The hard part of crypto is not buying. It is choosing.

A beginner meets this problem as a very long list of names, each with a price beside it, and no way to tell which of them are serious. The price does not help you, because a price tells you what other people have paid for something, not what the thing is or whether it works. Neither does the fact that you have heard of it.

What helps is asking the same small set of questions about every asset you look at, in the same order, and noticing which ones you cannot answer. Here are four, with what a thin answer actually looks like when you hit one.

A network and its token are two different things. The network can be genuinely interesting while the token attached to it has no documented claim on anything the network produces. So the question is not "is this project good". It is "what does this specific token do, and who needs to hold it".

If you cannot answer that in one sentence from the project's own material, you have learned something. That absence is a finding, not a gap in your reading.

Here is what that looks like in our research record. Gram, previously Toncoin, is an asset associated with TON, The Open Network, a general purpose blockchain on which developers can build applications and issue their own tokens. On our October 4, 2026 record Gram graded C / Speculative. Our published description of it, quoted word for word from that record:

And, further down the same description:

Read that carefully, because it is a statement about the evidence available, not a verdict on the network. The question "what is this token for" simply does not have a documented answer in the material we were able to examine. A reader asking question 1 for themselves would have arrived at the same wall.

Software rules are only as fixed as the people holding the keys. Before you compare one asset with another, find out who is able to alter the thing you are buying into, and whether those people are identifiable.

Two of the pass or fail screens in our published rating framework ask exactly this. One is about an anonymous or unaccountable team. The other is about custody arrangements that allow some party to freeze, mint or seize units of the asset.

You can ask both without any tooling. Are the people behind this identifiable, and accountable to anyone? Can any single party stop a transfer, freeze a balance, or create new units? A yes to the second question is not automatically disqualifying in every context, but it is something you want to know before you know anything else, rather than after.

These are two questions wearing one coat, and separating them is probably the single most useful habit on this list.

Usage asks whether real people and real volume are moving through the thing. Value accrual asks whether any of that activity reaches the holder of the token. A protocol can be busy and its token can still be entitled to none of the proceeds.

Our record on Aerodrome Finance shows both halves of that at once. Aerodrome is an automated market maker, meaning software that lets people swap one asset for another against a pool of funds rather than against another human being, and it is built around the Base network. On our October 4, 2026 record Aerodrome graded C / Speculative. From our published description of it, word for word:

And:

That is the second half of question 3 going unanswered and the first half going unmeasured, in one asset, in two sentences. Again, this describes the evidence that was available to examine. It is not a claim that nobody uses Aerodrome.

Most newcomers look at the supply figure and stop. The more useful version has three parts: how much of the supply exists and trades now, how much is waiting to be released and on what schedule, and who is holding the part that is not trading.

Our framework screens for both of the failure shapes here: supply concentrated in insider wallets, and a small portion of the supply trading freely while a large portion waits to be released. Aerodrome's description above happens to name the same gap from the other side: emissions and unlock schedule not established.

You do not need a model to ask this. You need the project's own supply documentation, and a willingness to treat "we could not find it" as an answer.

This is the part that is least intuitive, and it is how our own framework is built.

Our scoring has two layers. Six weighted dimensions produce a score, and the published bands put that score into a letter, as set out in our rating framework: A, Prime, is a score of 80 and above; B, Sound, is 60 to just under 80; C, Speculative, is 40 to just under 60; F, Failed, is below 40.

Those bands describe where a score sits. They are not the whole of what produces a letter grade. On top of the weighted score our framework also runs pass or fail disqualifier screens, and a tripped screen places an asset in F at any score, including a score in the A band. So an asset can be graded F while scoring well above 40.

Apply that to your own four questions and you get the right instinct. A strong answer on three of them does not survive a failure on the fourth. The answers do not average out.

Asking four questions about one asset is an evening's work if the documentation is good and a dead end if it is not.

The difficulty is not any single asset. It is that a comparison only means anything if every asset was asked the same questions, in the same order, against the same standard of evidence, and if the point where the answers ran out was recorded rather than filled in. Do that by hand across a few hundred assets and the first ones are no longer comparable with the last ones by the time you finish.

That consistency is the entire reason a rating exists. It is also why both of the quotations above read the way they do. Where the material did not establish something, our record says that the material did not establish it, in the same words, in the same place, for every asset.

CryptoGrade's AI researches tens of thousands of crypto assets to arrive at the top 200 by our own scores, then applies that framework to each of them. The result is a ranking, a letter grade, a score and a written research record per asset.

Our ranking is at cryptograde.ai/tokens. Logged out, you see the top 10 of it. With a free account you see the top 200 by our own score and the research record behind each grade, which is where sentences like the two quoted above live, one per asset, for every asset in the ranking. It costs nothing and it takes an email address.

One thing we will not do is tell you what belongs in your own portfolio. We publish what we think our data and our framework say about an asset, with the reasoning attached and the gaps marked. What you do with that is yours.

Four questions. Ask them in order, and treat the one you cannot answer as the answer.