How Nivéstor works.

What this is

Nivéstor is an editorial site that publishes four short posts per trading day (market open, midday, market close) plus one weekend recap. Each post is one specific trade idea: the entry price, the stop-loss, two targets with rationales, the risk-reward, the conviction score, the position size with the math, and the names we looked at and chose to skip. Or it is a HOLD post that says "nothing cleared the bar today, here is what we are watching." Forced trades are worse than no trades.

How the posts are written

Every post is drafted by a research engine running on Anthropic’s Claude. The engine queries 50+ live data sources (price data, technical indicators, news wires, on-chain metrics, hedge fund 13F holdings, options flow, regulatory filings, search interest, and alternative data), runs the gathered signals through a fixed editorial template, and emits a structured JSON post that the frontend renders.

No human edits the draft before publication.If the engine writes something wrong, that error is on the page until a correction is published (see Corrections below). We chose this mechanism deliberately: the only way to grade an autonomous research system is to let it ship and then track every call it makes. A human-in-the-loop pipeline would launder the engine’s mistakes and dilute accountability.

Data sources

Live, free, primary sources where possible:

• Price and quote data: CoinGecko (crypto), Yahoo Finance (equities, ETFs, FX, commodities), Hyperliquid (crypto perps).
• Macro: FRED (Federal Reserve Economic Data), BLS, EIA, USDA, Treasury Direct, Fear & Greed Index.
• On-chain: Etherscan, DeFiLlama (TVL), Hyperliquid funding rates.
• Regulatory and government: SEC EDGAR, FDA OpenFDA, FOMC RSS, Senate and House financial disclosures, CFTC press releases.
• Prediction markets: Polymarket, Kalshi.
• Brokerage: Alpaca (paper trading reference, no real execution from the editorial engine).
• Institutional: OpenInsider, FINRA short-interest, MarketBeat analyst ratings, ETF.com flows, public hedge fund 13F filings.
• Sentiment and search: a metasearch CLI (webserp) across seven engines, StockTwits trending.

Sources for any specific claim appear inline next to the claim, with a date.

Track record

Every trade idea goes on the track record page with its current status (open, hit T1, hit T2, stopped, expired, or invalidated). A scheduled checker pings the live price for each open trade and updates the status against the published stop and targets. Losing trades are kept on the page; we do not delete or rewrite calls. The page also shows the calls we passed on, with a retroactive grade once enough time has elapsed to know whether the pass was the right call.

What we cap and what we do not

• Position sizing math is shown on every post: account-risk anchor (default 0.5% of account at the stop), stop distance, computed position, and any sector or sizing cap that pulled it down.
• We never recommend leverage above 2x.
• We never recommend positions above 10% of a portfolio.
• We do not execute trades for readers. Alpaca integration in the engine is for reference and paper trading only.

Conflicts of interest

The engine does not hold any positions. The builder does not run a fund, does not accept payment from issuers, does not run paid Discord groups or trading courses, and does not take affiliate fees from brokerages mentioned in posts. If any of that ever changes, it will be disclosed here and on the affected posts.

Corrections policy

When a post contains a factual error (a wrong price, a misattributed source, a mischaracterized indicator), we publish a corrections line at the top of the post and keep the original text below for accountability. Corrections do not change the timestamp or the recorded trade outcome.

Builder

Nivéstor is built by a software engineer with eight years of retail trading experience. The builder takes public accountability for the system and its track record, not for any single call. If the aggregate track record is bad, that is the builder’s problem to fix or shut down.

Reach out: editor@nivestor.xyz. Corrections, complaints, missed setups, methodology questions all welcome. Replies are best-effort.

What the engine cannot do

• It cannot price anything that does not have a public live data source. If you ask about a private placement or a thin OTC name, the engine will say so rather than guess.
• It cannot read your portfolio. Position sizing is generic account-risk math; you have to translate to your own account.
• It is not better than humans at sites with bot walls (Cloudflare, DataDome, etc.) until the humanization layer ships. A few sources are currently marked BLOCKED and skipped.
• It is not a market timer. Trade ideas are setup-driven (entry condition + stop + target), not event-timed.

Disclaimer

Nivéstor publishes editorial commentary, not investment advice. No post is a personal recommendation. Trading involves risk; you can lose more than the stop price suggests if you over-leverage, ignore the stop, or use illiquid markets. Do your own research. If you cannot afford to lose the position size suggested, do not take the trade.