I built ETF Copilot (https://etf-copilot.com/go/show-hn) as analytical support for my own investing. It answers what financial screeners and platforms do not: which valuation, growth, quality and risk metrics matter for each ETF, and how do they compare to own history and peers. It covers 1,000+ US-listed equity ETFs and updates every night with fresh data.

Free version without signup offers 5 fund views. Overlap, concentration, and theme pages are unlimited. Pro $20/month tier adds unlimited views, full rankings, watchlists, alerts, and twice a week ETF market newsletter. For Show HN I opened a broader access showing most functionality for free.

> > > Background About a year ago I became much more interested in investing in ETFs. I have been in Finance for many years and saw many systems and tools from Yahoo Finance to expensive enterprise solutions. They all gave plenty of data, but getting to ETF-level conclusion still took hours.

> > > How it grew into a system I started building a small tool for myself. Rolling up stock level analysis to an ETF level was a key requirement, which enabled answering questions like “Is this ETF expensive” or “How profitable are companies in this ETF”. I automated it to run daily for US-listed equity ETF universe, which enabled ETF rankings based on valuation, growth, quality, and risk. I added features people were asking about on forums and social media: multi-fund comparison, overlap and concentration. Over time it evolved into an app.

> > > Technical problems solved 1) Mapping ETFs to stocks. ETFs hold many companies listed on international exchanges, reporting in different currencies, having different classes of securities. Tickers, ISIN, FIGI and company names can vary or be missing. I ended up building a fairly long set of rules to resolve the inconsistencies and map the data, as mapping quality directly influenced metrics downstream.

2) Metrics aggregation and rankings. Negative P/E, missing data, abrupt changes in some metrics caused wild swings in early versions, and a straightforward weighted average approach worked in some cases but broke in others. Developing a system of calculation and coverage rules was an integral part to raise quality of outputs. For rankings, “Gate first, weight second” filtering to remove poor performers proved to be a much more effective and stable approach than a seemingly logical weighted average across all.

3) LLM explanations grounded in facts and without hallucinations. I wanted to provide short daily explanations on each ETF and the market, instead of pushing the user to analyze dozens of metrics. I ran data analysis deterministically and built compact “factpacks” on each ETF and sector, as the only data LLM sees. Then fed those to LLM to write explanations and put a deterministic validator checking numerical claims against the source data, as well as a separate semantic LLM validator checking unclear or misleading wording.

> > > Questions 1. Can you quickly get the needed data in app to tell “what sets ETF X apart from similar funds”? 2. What would you improve to make it more actionable and useful for individual ETF investors? 3. Which steps and data points do you use, when you consider investing in ETFs?

> > > Website: https://etf-copilot.com/go/show-hn

Unlike many parents, children are fearless around AI

Polling has become a playground for grifters, liars, and frauds

Make Videos Programmatically with React

Disney+ and Hulu raise prices by up to 13 percent after doubling profits

The Year AI Came for Us: Teaching Entrepreneurship Will Never Be the Same

What Australia's Hidden Caves Can Teach Us About Martian Terrain

Show HN: Are My ETFs Overvalued?"

Mel: A Real Programmer

Air – nightly discovery scores for 1,800 US small-caps, with a paper trail

Show HN: Learn SQL free with no tracking or ads

Parenthood is not Instant Gratification

Invasions of Canada (1812)

Magnetic Thermal Transistor

Qualcomm Announces Renewed Agreement with Apple

Tutoring company tells parents to save their money and 'use AI instead'

Cloudflare addressed a cross-tenant data exposure vulnerability in Containers

Intent, Not Sophistication: The AI Attacker Is on the Record

Ask HN: How do you read, sign off, verify, manage the volume of generated slop?

America's biggest convenience-store chains overcharge consumers

Evidence: The Missing Runtime Stream (a.k.a. Not Logs or Telemetry)

Robot Basketball

Fedora Discussion Raises the Idea of Replacing LibreOffice with Collabora Office

Proof a Human Sent an Email

The Zig Journey

Entrepreneurship and Industry in Europe

Artificial Symbiotic Intelligence

Natural-Fiber Clothing Is in Demand–But Scientists Say It Needs More Scrutiny

Japanese used bookstores see 5x sales surge as books are being bought by the ton

Breaking Up with Google Play: Why Conversations Is Now Free

$5,999 Ryzen AI MAX+ 395 and Radeon R9700 workstation