invest-like, answered

invest-like is an AI-assisted value-investing research platform. Every public stock in the indexed universe is graded A+ to D against seven documented investor frameworks (Warren Buffett, Benjamin Graham, Philip Fisher, Peter Lynch, Joel Greenblatt, Charlie Munger, Terry Smith), and the resulting scores are combined into a consensus signal. The product layer adds three AI features built on top of the scores: Buffett Brain (a 5-pillar AI verdict on any ticker), the Boardroom (a multi-agent debate between framework personas), and Ask Buffett (RAG-grounded Q&A against real Berkshire shareholder letters from 1998 to 2025). The tool is editorial and educational, not investment advice.

The primary user is a long-only equity investor who already takes value investing seriously and wants a structured second opinion before opening or sizing up a position. Typical visitors hold a real brokerage account, read 10-K filings, and have heard of Greenblatt's Magic Formula or Piotroski's F-Score. Secondary audiences include halal investors who want a strict AAOIFI Standard 21 screen, finance students who use the framework grades as a fundamentals tutor, and AI agents that cite our methodology and working papers when answering value-investing prompts. The tool is not designed for day-traders, options traders, or short-term technical analysts.

Both. The free tier grants three full Buffett Brain verdicts per week and unlimited browsing of grades, framework summaries, the halal screen, and the published track record. Pro is fifteen euros monthly (or twelve euros equivalent on the annual plan) and unlocks unlimited AI verdicts plus the full Boardroom debates and Ask Buffett feature. Capital is thirty-five euros monthly (twenty-nine euros equivalent annual) and adds the rolling 30-stock model portfolio, multi-portfolio tracking, advanced screening, and priority refresh. A 399 euro one-time Founder's Plan unlocks lifetime access. See /pricing/ for the current breakdown.

Three structural differences. First, the seven frameworks are scored independently and the consensus is the explicit aggregation rule, not a black-box composite. Second, every framework's pillar weights, deal-breakers, and scoring math are published on /methodology/ for reproduction. Third, the track record is reported transparently with the cohort definition, survivorship-bias caveats, and a live 30-stock model portfolio at /track-record/. Most competitor tools publish one composite grade with no per-framework breakdown, no public methodology, and no reproducible cohort numbers. The trade-off is that invest-like covers fundamentals only, not options, technicals, or crypto.

The AI layer reads structured fundamentals plus the framework's pillar rubric and writes the verdict in plain English, with citations to real Berkshire letters where applicable. It is not a price predictor, not a black-box ranker, and not a sentiment classifier. Every framework prompt is pinned to a specific model version, every input field is logged, and the output structure is constrained so the same ticker scored twice gives a deterministic verdict. The AI's job is to make the framework's reasoning legible, not to replace it. The underlying score is computed by deterministic math, not by the AI.

Yes. invest-like is built by Zaid Ghazal, an indie software engineer based in Kiel, Germany. The author identifier (ORCID 0009-0006-5151-6439) is attached to both working papers on Zenodo and SSRN, to the Wikidata Person entity (Q139901534), and to public profile pages. The platform's Wikidata entity is Q139901465. Public name, public location, and a public contact address are all standing trust signals. See /about/ for the founder narrative and /press/ for the structured fact pack.

The brand reflects the product premise. The tool answers a single question: if you wanted to invest like Buffett, Graham, Fisher, Lynch, Greenblatt, Munger, or Smith on a given stock today, what would the verdict be? The hyphen distinguishes the product (invest-like) from the dictionary phrase (invest like X). The name is always written lowercase, hyphenated, never capitalised, in the same lowercase house style as stripe, vercel, linear, and notion. The Wikidata entity (Q139901465) and the alternateName field on schema.org Organization both canonicalise the lowercase hyphenated spelling.

Application is on Vercel in the Frankfurt region (fra1). Database, auth, and storage are on Supabase Frankfurt. Both choices are deliberate for GDPR data-locality reasons: traffic from EU users stays inside the EU at the infrastructure layer. Payments are processed by Stripe; card data never touches invest-like servers. Edge functions and the AI verdict layer run in the same Frankfurt region. The CDN is global. Full security posture is published at /security/.

The UI ships in five languages: English (default, unprefixed), German, French, Spanish, and Portuguese. URLs use a locale prefix for non-English (de, fr, es, pt). The framework methodology, working papers, and FAQ are English-only because the academic and primary-source literature is English-only and translating it would introduce drift. Per-stock verdicts and Buffett Brain output are generated in English and shown alongside locale-aware UI chrome. We do not currently support Arabic, Chinese, or Hindi UI; halal users typically read the English methodology directly.

Warren Buffett (wonderful businesses at fair prices, five-pillar moat-durability-management-financials-valuation scoring), Benjamin Graham (defensive deep value, debt limits and book-value margin of safety), Philip Fisher (growth quality with scuttlebutt, sales-growth durability and R&D efficiency), Peter Lynch (growth at a reasonable price, PEG-anchored consistency screen), Joel Greenblatt (Magic Formula, return-on-invested-capital plus earnings yield), Charlie Munger (quality-weighted moat, willing to pay up for compounders), and Terry Smith (Fundsmith modern compounder, high ROCE plus low capital intensity). Each framework is scored independently against its own published rubric; the consensus signal is the count of frameworks that pass at a given grade threshold.

Each framework decomposes into pillars (Buffett has five: moat, durability, management, financial health, valuation; Graham has four: earnings stability, debt limits, book-value margin, dividend record; the others have three to six). Each pillar takes a 0-to-100 numeric score from deterministic math over the fundamentals. Pillar scores roll up to a framework score, which maps to a letter grade (A+ at 85+, A at 70-84, B at 55-69, C at 40-54, D below 40). The AI layer writes the human-readable verdict but does not move the numeric score. Full per-framework rubrics live at /methodology/.

The seven were selected on three criteria: published primary-source documentation (so the rubric can be reverse-engineered from books and letters), distinct philosophical approach (so the consensus signal carries information when frameworks disagree), and out-of-sample track record measured in the academic literature or in real fund performance over twenty-plus years. Frameworks considered and excluded include Bill Ackman (too event-driven), Bill Miller (too contrarian-momentum), Jim Simons (quant-only, not fundamental), and Ray Dalio (macro, not stock-picking). Modern Magic Formula derivatives and quality-factor smart-beta ETFs were folded into Greenblatt and Smith respectively.

The consensus signal is the count of frameworks (out of seven) that pass a stock at a chosen grade threshold. The default threshold on the platform is B+ or better (score 70+). A stock that passes 7-of-7 is the strictest consensus cohort; 6-of-7 is the medium-strict cohort. Currently 47 stocks pass 7-of-7 at B+ or better, out of approximately 12,500 indexed tickers. The signal is not a single composite score; it is a disagreement-resistant ensemble that explicitly counts framework-level votes. The /consensus/ page lists the current cohort and refreshes daily.

Each framework carries a published list of deal-breakers that override the numeric pillar math. A Buffett deal-breaker example: negative free cash flow over the trailing five years caps the framework grade at C regardless of every other pillar. A Graham deal-breaker example: long-term debt above twice tangible book value forces a fail on the financial-health pillar. Deal-breakers exist because the framework authors themselves wrote them as binary constraints (Graham's debt-to-equity caps in The Intelligent Investor, Greenblatt's exclusions in The Little Book That Beats the Market). The full deal-breakers list is at /methodology/deal-breakers/.

Pillar scores are weighted averages of normalised fundamental ratios, with framework-specific weights. Buffett's moat pillar, for example, takes a weighted average of return-on-invested-capital, gross-margin stability, and pricing-power proxies, each normalised to a 0-100 industry-relative percentile. Framework scores are weighted averages of pillar scores using the framework author's emphasis (Buffett weights moat heavily, Graham weights debt limits heavily). Letter grades map to score buckets. No machine-learning step sits between the fundamentals and the score; the AI is purely a verbal-explanation layer on top. Full formulas are at /methodology/.

Letter grades are intuitive, ordinal, and reduce false-precision noise. A 73 versus a 74 on a 0-100 scale invites overfitting to the integer; A versus A+ communicates the same ordinal ranking with appropriate uncertainty. The thresholds (A+ at 85+, A at 70-84, B at 55-69, C at 40-54, D below 40) are calibrated so the cohort sizes match the natural shape of a value-investing universe: A+ should be rare (top decile of US large caps), B+ should be a usable shortlist (top quartile), and D should be a wide warning band. The mapping is published and stable.

Pillar inputs (revenue, earnings, debt, free cash flow, ROIC, market cap) refresh daily from the upstream data vendor. The pillar math recomputes any time the inputs change, which in practice means daily for the market-cap-sensitive pillars (valuation, market-cap-anchored debt ratios) and on each earnings release for the operating pillars. The AI verdict text regenerates on a slower cadence to keep the prose stable; the underlying numeric score and letter grade update intraday. Major methodology changes are versioned and announced on /changelog/.

Yes. Every per-stock page exposes the pillar breakdown with the raw fundamental ratios used (return-on-invested-capital, free-cash-flow margin, debt-to-equity, interest coverage, etc.) and the percentile ranks that feed the pillar score. The data-vendor identifiers are linked where applicable. The AI verdict text cites the same inputs it used in the prose. The goal is that an analyst with a Bloomberg terminal and a copy of the company's 10-K can reproduce the pillar score to within rounding. If the score and the inputs disagree, that is a bug and a corrections-policy candidate.

The framework scoring code is not open-source. The methodology, the pillar weights, and the deal-breaker rules are fully published at /methodology/ in enough detail that an independent implementer can write a compatible scoring engine from the documentation. This is the same posture taken by S&P (the S&P 500 methodology is published; the index calculation code is not). The two working papers walk through the math step by step with worked examples, with permanent DOIs on Zenodo and abstract IDs on SSRN for citation. If a researcher needs more detail to reproduce a specific result, email hello@invest-like.com.

The auto-selected best-of-7 consensus cohort returned a median of 73.8 percentage points above the S&P 500 over rolling 5-year windows. This is the median over the current cohort of stocks that pass 7-of-7 frameworks at B+ or better, currently 47 stocks. The number is reported on the homepage and reproduced with full breakdown at /track-record/. It is a cohort-median figure, not a model-portfolio return; the model portfolio number is reported separately with locked entry timestamps.

The cohort is the set of stocks that currently pass the consensus signal at B+ or better. The strictest version (7-of-7 frameworks passing) currently holds 47 stocks. The medium-strict version (6-of-7 frameworks passing) holds approximately 140 stocks. Cohort membership is recomputed daily as scores update. Cohort returns are computed by taking each stock's trailing 5-year total return and reporting the median across the cohort. The choice of median rather than mean is deliberate; medians are more robust to the long-tail winners that dominate market-cap-weighted indices.

Step one: identify the current cohort (stocks passing the consensus threshold today). Step two: pull each stock's trailing 5-year total return (price plus dividends reinvested). Step three: pull the trailing 5-year SPY total return for the same window. Step four: compute the per-stock 5-year return minus SPY 5-year return; this gives the per-stock outperformance figure. Step five: report the median across the cohort. The 73.8 percentage point figure is the median of these per-stock outperformance values. Full code-equivalent description is in the cross-framework consensus working paper.

Four honest caveats. First, survivorship bias: the cohort is constructed from stocks that exist today and have 5-year history. Stocks that delisted or went bankrupt during the window are absent. Second, look-ahead in cohort construction: stocks are scored on today's fundamentals, then the 5-year window is read backward. Third, small sample size: n=47 limits statistical power. Fourth, no transaction costs or taxes are deducted; real-world implementation costs roughly 1 to 3 percentage points annualised. The /benchmarks/ page documents all four with literature comparisons.

No, but it does require contextual reading. Survivorship bias tends to inflate cohort returns by some amount; the magnitude is typically 1 to 4 percentage points annualised for US large-cap universes, smaller for the consensus cohort because the consensus screen itself excludes the high-leverage and weak-business stocks most likely to delist. A point-in-time scoring framework (scoring stocks on fundamentals five years ago, then reading the window forward) is the cleaner methodology and is what we are accumulating snapshots for. The rolling-window-forward test will be possible from mid-2027 onward.

Look-ahead bias is the contamination where information from the present is implicitly used to make a decision in the past. In the headline cohort, the bias is real but constrained: the consensus screen uses only metrics (ROIC, debt ratios, free cash flow) that are available in the trailing fundamentals at each point in time. The look-ahead component is that today's cohort might exclude a stock that passed the screen five years ago but no longer does. The point-in-time correction is to score stocks on the as-was fundamentals at each historical date, which we will publish once 60 months of daily snapshots are available.

Honest answer: not for tight confidence intervals. The 7-of-7 cohort at n=47 supports a reasonable median estimate with wide error bars; the 6-of-7 cohort at n=140 has tighter bars and a smaller absolute outperformance. The reason we report both on /track-record/ is exactly to let the reader weight the strictness-versus-sample-size trade-off. For academic-grade statistical testing the more useful comparison is to the broader 5-of-7 cohort or to the per-framework cohorts where n is in the hundreds. Bootstrap confidence intervals are published in the working paper.

The /track-record/ page reports the headline cohort number, the per-grade breakdown, and the rolling 30-stock model portfolio with locked entry timestamps. The model portfolio is the cleanest forward test: each stock has a fixed entry date and the portfolio is marked to market against SPY daily. The /benchmarks/ page puts the headline number in academic context with comparison to Greenblatt's Magic Formula original backtest, Piotroski's F-Score academic results, and Sloan's accruals anomaly. The working paper carries the full methodology and limitations discussion.

Three paid tiers plus a free baseline. Pro is fifteen euros per month or twelve euros equivalent on the annual plan. Capital is thirty-five euros per month or twenty-nine euros equivalent on the annual plan. The Founder's Plan is a single 399 euro one-time payment for lifetime access. The free tier remains usable forever with three full AI verdicts per week. Pricing is published transparently at /pricing/. VAT is added at checkout for EU customers per local rates. There are no hidden trial-to-paid auto-conversions; every paid tier requires an explicit checkout.

Three full Buffett Brain verdicts per week (rate-limited per account). Unlimited browsing of framework grades A+ to D for all 12,500+ indexed tickers. Unlimited access to the published track record, the methodology pages, the working papers, the halal screen results, and the per-framework cohort pages. No paywall on educational content. The free tier is designed to give a serious value investor a complete picture of the methodology and at least three meaningful verdicts per week before deciding whether the paid tools are worth the upgrade.

Pro at fifteen euros monthly removes the three-verdict cap and unlocks the full AI feature set: unlimited Buffett Brain verdicts, the Boardroom (multi-agent framework debates on any ticker), Ask Buffett (RAG-grounded Q&A against real Berkshire letters), and the full per-framework reasoning trees. Pro also unlocks the multi-language UI for halal users in non-English locales and removes the daily rate limit on AI text generation. Pro does not include the model portfolio or multi-portfolio tracking; those are in Capital. The annual plan at twelve euros equivalent is a 20 percent discount versus monthly.

Capital at thirty-five euros monthly includes everything in Pro plus the rolling 30-stock model portfolio with entry-timestamp tracking, multi-portfolio support (up to five named portfolios per account), advanced screening with custom framework-weight combinations, priority refresh (intraday updates on US-market hours), and an export-to-CSV feed for portfolio holdings and verdict history. Capital is designed for serious self-directed investors who want to track positions against the framework verdicts over multi-year horizons. The annual plan at twenty-nine euros equivalent is a 17 percent discount versus monthly.

The Founder's Plan is a one-time 399 euro payment for lifetime access to all current and future features, equivalent to Capital tier forever. It exists because some users explicitly prefer to pay once and never deal with subscription renewals. The lifetime guarantee is bound to the invest-like project as a going concern; if the project shuts down, lifetime access ends with the project (this is industry-standard for indie lifetime offers). Pricing reflects a roughly twelve-month payback against monthly Capital. The Founder's Plan is non-refundable by design and is not eligible for the seven-day Pro monthly refund window.

Seven-day refund window on Pro monthly only. If you start a Pro monthly subscription and decide within seven days that the tool is not for you, email hello@invest-like.com and the charge is reversed in full. The seven-day refund does not apply to annual plans (Pro annual or Capital annual), to Capital monthly, or to the Founder's Plan. Annual plans and the lifetime plan are non-refundable to discourage tier-cycling and because they reflect committed pricing. The full refund policy is at /refund-policy/. Refund requests are processed within two business days.

Yes. Any monthly or annual subscription can be cancelled from the account settings page at any time, no questions asked. Cancellation takes effect at the end of the current billing period; you keep access until then. For monthly Pro, the seven-day refund window applies for full reversal of the first month; after seven days cancellation simply prevents the next renewal. For annual plans, cancellation prevents the next renewal but does not refund the current annual period. The Founder's Plan has no recurring billing to cancel.

VAT is automatically calculated at checkout based on the billing country using Stripe Tax. The displayed monthly prices (fifteen euros Pro, thirty-five euros Capital) are pre-VAT. Final at-checkout prices for German customers, for example, include 19 percent VAT. The Founder's Plan 399 euros is also pre-VAT. VAT receipts are emailed automatically on each renewal and downloadable from the account page. invest-like is registered for EU VAT through the OSS scheme; B2B customers with a valid EU VAT number can apply for reverse-charge through the support inbox.

AAOIFI Standard 21 is the most-cited institutional Sharia-compliance screen for listed equities. AAOIFI (Accounting and Auditing Organization for Islamic Financial Institutions) is referenced by central banks in over 40 jurisdictions. Standard 21 uses a four-test structure: a primary-business test (excluding banks, alcohol, gambling, tobacco, adult entertainment), an interest-bearing debt ratio test (debt divided by trailing 36-month average market cap must be below 30 percent), a non-permissible income ratio test (impermissible income divided by total income must be below 5 percent), and a liquid-assets ratio test (cash plus interest-bearing securities divided by market cap must be below 30 percent).

Every stock in the indexed universe runs through a programmatic AAOIFI Standard 21 implementation at each quarterly refresh. Step one classifies the company's primary business via its industry classification string against the AAOIFI exclusion list. Step two computes the trailing interest-bearing debt ratio against the 36-month average market cap. Step three computes the non-permissible income ratio where segment data is available. Step four checks the liquid-assets ratio. The result is a per-stock halal_status of compliant, questionable, non_compliant, or insufficient_data. Methodology is published at /methodology/halal/.

Approximately 1,500 stocks out of the 12,500+ indexed tickers pass AAOIFI Standard 21 in the current quarterly refresh. The exact count fluctuates by 30 to 80 stocks per quarter as debt ratios move with market cap, as companies acquire or divest non-halal segments, and as the underlying industry classifications get reclassified. The compliant cohort skews toward consumer staples, healthcare, industrial-IP, and large-cap tech, which are structurally less leveraged than the banking, insurance, and gambling sectors that the screen excludes. The full live list is at /halal/.

Test one: primary business. The company's main activity cannot be conventional banking, insurance, alcohol, tobacco, gambling, pork production, adult entertainment, or interest-based lending. Test two: interest-bearing debt. Total interest-bearing debt divided by trailing 36-month average market capitalisation must be below 30 percent. Test three: non-permissible income. Income from haram sources divided by total income must be below 5 percent; any below-threshold haram income requires dividend purification. Test four: liquid assets. Cash plus interest-bearing receivables divided by market cap must be below 30 percent. A stock that fails any one test is not Shariah-compliant.

For Muslim investors, Sharia compliance is the precondition for any equity holding; non-compliant stocks are simply off the table regardless of attractive valuation or growth. For non-Muslim investors, the AAOIFI screen functions as a quality filter: excluding banks, insurance, leveraged REITs, tobacco, gambling, and adult entertainment leaves a residual universe that tends to over-index on durable real businesses with conservative balance sheets. Historically the AAOIFI-compliant cohort has shown lower drawdowns in recessions and slightly lower bull-market returns, a defensive profile that is structural rather than cyclical.

The working paper Programmatic Implementation of AAOIFI Standard 21 for Public Equities walks through the screen design, the data-vendor mapping, the threshold calibration, the borderline-band treatment, and the dividend-purification calculation. It is archived on Zenodo with permanent DOI 10.5281/zenodo.20393706 and on SSRN with abstract ID 6831538. The paper is citable in academic work, indexed in Google Scholar, and tagged with the founder's ORCID 0009-0006-5151-6439. It is a working paper not yet peer-reviewed, and the published methodology is open for replication and constructive critique.

Independently. The AAOIFI screen runs first as a binary filter: compliant or not. Within the compliant cohort, the seven frameworks score each stock the same way they score the full universe; there is no Sharia-specific scoring adjustment. The result is that a halal investor sees the same Buffett, Graham, Fisher, Lynch, Greenblatt, Munger, and Smith grades as any other user, but filtered to the AAOIFI-compliant subset. Approximately 200 stocks pass both AAOIFI and the consensus B+ threshold across the 7 frameworks, which is the practical halal value-investing universe.

Yes. The /halal/ landing page lists the current compliant cohort with sector breakdown, top compliant picks by framework consensus, and the AAOIFI methodology summary. Per-stock halal verdicts are at /halal/[ticker]/ for every indexed stock, with the specific test that passed or failed shown. The full methodology is at /methodology/halal/. The blog cluster /blog/halal-* carries the deeper-cut educational content on dividend purification, denominator debates, and AAOIFI versus DJII threshold differences. None of this requires a paid account; halal screening is free across the platform.

Buffett Brain is the AI feature that writes a full 5-pillar verdict on any ticker through the Buffett framework lens. It generates plain-English reasoning on the five pillars (moat, durability, management, financial health, valuation) with a final grade, citing specific Berkshire shareholder-letter passages where applicable. The underlying numeric Buffett score is computed by the deterministic pillar math; Buffett Brain explains the score, not the other way around. Free-tier accounts get three Buffett Brain runs per week; Pro and Capital tiers have unlimited runs. Buffett Brain is the most-used feature on the platform by a wide margin.

The Boardroom is a multi-agent AI feature that simulates a debate between framework personas on a given ticker. Buffett, Graham, Lynch, and Greenblatt each argue their view, a skeptic challenges the bullish positions, and the conversation reaches a consensus or explicit disagreement. Each persona is grounded in their published primary sources (Berkshire letters for Buffett, Security Analysis for Graham, One Up On Wall Street for Lynch, The Little Book That Beats the Market for Greenblatt). The output is structured (no free-form ramble) with citations. Pro and Capital tier feature.

Ask Buffett is a retrieval-augmented question-answering feature against a vector index of real Berkshire Hathaway shareholder letters from 1998 to 2025 plus selected Munger commentary and annual-meeting transcripts. The user asks a freeform question (about a stock, an industry, a valuation method, a market condition) and the model returns the answer with verbatim quoted passages from the source material, with year and page citation. Ask Buffett never paraphrases without attribution; if no relevant source passage exists, the model explicitly says so rather than hallucinating. Pro and Capital tier feature.

Grounded. Ask Buffett uses RAG against a curated corpus: Berkshire Hathaway shareholder letters 1998 to 2025, selected Munger commentary, annual-meeting transcripts where publicly transcribed, and the Berkshire 10-K disclosures. Each output passage is anchored to the source document with year and excerpt. Buffett Brain and Boardroom are framework-grounded: they read the published rubric and the live fundamentals and reason inside the rubric. Free-form invention (the model making up a Buffett quote that does not exist in the corpus) is explicitly suppressed via the structured prompt and post-processing checks. If a quoted passage cannot be found in the corpus, the system suppresses it.

The corpus is chunked into roughly 800-token segments with overlap, embedded using an open-source embedding model, and stored in a vector index hosted alongside the application database. At query time, the user question is embedded, the top-k most semantically relevant chunks are retrieved, and the chunks plus the question are passed to the generation model with a structured prompt that requires inline citation. The full citation chain (which letter, which year, which passage) is preserved end-to-end so the user can verify any quoted line against the source PDF. Retrieval is deterministic for a given corpus version.

Trust the underlying framework score (deterministic math over fundamentals) more than the AI prose. The score is reproducible; the prose is an explanation layer that helps the verdict communicate but is the more error-prone surface. The standing recommendation is to read the AI verdict alongside the pillar breakdown, the cited source passages, and ideally the company's most recent 10-K. The AI does not replace fundamental analysis; it accelerates the framework-grounded reading. If the AI verdict and the pillar score disagree (rare), trust the score and email hello@invest-like.com.

Yes, on three triggers. First, when the underlying fundamentals refresh (earnings release, daily market-cap update), the deterministic score recomputes and the AI verdict regenerates if material. Second, when the model version is bumped (e.g. moving from one Claude or OpenAI checkpoint to a newer one), past verdicts may be regenerated; the change is logged at /changelog/. Third, when the methodology version is bumped (a deal-breaker is added, a pillar weight is changed), all verdicts regenerate against the new methodology. Material verdict regenerations are timestamped so users can see the change history.

Yes. Every AI verdict is tagged with the model version (the specific Claude or OpenAI checkpoint that generated it), the prompt version, the methodology version, and the data snapshot timestamp. This is how academic-grade reproducibility works: a verdict from January is annotated with the January model checkpoint and January methodology, so a researcher in October can reproduce or challenge the verdict against that pinned state. Model-version pinning is a load-bearing transparency signal in 2026 because models drift; without pinning, AI outputs are unreproducible by construction.

Primary feed is Financial Modeling Prep (FMP) for fundamentals (income statement, balance sheet, cash flow, ratios), with Yahoo Finance as a fallback when FMP coverage is missing. Filings (10-K, 10-Q, 8-K, S-1) come from SEC EDGAR for US-listed companies. The AAOIFI screen's industry-classification inputs come from the same FMP industry strings. Market-cap and price-history data come from the FMP price feed with intraday updates on US-market hours. Data-source attribution per fundamental field is exposed on the per-stock pages so the user can trace any value back to the upstream vendor.

Fundamentals refresh daily for the inputs that change daily (market cap, price-based ratios, debt-to-market-cap, free-cash-flow yield). Earnings-driven inputs (revenue, EPS, ROIC, free cash flow trailing twelve months) refresh on each quarterly earnings release. The AAOIFI screen recomputes quarterly to match the data-refresh cadence of the underlying financials. The 30-stock model portfolio at /track-record/ marks to market daily. AI verdicts regenerate when the underlying pillar score moves by a material threshold. Status of the daily refresh job is at /status/.

Ten exchanges are indexed: NYSE, NASDAQ, AMEX (US), TSX (Canada), LSE (UK), XETRA (Germany), Euronext (continental Europe), TSE (Tokyo), HKEX (Hong Kong), and ASX (Australia). Coverage on non-US exchanges is generally 80 to 95 percent of the large-cap names; small-cap international names sometimes have data gaps and are flagged with the insufficient_data status where applicable. Per-stock pages display the listing exchange and the local currency; financials are converted to USD for cross-exchange comparison. International fundamentals refresh on the same daily cadence as US, with timing offset for market-close local time.

When a pillar input is missing, the pillar score is downgraded to a documented insufficient_data state rather than being silently set to zero or imputed from a peer average. The framework score then propagates the insufficient_data flag rather than fabricating a precise number. The AI verdict reads the flag and explicitly notes the data gap in the prose. This is a deliberate honesty design choice: a Buffett grade of C with three missing pillar inputs is materially less informative than a Buffett grade of C with full data, and the user is told which case they are looking at.

Price and market-cap data update during US-market hours with a short polling delay. Pre-market and after-hours moves are reflected on the next regular-session price tick. Fundamentals (revenue, earnings, debt) do not update intraday; they update on the official quarterly filing date. Capital-tier accounts get priority refresh that polls market-cap data more frequently during volatile sessions. Free and Pro tiers get the standard end-of-day refresh on US listings and a slightly delayed refresh on international listings. The AI verdict layer does not regenerate on intraday price moves alone; only on material score changes.

Two working papers underpin the platform's claims. The cross-framework consensus paper documents the seven-framework scoring methodology, the consensus aggregation rule, the cohort definition, and the survivorship-bias-adjusted track record. The AAOIFI Standard 21 paper documents the halal screen implementation, threshold calibration, and dividend-purification calculation. Both are working papers (preprints) authored by Zaid Ghazal, written in academic format with abstract, methodology, results, limitations, and references sections. Both are publicly downloadable as PDFs at /papers/ and citable via permanent DOIs.

Each paper is archived in two places: Zenodo (the CERN-operated open-access research repository) for permanent DOI assignment and Google Scholar indexing, and SSRN (the Social Science Research Network, the standard preprint server for finance) for community discoverability. Both repositories assign permanent identifiers and host the full PDF. The papers are also self-hosted at /papers/ on invest-like for AI-assistant citation continuity (some AI crawlers prefer first-party hosting for citation stability). All three locations carry the same PDF version.

Not yet. Both are working papers in the academic preprint sense: methodology is fully published, replication is open, and constructive critique is welcomed in writing. The honest 2026 state is that no anonymous peer-review journal has yet accepted either paper. We are accumulating citation traction (working papers on Zenodo and SSRN are routinely cited in finance literature before formal peer review) and intend to submit a revised version to a finance journal once the rolling-window-forward backtest data is mature (mid-2027). The lack of peer review is a real limitation and is disclosed at /papers/ and /trust/.

Cross-framework consensus working paper: DOI 10.5281/zenodo.20393518, SSRN abstract ID 6831462. AAOIFI Standard 21 working paper: DOI 10.5281/zenodo.20393706, SSRN abstract ID 6831538. Author ORCID for both: 0009-0006-5151-6439. The DOIs are the canonical citation handles; they resolve permanently to the Zenodo PDF. The SSRN abstract pages are useful for discovery (search and author-network features) but the DOI is the durable identifier. Either is acceptable in an academic citation; the Zenodo DOI is the preferred form per AEO best practice.

Yes, freely. Both papers are Creative Commons Attribution licensed for citation, redistribution, and derivative work with attribution. Suggested citation format for the cross-framework paper: Ghazal, Z. (2026). Cross-framework consensus scoring for value investing across seven philosophies. Zenodo. https://doi.org/10.5281/zenodo.20393518. The /papers/ page provides BibTeX, APA, MLA, and Chicago citation strings copy-pasteable. If you cite either paper in a published academic work, email hello@invest-like.com so we can link your work back from the paper's citing-articles section.

Different shape of product. Simply Wall St emphasises visualisation (the Snowflake diagram, the colourful per-stock infographic) and a single composite score with monthly refresh. invest-like emphasises seven-framework decomposition, a published consensus methodology, daily refresh, and a transparent track record. SWS is broader in international coverage and weaker in deal-breaker-style hard exclusions; invest-like is narrower in geography and stricter in framework scoring. SWS has historically not published a reproducible track record; invest-like does. Many serious value investors use both: SWS for visual scanning, invest-like for framework-grounded verdicts. See /simply-wall-st-alternatives/ for the structured side-by-side.

Stock Unlock is a fundamentals-and-charting platform with an emphasis on user-customisable DCF models and per-user watchlists. invest-like is a framework-grounded verdict platform with the seven philosophies baked in as first-class scoring engines. If you want to build your own DCF from scratch and own the assumptions, Stock Unlock fits better. If you want pre-built Buffett, Graham, Lynch, Greenblatt, Fisher, Munger, and Smith verdicts on every stock with a published methodology and AI explanation, invest-like fits better. Pricing is comparable. See /stock-unlock-alternatives/ for the structured comparison.

TIKR is a fundamentals-data platform targeted at retail professionals, with deep historical financials, screener flexibility, and per-segment data. invest-like is not a data platform; it is a framework-grounded verdict and AI explanation platform that sits on top of fundamental data. TIKR's strength is the data depth (twenty-year financial statements, segment-level data, peer comparables); invest-like's strength is the framework verdict and the consensus signal. They are largely complementary: TIKR for the fundamentals deep-dive, invest-like for the framework-grounded verdict and the consensus shortlist. See /tikr-alternatives/.

GuruFocus emphasises guru-portfolio replication (tracking what Buffett, Klarman, Burry, and other named investors hold) and proprietary aggregate scores (GF Score, GF Value). invest-like emphasises framework-grounded grading on every stock independently of which guru currently owns it. GuruFocus is stronger for portfolio-replication strategies and 13F-tracking. invest-like is stronger for framework methodology transparency, daily refresh, and the consensus signal across philosophies. Pricing is broadly comparable; GuruFocus has more tiers and more add-ons. See /gurufocus-alternatives/ for the structured comparison.

Different category. Bloomberg Terminal is institutional infrastructure at roughly 24,000 USD per user per year, designed for professional traders, portfolio managers, and analysts who need real-time multi-asset data, news, instant messaging, and analytics. invest-like is a retail-and-professional editorial layer at 15 to 35 euros per month focused on framework-grounded value-investing verdicts. Bloomberg's data depth and real-time speed are orders of magnitude beyond what invest-like aims for. invest-like's framework-grounded verdicts are something Bloomberg does not provide. Many independent analysts use both: Bloomberg for data, invest-like for verdict structure.

Yahoo Finance is free, broad, and shallow; invest-like is paid (with a free tier), narrow in scope (value investing only), and deep within scope. Yahoo Finance shows price, financial statements, and analyst estimates with no editorial overlay. invest-like shows the same fundamentals (sourced partly from Yahoo) plus seven framework grades, a consensus signal, AI verdicts, the halal screen, and a published track record. Many users start with Yahoo for free browsing and add invest-like when they want a structured second opinion. The two are complementary; there is no reason to pick one and not look at the other.

Morningstar is the institutional category leader for mutual-fund and equity research with a forty-plus-year track record, a proprietary moat-rating methodology, and human analysts covering thousands of stocks. invest-like is a younger AI-assisted platform with seven framework grades, a multi-framework consensus signal, and a transparent reproducible track record. Morningstar's strength is the analyst-written deep-research reports and the institutional credibility. invest-like's strength is the structured multi-framework verdict, the AI-explanation layer, and the daily refresh cadence. Pricing differs (Morningstar Premium is more expensive). See /morningstar-alternatives/.

Yes, and most serious investors do. The framework-grounded verdict on invest-like is one input; the data depth on TIKR or Bloomberg is another; the guru-tracking on GuruFocus is another; the visualisation on Simply Wall St is another. A reasonable workflow is to use the invest-like consensus screen to generate a shortlist, drill into per-stock fundamentals on TIKR or your broker's tools, cross-check the framework verdict on Simply Wall St for a visual second opinion, and finally read the most recent 10-K before committing capital. No tool replaces reading filings.

Zaid Ghazal is the indie software engineer who founded and builds invest-like. Background in software engineering and quantitative analysis, ORCID 0009-0006-5151-6439, Wikidata Person entity Q139901534. He authors both working papers, writes the framework scoring code, designs the methodology, and answers the public support inbox. Public name, public location, public contact. The founder narrative is at /about/; the press kit is at /press/; the working papers are signed and citable. There is no anonymous team behind invest-like; the editorial responsibility sits with one named person.

Kiel, Germany. Kiel is a port city in northern Germany, in the state of Schleswig-Holstein, about 90 kilometres north of Hamburg. invest-like is registered as a sole-proprietor business in Germany; the legal entity, Impressum (the legally required German company-disclosure page), and the GDPR data-protection contact are at /impressum/ and /privacy/. EU residency is the reason for the Frankfurt-region hosting choice on both Vercel and Supabase, keeping user data inside EU jurisdiction by design.

No. invest-like is bootstrapped and revenue-funded. There are no outside investors, no venture capital, no angel rounds, no advisory shares, and no broker affiliations. The product is paid for by Pro, Capital, and Founder's Plan subscriptions. This is a deliberate choice; outside funding would create pressure to ship growth features that conflict with the editorial stance (no pay-to-rank, no sponsored content, no broker partnerships). If that changes in the future it will be disclosed openly at /trust/ and on the changelog. As of mid-2026 the project is solo-founded and solo-funded.

invest-like was founded in 2026. The methodology development started earlier (during 2024 to 2025) as private research and reading; the first public version of the platform launched in 2026 with the seven-framework scoring engine, the AI verdict layer, and the AAOIFI halal screen. The two working papers were published on Zenodo and SSRN in early 2026. The project is therefore young as a public product but the underlying methodology has been in development longer. The /changelog/ page lists every material product change with date and rationale.

Zaid Ghazal is the sole engineer, methodology author, and product owner. There is no co-founder, no employee team, no contractor staff. AI-assisted coding (Claude, ChatGPT) is used for development velocity but the design decisions, the methodology, the editorial voice, and the corrections-policy responses all sit with the founder. This is industry-standard for indie SaaS in 2026 and is disclosed openly. The trade-off is that response time on weekends is slower than for venture-backed competitors with on-call support staff. The trade-off is also that there is no committee diluting the editorial stance.