Method: How This Desk Tests and What It Will Not Say
QA Ground is a testing desk for trading automation. It publishes procedures rather than verdicts, and this page states how those procedures are produced, what the desk uses as evidence, and the categories of content it does not publish at all.
QA Ground exists to answer one question in a repeatable way: how do you find out whether a piece of trading automation does what it claims, before it handles an amount you would miss. Everything published here is written so a reader can run it, and everything that cannot be run is marked as an open question rather than presented as a conclusion.
What the desk covers
The subject is automation that signs transactions on your behalf on Solana: trading bots, volume tools, execution helpers and anything else holding a key and acting without a human at each step. The material splits into two halves. Before you trust it covers acceptance criteria, test clusters, transaction simulation and sized rehearsals. When it misbehaves covers reproduction, on-chain evidence, regression protection and incident review.
The desk does not cover strategy. Whether a plan is a good idea is a separate argument from whether the software executes it correctly, and mixing the two produces material that is weak at both. A well-tested system faithfully executing a poor plan is a real and common outcome, and testing is not the discipline that prevents it.
How a procedure gets published
Every procedure is written as steps with an expected result, then checked against protocol documentation and against what the desk can perform itself. Where a step depends on the behaviour of a specific program or provider, the documentation for that program or provider is the source, and it is linked so a reader can verify the claim rather than take it on trust.
Where arithmetic appears, it is built from published protocol constants and clearly labelled assumptions, and it is marked as illustrative. Fee examples, rent figures and compute budgets are calculated in the text so that a reader can substitute their own numbers and follow the same path. No figure on this site is presented as a measurement from a run the desk has not described.
Sources
Primary sources are the Solana documentation, the validator client documentation, the published references for the programs a procedure touches, and the on-chain record itself. The ledger is treated as the authority whenever a tool's own report disagrees with it, which is a working rule rather than a rhetorical position: one of the two is durable, public and verifiable, and the other is not.
Where a claim cannot be sourced, it is either omitted or marked as untested. The desk would rather publish a shorter page with a stated gap than a complete-looking page with an invented middle.
Four things this desk does not publish
No fabricated numbers. There are no defect rates, pass rates, reliability scores, review counts or ratings anywhere on this site, because the desk has not run the sample that would justify them and inventing one would make every other number here worthless.
No verdicts about named products. QA Ground publishes methods for evaluating tools, not judgements about specific ones. A published accusation the desk cannot evidence would be unfair, and it would also be obsolete the moment a build changed. Readers are given the procedure and reach their own verdict.
No safety guarantees. A passing test establishes that on a stated build, with a stated configuration, over a stated window, observed results matched expected ones. It does not establish that software is safe, and any page here that seems to suggest otherwise is a page that needs correcting.
No procedure that exposes keys. Nothing published here asks a reader to enter a seed phrase or a private key into a website, a form, a chat or a support ticket. Where a step carries irreversible risk, the warning appears in the step itself rather than in a disclaimer at the bottom of the page.
Corrections
Technical material ages, and some of it is wrong on the day it is published. If a procedure here does not work, produces a different result, or describes behaviour that has since changed, the desk wants the report with enough detail to check it: the page, the step, what you observed, and any artefact such as a transaction signature. Corrections are made to the page rather than appended as a note, and the page's modification date reflects the change.
Authorship
Pages are published under the QA Ground Desk. There is no invented author profile, no photograph and no list of credentials, because the desk would rather be judged on whether the procedures work than on a biography. Reach the desk through the contact page.