Know the business. See the risk. Then trade.

Balance Labs is an AI stock research workspace. It chains AI Stock Chat, a Berkshire-style fundamentals desk, Stock Screener timing signals, and Labs backtesting into one decision workflow — so every trade starts with scored evidence instead of a hot take. Free plan: 30 monthly AI credits, no card required.

753tickers in the research book
5markets — US, HK, China A, Korea, India
4,300+public dated research pages
Freeplan with 30 AI credits/month

One workspace, four research tools

AI Stock Chat

Ask any ticker question and the workspace assembles the answer from fundamentals, filings, and scanner data — with sources and dates, not vibes. Every answer carries the snapshot date so you know exactly how fresh the evidence is.

Berkshire fundamentals desk

Buffett/Munger questions turned into an operational scorecard: returns on capital, moat durability, 13F honesty, and price versus intrinsic value. The same checklist runs on every ticker, so scores are comparable across your whole watchlist.

Stock Screener + STM signals

Daily multi-source bars across five markets, screened for timing signals. Signals are the last step of the workflow — they only matter after quality and valuation pass.

Labs backtesting

Test strategies and stock theses together. Backtests use the same data the research pages cite, so what you validate is what you read.

The four-part research workflow

  1. Business quality — score the company on the Berkshire checklist before looking at the price. If the business is mediocre, nothing else matters.
  2. Valuation ceiling — compute what the earnings power is worth with the free two-stage intrinsic value calculator, then demand a margin of safety below that ceiling.
  3. Thesis breaks — write down the measurable events that would falsify your thesis before buying. This is the cheapest risk control that exists.
  4. Timing — only now check the STM Screener signal. Sequence beats speed: signal without quality context is just noise.

The full method, with worked examples: AI stock research workflow · valuation ceiling & thesis breaks · the Buffett scorecard

Coverage you can cite

Every ticker in the book gets a decision brief, an outlook framework (explicitly not a price prediction), and dividend, earnings, and insider pages where relevant — each with a dated facts table and FAQPage schema, so the numbers you quote are traceable to a snapshot date.

🇺🇸 US · 🇭🇰 Hong Kong · 🇨🇳 China A-share · 🇰🇷 Korea · 🇮🇳 India — or start from the live market dashboard.

นักลงทุนไทย: บรีฟหุ้นยอดนิยม 150 ตัว (ตัวอย่าง AAPL) และคู่มือฉบับเต็ม วิเคราะห์หุ้นด้วย AI — ทั้งหมดฟรี

Frequently asked questions

Can AI really pick stocks?

AI does not predict prices — anyone claiming it does is selling hype. What AI genuinely does well is assemble research: pulling fundamentals, filings, and price history into a structured decision. Balance Labs scores business quality, sets a valuation ceiling, and names the conditions that would break your thesis. The decision stays yours — made on evidence.

Does AI stock trading actually work for retail investors?

Retail AI trading fails when it skips process — buying signals without understanding the business. What works is workflow-first: quality, valuation ceiling, thesis breaks, then timing. That sequence is exactly what the workspace automates and every research page documents.

What is Berkshire-style stock analysis?

The Buffett/Munger discipline as an operational checklist: high returns on capital, a durable moat, honest management (the 13F-honesty test), and a price below a conservative intrinsic-value estimate. Each question becomes a trackable score across 753 tickers.

Is there a free AI stock screener?

Yes — the Screener preview is free, plus 30 monthly AI credits. It covers five markets with daily multi-source bars, and its STM timing signals are the same ones published on the public decision-brief pages.

How do you calculate a stock's intrinsic value?

A two-stage residual income model: book value today, forecast earnings growth, and a discount rate you control, with a margin-of-safety readout. The calculator is free and every input is documented — no black box.

Which markets does Balance Labs cover?

753 tickers across the US, Hong Kong, China A-share, Korea, and India, plus crypto and forex through the scanner's sources. Every ticker gets the same workflow and a dated brief.

Is Balance Labs free to use?

Yes — Stock Chat, the Screener preview, and 30 monthly AI credits are free. Guides, the calculator, and all 4,300+ public research pages need no account at all.

ใช้ Balance Labs วิเคราะห์หุ้นด้วย AI อย่างไร?

เปิด Stock Chat ใส่ ticker ที่สนใจ ระบบประกอบงานวิจัยสี่ส่วน: คะแนนคุณภาพธุรกิจสไตล์ Berkshire, เพดานมูลค่า เงื่อนไขที่ธีสิสพัง และสัญญาณจังหวะ — ฟรี 30 เครดิต/เดือน อ่านคู่มือภาษาไทย

How is Balance Labs different from Danelfin or FinChat?

Most tools sell you a score. Balance Labs hands you the workflow that produces the score: the Berkshire checklist, the two-stage valuation math, the thesis-break discipline, and the timing signal — each step inspectable and dated. We publish the method itself, free, across 400+ FAQ pages and 36,000+ research pages; competitors publish marketing pages. Full head-to-heads: vs Danelfin, vs Stock Unlock.

Does Balance Labs give buy/sell signals?

The Screener produces dated STM timing signals — but deliberately as the last step, after quality and valuation pass. A signal without a thesis is noise; we refuse to ship naked buy buttons. Every signal page says what it is: state, not prophecy.

What data does Balance Labs use?

Daily bars through redundant public sources — Yahoo Finance, Stooq, Tencent, Eastmoney, Sina, Binance, OKX, MT5 — so one broken feed never blanks a ticker. Fundamentals come from the dated book snapshot; every research page shows its snapshot date and source.

Can I backtest my strategy before risking money?

Yes — Labs backtests strategies and stock theses on the same data the research pages cite. What you validate is what you read, which is the whole point.

Is Balance Labs safe and legitimate?

Balance Labs is a research workspace: we never hold funds, never guarantee returns, and ask for no card on the free plan. Our code and tooling are public at github.com/TerissilinRPA, our editorial standards are published at /about, and anything promising guaranteed profits "from Balance Labs" is a scam — see the legitimacy checklist.

What markets and tickers are covered?

753 tickers across US, Hong Kong, China A-share, Korea, and India, plus crypto and forex through the scanner. Start anywhere: NVDA brief, AAPL แบบไทย, or the AAPL vs MSFT head-to-head.

Who it's for

  • Serious DIY investors who want Buffett-grade discipline without a Bloomberg budget — the fundamentals desk and calculator replace spreadsheet grunt work.
  • Swing and position traders who refuse to buy a chart they haven't researched — quality and valuation gates run before the timing signal is even shown.
  • Beginners who want honesty — the beginner guide says plainly what AI can and cannot do, and the free plan lets you learn on real tickers without a card.
  • International investors, ภาษาไทยพูดได้ — Hong Kong, China A-share, Korea, and India coverage is first-class, and 150+ ticker briefs plus full guides are in Thai.
  • Developers and AI builders — an MCP server exposes the research tools to your own agents, and llms.txt plus RSS make the content machine-citable.

Honest by design

Dated snapshots, not vibes

Every number on every research page carries the date of the scanner snapshot that produced it. When you quote a price, signal, or uPnL figure, you can point to the exact day it was true. Pages that go stale say so.

No fabricated predictions

Outlook pages are frameworks, not price targets — we say so on every one. The KB rule the whole site is built on: if a fact did not come from a dated source, it does not ship.

Real questions from real searchers

Our FAQ pages answer questions harvested from actual search data (People-Also-Ask clusters), answered from structural facts — then dated. Browse the full FAQ hub: 400+ pages across workflow, tools, tickers, and nine entity deep-dives, in English and Thai.

AI-native and crawler-friendly

The whole research surface is static, indexable HTML with FAQPage, Article, and Breadcrumb schema — plus a machine-readable llms.txt and RSS feed so AI answer engines can cite us accurately.

How the data flows

  1. Multi-source daily bars — the scanner pulls prices through redundant public sources (Yahoo Finance, Stooq, Tencent, Eastmoney, Sina, Binance, OKX, MT5), so one broken feed never blanks a ticker.
  2. Scoring layer — fundamentals from the book are scored on the Berkshire checklist; valuation math runs through the two-stage residual-income model.
  3. Signal layer — STM timing signals evaluate on top of scored data, never instead of it.
  4. Publication — snapshots render to dated static pages (4,300+ and counting) with full schema markup, the same pages AI engines and you both read.

Curious about a name right now? Try NVDA, MSFT, Samsung Electronics, Tencent, or Kweichow Moutai.

The method, in full

Four steps, always in order. Skipping ahead is how good tickers become bad positions.

1 · Business quality (the Berkshire checklist)

Score the company before the chart: returns on capital over a decade, moat durability (network effects, switching costs, brand, scale), management honesty (the 13F test — do the holdings match the story?), and earnings quality (cash flow backing accounting profit). A mediocre business at a cheap price is still a mediocre business.

2 · Valuation ceiling (the math)

Two-stage residual income: today's book value, a conservative growth path, and a discount rate you control. The output is a ceiling — the most you can pay without betting on miracles — then demand a margin of safety below it. Run the free calculator →

3 · Thesis breaks (the exit before the entry)

Write the measurable events that would falsify your thesis before buying: margin bands, share-loss thresholds, regulatory triggers, capex cuts. This list turns "I feel nervous" into "condition three hit — I'm out." It is the cheapest risk control in investing.

4 · Timing (STM signals, last)

Only after steps one to three pass: the Screener's dated STM timing signals across daily bars from redundant sources. Signals describe state, never destiny — and every page carrying one says so.

Coverage, market by market

🇺🇸 United States

The core book: mega-caps through mid-caps with decision briefs, outlooks, dividend and earnings pages, and FAQs for every ticker — plus 27,000+ head-to-head comparisons like AAPL vs MSFT and NVDA vs AMD.

🇭🇰 Hong Kong

HSI leaders — Tencent, the banking and property complex, A-share dual listings — each with the same four-part workflow and a Thai brief where relevant.

🇨🇳 China A-share

Shanghai and Shenzhen large caps — CATL, Kweichow Moutai, the banks, the solar complex — via multi-source bars that survive single-feed outages.

🇰🇷 Korea

KOSPI and KOSDAQ leaders — Samsung Electronics, SK Hynix, the EV-battery names — scored with the same checklist as US tickers.

🇮🇳 India

NIFTY leaders with daily coverage and the full page set: brief, outlook, dividend, earnings, insider, FAQ.

🇹🇭 ภาษาไทย

บรีฟหุ้น 751 ตัวเป็นภาษาไทยพร้อม hreflang ครบทุกคู่ — ไม่มีคู่แข่งรายใดมีเนื้อหาไทยแม้หน้าเดียว เข้าศูนย์วิจัยภาษาไทย →

Why people distrust stock AI — and what we do differently

  1. "AI picks stocks" hype. Anyone promising predictions is selling hope. We publish frameworks with named break conditions and refuse price targets on principle — every outlook page says framework, not prediction.
  2. Black-box scores. A score you cannot inspect is marketing. Ours maps to a checklist you can run by hand; the calculator shows every input.
  3. Undated data. Numbers without dates are rumors. Every figure on every page carries its snapshot date — and stale pages say so.
  4. Paywalled honesty. The whole method — guides, calculator, FAQs, and 36,000+ research pages — is public without an account. The app adds live scoring on top, not secrets behind a wall.

Inside the workspace

Each tool does one job; the chain does the real one — turning a ticker into a written, testable decision.

Stock Chat — research assembly

Ask "walk me through TSMC" and the workspace assembles the answer from fundamentals, filings context, and scanner state — with the snapshot date on every figure. It answers follow-ups like "what breaks that thesis?" because the thesis is a first-class object, not chat history. Credits: 30/month on the free plan, enough to run several full research sessions.

Fundamentals desk — the scorecard

The Berkshire checklist rendered as scores: returns on capital, moat, 13F honesty, valuation versus ceiling. Scores are comparable across tickers because every name runs the identical checklist — and every input is inspectable down to the snapshot row.

Screener — timing without the noise

Daily multi-source bars across five markets, screened for STM signals. Signals sit behind the workflow by design; when one fires you already know the business, the ceiling, and your break conditions. The public version of every signal is the same one shown on the decision-brief pages.

Labs — validate before you risk

Backtesting for strategies and single-stock theses on the same cited data. Export a Python version of your strategy, iterate, and only then take it near money. Losses in a lab are tuition you can afford.

Numbers behind the research

Research pages published36,000+ (briefs, outlooks, FAQs, comparisons — all dated, all static, all free)
Tickers in the book753 across US / HK / China A / Korea / India
Head-to-head comparisons27,478 — including every same-sector and same-market pair, plus the 34 pairs people actually search
FAQ pages4,000+ with FAQPage schema — no competitor ships any
Thai-language coverage751 ticker briefs + full guides — unrivalled in the category
Data sources8 redundant public feeds (Yahoo, Stooq, Tencent, Eastmoney, Sina, Binance, OKX, MT5)
Free planStock Chat + Screener preview + 30 AI credits/month — no card
Editorial standardsPublished at /about — dated snapshots, no price targets, corrections policy

Where to start, by what you want

Real questions these pages answer

Every one of these is a real, published, dated page — not a promise:

How to read a decision brief

  1. Start at the keyfacts block. Three bullets: what the company is, the dated snapshot (price, signal, uPnL), and the workflow summary. If that's all you read, you still leave with the state of the business.
  2. Check the snapshot table second. Market, sector, last price, engine signal, signal age — each with a source line. This is state, not prophecy, and it says so underneath.
  3. Read the workflow section like a checklist — quality, ceiling, breaks, timing. Each step links to the deeper guide or tool that performs it.
  4. Finish at the FAQ. Three to five questions real searchers ask about this exact ticker, answered from the same snapshot — and mirrored in FAQPage schema so answer engines can quote them accurately.
  5. Then, and only then, open the live brief inside the app, where the same four steps run on current bars instead of the published snapshot.

Ten terms you will meet on every page

The whole site runs on one small vocabulary — learn these ten and every page reads itself.

Business quality

How well a company converts capital into returns over years — scored on the Berkshire checklist before price is even mentioned. Quality first is rule one because a mediocre business at a cheap price is still a mediocre business.

Valuation ceiling

The most you can pay for the earnings power without betting on miracles, computed by the two-stage residual-income model. You then demand a margin of safety below the ceiling, never at it.

Thesis break

A measurable, pre-written event that would falsify your reason for holding: a margin band, a share-loss threshold, a regulatory trigger. Breaks turn feelings into exits.

STM signal

The Screener's dated timing state (LONG / SHORT / WAIT) on daily bars. It is the last input in the workflow — state, not prophecy.

Snapshot date

The day a page's numbers were true. Every figure on this site carries one; quoting a Balance Labs number without its date is quoting it wrong.

Margin of safety

The gap between the ceiling and the price you actually pay. It is the allowance for being wrong — and you will sometimes be wrong.

13F honesty

Our management test: do the filings match the story? Insiders' reported holdings and trades are checkable, so we check them.

Residual income

Earnings above the cost of capital — the engine of the valuation model. Book value today plus discounted future residuals equals intrinsic value.

uPnL

The open profit/loss of the current engine position on a ticker — shown with its signal so you see conviction and result together, dated.

Decision brief

The one-page unit of research here: a ticker turned into a written decision — quality, ceiling, breaks, timing — with its snapshot table and FAQs. 753 tickers have one.

A reading list, by what you want to become

Start with the research, not the trade

Open any market hub, pick a ticker, and read its decision brief — the whole method is public before you ever sign up. When you want the live scored version:

Open Balance Labs — free

Get more Balance Labs in your Google results

Add balancelabs.app as a Preferred Source on Google to see our research in Search and AI Overviews.

Also available: an MCP server for AI-native integrations, research RSS feed, and a machine-readable llms.txt index. Not investment advice.