How Thry Analyses 93 Channels in 60 Seconds

When you submit a hypothesis to Thry, a lot happens in the next 60 seconds. Here's what each stage does, and where the boundaries are between AI judgement and deterministic computation.

The pipeline

Thry processes your hypothesis through seven stages. Three of them run in parallel, which is how we deliver a full report in under a minute.

1. Parse — AI extracts the investment thesis, stance, themes, risk factors, bull/bear cases, and trade implications
2. Narrative search — Headlines from financial-news feeds and real-time web search, plus voice matching against 93 tracked channels
3. Equity discovery — An AI research agent identifies thesis-aligned companies with alignment scores and written rationales
4. Technical analysis — Six indicators computed deterministically from market data, combined into a composite score
5. Alignment analysis — Equities ranked by thesis alignment, adjusted for technical context and volatility
6. Backtesting — Portfolio tested against SPY with up to a year of daily data
7. Report generation — An AI-generated executive summary ties everything together

Steps 2, 3, and 4 run in parallel. That's how we get a full report in under a minute.

Parsing your hypothesis

The first thing Thry does is understand what you're actually claiming. The parser extracts:

• Stance: bullish, bearish, or neutral
• Bull and bear cases: the strongest arguments for and against your thesis
• Themes and sectors: the market areas your thesis touches
• Trade implications: what would happen in markets if your thesis is correct
• Evidence framework: what data would validate or invalidate your view

Before any analysis begins, a safety gate screens for PII and non-investment content. If you accidentally paste personal information, it gets caught here.

The structured output from this stage drives everything that follows. The narrative search knows what to look for. The equity discovery agent knows what kinds of companies to find. The entire pipeline is shaped by your specific thesis, not a generic template.

Narrative search

Thry searches for relevant market narrative through two tracks simultaneously.

Financial news and web headlines are pulled from configured market-data feeds and real-time web search. Articles are scored by how many of your hypothesis's key terms appear, filtered by a relevance threshold, and classified by their sentiment relative to your specific thesis. This last part matters: if your hypothesis is bearish, then a bearish article aligns with your view and a bullish one contradicts it. Polarity-aware alignment means Thry understands the direction of your claim, not just the topic.

Voice matching searches a database of 1,600+ pre-extracted opinions from 93 tracked financial YouTube channels with a combined subscriber reach of 94M+. The search runs in two stages: keyword matching against your hypothesis terms, and a recency pass that catches relevant commentary using different language to yours.

Each voice carries an individual credibility weight from our source database — not a blanket score for all influencers. Display influence is log-scaled by subscriber count, so smaller expert voices aren't drowned out by raw audience size, but channels with large followings still carry appropriate weight. The aggregate voice sentiment accounts for both alignment direction and source credibility.

After deduplication, every source — headline or voice — is classified into three buckets: aligned, contrary, and neutral.

Why contrary evidence matters

Most people look for information that confirms what they already believe. That's confirmation bias, and it's the enemy of good investing.

Thry deliberately surfaces contrary evidence. Every report shows you not just the sources that agree with your thesis, but the strongest arguments against it. The top aligned and top contrary sources are highlighted by credibility score, so you're seeing the most authoritative voices on both sides.

We think the most valuable thing we can show you is a well-reasoned opposing view you hadn't considered.

Technical analysis

For each candidate stock, Thry computes six technical indicators from market data: RSI, MACD, Bollinger Bands, ADX, 50-day SMA, and 200-day SMA. All prices are converted to USD using live exchange rates.

These are combined into a composite technical score using a deterministic formula. Each indicator contributes independently to the final score, which ranges from -1 (strongly bearish) to +1 (strongly bullish). The computation is fully reproducible — the same market data always produces the same score.

This doesn't override the thesis. It refines timing. Two stocks with equal alignment get differentiated by which has better technicals right now. Each holding also gets a plain-language technical narrative explaining what the indicators show and what they mean in context.

We also estimate annualised volatility from the market data. This feeds directly into the alignment analysis.

Alignment analysis

This is where thesis alignment, technical context, and risk adjustment come together.

Equities are ranked primarily by how directly they express your thesis, supplemented by their technical profile and adjusted for volatility. Higher-volatility stocks receive smaller analytical weight so no single name dominates the view. Hard constraints cap individual positions, enforce minimum significance, and limit total holdings to keep the analysis focused and diversified.

The output is a ranked analytical view showing which stocks best express your thesis, not a portfolio recommendation. Each holding includes an alignment score, a technical profile, and a written rationale explaining its inclusion.

Critically, if your thesis is bearish ("banks will decline"), the system finds the beneficiaries of that scenario — not the banks themselves. This is a long-only analytical framework that identifies what would perform well if your thesis is correct.

Backtesting

The ranked holdings are tested against the S&P 500 (SPY) using up to a year of daily price data. The key metrics computed are:

• Total return: simple percentage change over the period
• CAGR: annualised compound growth rate
• Alpha: return above or below the benchmark
• Sharpe ratio: risk-adjusted return relative to a risk-free rate
• Max drawdown: largest peak-to-trough decline
• Annualised volatility: standard deviation of daily returns, annualised

Contribution analysis breaks down which holdings drove performance across multiple time horizons — from 7 days to the full period — so you can see whether returns came from broad thesis alignment or a handful of names.

The most important thing about the backtest? It's hypothetical. We're very upfront about this. It uses historical data with perfect hindsight. It doesn't account for transaction costs, slippage, or taxes. We include these caveats in every report because they matter, and because ignoring them would be dishonest.

What's AI and what's not

We think this distinction matters, so we're transparent about it.

AI-driven stages: hypothesis parsing, equity discovery, voice relevance scoring, executive summary generation. These use AI judgement and may produce different results on repeated runs.

Deterministic stages: technical indicator computation, composite scoring, alignment weighting, volatility adjustment, constraint enforcement, backtesting. These use fixed formulas — same inputs always produce same outputs. No AI involved.

This matters because it means the quantitative backbone of every report is fully reproducible. The AI handles the parts that require interpretation — understanding your thesis, finding relevant companies, assessing narrative relevance. The maths is just maths.

Try it yourself

You can test two hypotheses per month for free at thry.ai. Type a macro theme, a sector view, a contrarian position — anything you want to pressure-test. And see what happens when you put it through a proper process.

Disclaimer: Thry is a research tool, not a financial adviser. All analysis is AI-assisted and may contain inaccuracies. Past performance does not guarantee future results. Always do your own research and consult a qualified financial adviser before making investment decisions.