re-running the rule on a schedule to refresh holdings and weights. Needed because prices move and the basket drifts from the target.

Cadence

How often rebalancing happens: monthly, quarterly, semi-annual, annual. Trade-off between staying on target and trading costs.

Per-holding cap

A ceiling on how much of the index any single stock can be, e.g. 8%. Protects against concentration risk.

Concentration risk

The danger of one company (or sector, or country) being too big a share of the basket.

Dividend

Cash a company pays its shareholders out of profits. Received automatically by your broker if you own the stock.

ADR

American Depositary Receipt. A wrapper that lets US investors buy a foreign company's shares in USD on a US exchange.

Exchange

Where stocks trade. NYSE and Nasdaq (US), LSE (London), etc. The ticker suffix tells you which.

Factor

A measurable characteristic of a stock (size, value, momentum, quality, low-volatility) that historically explains part of its return.

Smart beta

Marketing term for factor-weighted index funds. More systematic than active, more opinionated than plain market-cap weighting.

ESG

Environmental, Social, Governance. A set of non-financial criteria used to include or exclude companies on ethical grounds.

Slippage

The gap between the price you planned to trade at and the price you actually got. Small on big stocks, bigger on thin ones.

Turnover

The % of the basket that changes at each rebalance. High turnover means more trading costs and more taxable events.

Snapshot

A frozen-in-time version of the index's holdings and weights. Every rebalance produces a new snapshot.