How We Review Products
How Chord Lab's guides are made today, what is being built to replace them, and what the AI is and isn't allowed to do.
The short version
I don't run a testing lab or a studio, and I won't pretend to. Every guide on this site today was drafted in April 2026 by an AI model from a topic prompt, using only what the model already knew. Since then the drafts have been edited through automated and AI-assisted rewrites that I direct; the November 2026 pass removed the invented scores and prices and the claims that didn't hold up. What I'm building to replace them: a catalog of real products for each category, picks chosen from it by fixed rules weighted for the specific use-case each guide covers, and an AI that writes the prose without the ability to choose, re-rank, or price a single product. No guide has been rebuilt on that method yet.
Where things stand
As of November 2026, no guide has been rebuilt. Until a guide is rebuilt, it shows no product prices and no scores, it carries a note at the top saying how it was made, and its links open Amazon so you can see current prices and availability there.
Who's behind this
Chord Lab is run by Vin Falkiewicz, who has played guitar and drums since age 12 and is a Senior Manager in Technology & Transformation at EY, leading AI Mobilization for EY's Financial Services practice. That's where the approach comes from — data pipelines and tightly-constrained AI systems are the same patterns I work with in production for enterprise clients. That is the approach being built here, for the question of which music gear is actually worth buying.
When this site says “we,” it means me plus the software pipeline described below. There is no editorial team, no test studio, and no luthier on staff — I'd rather tell you that plainly than invent one.
How today's guides were made
Each guide started as an AI draft. A language model was given the guide's topic and asked for a ranked list of real products with specifications, pros and cons, a buying guide and FAQs, written from what it knew rather than from live data. The products were then linked to Amazon search results for their names. Nothing was tested, and no listing was verified.
Since then the drafts have been edited through automated and AI-assisted rewrites, run at my direction, and the checks below now block the claims an AI draft tends to make. A guide that hasn't been rebuilt is still an edited AI draft, and it says so at the top.
Where the data will come from
Each rebuilt guide will start from a catalog of real products, and every product in it will carry its provenance:
- Amazon listing data, through Amazon's own API — used for one thing: confirming that the exact listing a guide links to exists, is in stock, and is the product the guide names. Where you see a price on this site, it will be Amazon's current price with the time it was fetched. This site does not use Amazon customer ratings or review counts, and does not store Amazon's data.
- Manufacturer specifications — used for product identity and the specs that matter for each instrument: wattage and speaker size for amps, pickup configuration and scale length for guitars, key action and polyphony for keyboards. Specs are marketing until something independent confirms them, so they're presented as the manufacturer's claims.
- Independent sources, as they're added — published independent reviews and measurements get onboarded only if they're free to access, attributable, and clear about their methods. Each rebuilt guide will say which sources stand behind it.
How products will get picked
In rebuilt guides, rankings will come from fixed rules applied to the catalog — math over the data, not an AI's opinion:
- Weighted per use-case. A guide for metal players weighs humbuckers, fast necks, and 24 frets; a blues guide weighs single-coils and a vintage voice. The same guitar can rank first in one guide and fourth in another — that's the system working, not a contradiction.
- Cheap won't automatically win. Price will affect a pick only through a guide's budget limit, never as a score.
- Budget limits will be hard limits. A “best under $500” guide won't be able to include a $700 product, and a $12 strap will never be ranked as a guitar. Today's guides have no such check beyond a manual pass, so an over-budget pick found in one is removed rather than excused.
What the AI does — and what it can't
Today, AI drafted the guides you're reading (see above). In rebuilt guides, AI will write the prose and nothing else: a constrained writer will turn the locked picks into readable text. It won't be able to add, drop, reorder, or rename products, and every spec and link will be filled in from the data, not generated by the model. Prices will never appear in the prose at all; when a price is shown, it comes live from Amazon with a timestamp.
Automated checks run before anything publishes, on every guide: common fabricated-authority phrasings (a pretend studio test, an “editorial team,” claims attributed to unnamed users or experts) are blocked, and the Amazon Associates statement, the no-scores rule and the no-product-prices rule are pinned by tests. When I find a guide whose products don't match its title, it comes down; four are down now. I will read every rebuilt guide before it ships.
Scores and ratings
Guides don't show a score or a star rating. Earlier versions carried a “Chord Lab estimate” out of 10 that came from the AI draft; those were removed in November 2026 because nobody had evaluated the products behind them. If a rebuilt guide shows a score in future, it will be computed from the product's specifications against the guide's use-case, with the basis stated. Amazon customer ratings and review counts will never be an input, and never appear on this site.
When we don't publish
- If the products in a guide don't match what it promises — computer keyboards in a guide to musical keyboards, picks above the guide's stated budget, or a model that doesn't appear to exist — it comes down until it's fixed. Four guides are down for those reasons.
- A rebuilt guide will be held if fewer than three products clear the data bar.
What we don't claim
- We don't claim to have physically tested products. When a guide cites a measurement or a review, it credits the source.
- We don't pretend any single source is gospel. Manufacturer specs are marketing until something independent confirms them, and each claim should be attributable.
- We don't pretend the AI is infallible — which is exactly why the plan keeps it out of every decision and behind automated guards.
- The rebuild hasn't shipped yet. Today's guides predate this method and will be replaced, not retrofitted.
Editorial independence
- We do not accept payment for product placement or favorable coverage.
- Commission rates play no part in how picks are chosen. In rebuilt guides, price will affect a pick only through the guide's budget limit; in today's AI drafts, the “Best Value” label was the model's own judgment.
- Every pick lists drawbacks as well as strengths. A recommendation that fits nobody in particular fits nobody.
How we make money
Chord Lab is reader-supported. As an Amazon Associate I earn from qualifying purchases.When you click a “See price on Amazon” link and buy something, I may earn a small commission at no additional cost to you. That revenue funds the data the method above depends on.
Keeping guides current
Every guide shows the date it was last updated. Once the catalog method is live, each rebuilt guide's links will be re-checked against live Amazon listings daily, and when the data changes the guide will be re-run through the same pipeline — not hand-patched.
Spot an error? Question the method?
Reach me directly at contact@chord-lab.com. If a recommendation is wrong, I want to know. If the method doesn't hold up under scrutiny, I want to fix it. My LinkedIn is public.
— Vin