How to idenfity good crypto projects

---

Four Filters: How to Choose What Belongs in a Crypto Portfolio

Most people do not pick cryptocurrencies. They collect them.

A friend mentions a ticker. A chart looks like a staircase. A founder gives a talk that makes the future sound already finished. The coin goes into a wallet the way a souvenir goes into a drawer - because the moment felt important, not because a question was answered.

That habit is understandable. Crypto is noisy, and noise feels like information. It is also an expensive way to learn. There are tens of thousands of tokens. Almost all of them will fail to matter. Some will fail politely, by fading. Some will fail by design, by transferring wealth from late believers to early ones. A few will do something the old financial system cannot do as well, or cannot do at all.

The process to use is not a crystal ball. It will not tell you which coin goes up next month. It is a sequence of filters meant to keep you from confusing a good story with a good holding. Each step is stricter than the one before it. A project can look brilliant at step one and still die at step three. That is the point. The goal is not to find reasons to buy. The goal is to find reasons to walk away while there is still time.

The four filters are simple on purpose:

1. Niche: Identify the crypto’s niche or edge.
2. Proof: Verify the technology can deliver that advantage.
3. Demand: Ask whether adoption and economics turn the advantage into durable demand.
4. Size: Size the position so a bad outcome is survivable.

What follows is not a checklist you can sprint through in five minutes. It is a way of thinking. Used honestly, it makes the market feel less like a carnival and more like a set of engineering problems with prices attached.

---

Filter one: What job is this thing actually trying to do?

Every serious project is an argument about a missing job.

Bitcoin’s argument is that money should be hard to seize, hard to inflate, and possible to send without asking a bank. Ethereum’s argument is that agreements should be able to run themselves. Chainlink’s argument is that a self-running agreement is useless if it cannot learn anything about the world outside its own ledger. Monero’s argument is that money which publishes a diary of every payment is not quite money. Uniswap’s argument is that a market should not need a listing committee. Hyperliquid’s argument is that a derivatives exchange can live on-chain without feeling like a toy. Bittensor’s argument is that intelligence itself might be produced by a market instead of a campus.

Notice what those sentences share. They do not begin with a price target. They begin with a job.

A niche is not a slogan. “Web3,” “the future of finance,” and “AI plus crypto” are not niches. They are weather systems. A niche is a specific problem, felt by specific people, that this design claims to solve better than the obvious alternative. The alternative might be a bank, a cloud company, a centralized exchange, another blockchain, or nothing at all - the old world simply leaving the problem unsolved.

Try to write the edge in one plain sentence a high-school student could repeat. If you cannot, you do not understand the project yet. “It will go up because the sector is hot” is not a sentence. “It lets a smart contract know a price without trusting a single messenger” is a sentence. “It hides sender, receiver, and amount on every ordinary payment” is a sentence. “It lets anyone create a market by dropping two tokens into a pool” is a sentence.

That sentence has to survive a rude follow-up: compared with what?

If the answer is “compared with doing nothing,” the project may still matter - some problems really were unsolved. If the answer is “compared with Coinbase,” or “compared with Amazon’s servers,” or “compared with Bitcoin itself,” then the edge has to be real enough that users would accept crypto’s extra friction to get it. Crypto is almost never the most convenient option. Convenience is what the old system is good at. An edge that is only slightly better than a bank app is not an edge. It is a science fair.

This first filter also throws out a surprising amount of the market. Tokens that exist mainly to have a ticker do not have a niche. Tokens that copy a neighbor with a new mascot do not have a niche. Tokens whose only stated purpose is “governance” of a product nobody needs do not have a niche. Governance is a tool. It is not a reason to exist.

A useful trick: imagine the token disappearing tomorrow and ask what breaks. If the honest answer is “a Discord channel gets quieter,” there was no edge. If the honest answer is “oracles stop, or private payments become public again, or a live market has no place to sit,” you are looking at a candidate. You are not looking at a purchase yet. You are only looking at a claim.

---

Filter two: Can the machine actually do the job?

Crypto is full of brochures for machines that have not been built, or have been built and do not work as advertised.

Step two is where enthusiasm has to sit down and look at the engine. The question is not “does the white paper sound intelligent?” The question is “does the technology deliver the advantage named in step one - today, or on a path that does not require a miracle?”

This is the part people skip because it feels like homework. It is also the part that separates a protocol from a pitch deck.

Start with the mechanism, not the brand. If the edge is privacy, how does a payment hide? If the edge is an oracle, how do many messengers agree without becoming one messenger in a trench coat? If the edge is a decentralized exchange, what actually prices the trade - a pool, an order book, a hybrid - and what happens when markets get violent? You do not need a computer-science degree to follow the shape of the answer. You do need to refuse answers that dissolve into adjectives.

Then ask whether the design has been punished by reality. A testnet is a rehearsal. A mainnet with real money is a trial. Has the code been used under stress? Has it been audited, broken, patched, and used again? Has it survived a year when nobody was cheering? Technology that only works in a bull market is not technology. It is décor.

There is a difference between hard and hand-wavy. Some problems are genuinely difficult. Getting data onto a blockchain without re-creating a middleman is difficult. Hiding transaction graphs is difficult. Building an on-chain exchange that can handle leverage without melting is difficult. Difficulty is allowed. Fog is not. If the team cannot explain the hard part without retreating into “our novel consensus” or “our AI layer,” the advantage from step one is still a hope.

Also watch for a quiet kind of failure: the product works, but not at the layer that matters. A chain can be fast in a demo and centralized in production. A privacy coin can hide amounts and still leak a graph. An “AI network” can pay people to run models and still be a thin wrapper around the same few companies that already train models. The test is whether the claimed edge survives contact with the actual architecture.

This filter should make you slower, not smarter-looking. If you cannot get a clear picture of how the machine works, the correct move is not to invent confidence. The correct move is to keep reading, or to pass. Passing is a position. In a market this large, passing is the default position.

One more honesty check belongs here. Verify the people only after you verify the mechanism. Charisma is not a consensus algorithm. Founders matter - they ship, they freeze, they vanish - but a biography is not proof that the code does what the story says. Work backward from the machine. If the machine is real, the names become context. If the machine is vapor, the names are theater.

---

Filter three: Will anyone keep paying for that advantage?

A working machine is not the same thing as a business, and a business is not the same thing as a token that deserves to exist.

Step three is the coldest filter, and the one crypto culture is worst at. It asks whether the advantage becomes demand that lasts, and whether that demand has a reason to touch the token.

Adoption is the first half. Do real users show up when the price is boring? Are they using the product because it is the best tool for a job, or because they are farming a reward that will end? A protocol that lights up when tokens are being printed, then goes dark when the printing stops, did not find product-market fit. It found a temporary job program.

Useful questions, asked without romance:

• Who needs this badly enough to use it even if the token is not going up?
• What do they use if this project disappears - and is that substitute good enough that they would shrug?
• Is usage growing in the part of the product that matters, or only in the part that is easy to screenshot?
• Are the users diversified, or is one fund, one chain, or one incentive program the whole story?

Economics is the second half, and it is where many good inventions become poor investments.

A token can be many things: a right to use a network, a right to govern it, a claim on fees, a reward for work, a unit of account inside a machine, or a souvenir of a moment when people felt early. Only some of those create durable demand. “Durable” is the word that matters. A token that people buy because they hope someone else will buy it later is not an economic design. It is a hot potato with a logo.

Tokenomics, translated out of jargon, is just the set of rules for how coins appear, who gets them, what they are for, and whether anyone has to hold them to get the thing the project is good at.

Supply matters, but not as a magic number. A fixed cap can be a feature if the token is trying to be money. It can be a trap if the network still needs to pay workers forever and has no other budget. A large “community” allocation can be decentralization. It can also be a schedule of future selling. Unlocks are not evil. Unlocks that dwarf actual usage are a weather report: expect wind.

Utility matters more than the cap. Does anyone need the token to do the job from step one? Chainlink’s LINK is used to pay node operators. That is not a guarantee of price. It is at least a loop that touches the real work. A governance token for a protocol that would function the same way if the token vanished is a weaker loop. It might still be valuable if governance is scarce and the protocol prints cash. It might be a voting sticker glued to a product that does not need voters.

Incentives matter because they reveal what the project actually pays for. If the largest reward goes to mercenaries who arrive, extract, and leave, the network is renting appearance. If the reward goes to people who supply a scarce resource the machine needs - security, privacy-preserving computation, honest data, inventory in a market - then the economics might be attached to the edge instead of attached to the chart.

Fees and value capture are the last door in this filter. Even a beloved product can fail the token. Users may pay in some other asset. Fees may go to a company, not to the coin. The chain may thrive while the token is diluted into irrelevance. This is not a scandal. It is a reminder that “the protocol is used” and “the token must rise” are different sentences. You want a story for how usage creates a reason to hold or spend the coin that does not depend on a new wave of buyers arriving out of nowhere.

If you cannot tell that story without using the word “narrative,” the position is not an investment in infrastructure. It is a bet on attention. Those bets can pay. They are also how people confuse a casino chip with a tool.

---

Filter four: If I am wrong, am I going to be ok?

The first three filters try to reduce the chance of being wrong. They cannot eliminate it.

Crypto is a young field attached to a violent market. Regulation can change the legality of a niche overnight. A better design can appear and eat the edge you liked. A working protocol can be hacked. A token with real use can still go down eighty percent and stay there while the market decides that some other machine is the one that matters. You can do the work and still lose.

Step four exists because survival is a strategy.

Position size is the unglamorous half of investing, which is why it is the half that keeps people in the game long enough to learn anything. DO NOT size a position by how exciting the idea is. Excitement is how oversized positions are born. Size it by how a bad outcome would feel in a life that still has rent, time, and other work to do.

A simple way to say it: a position should be large enough that being right matters, and small enough that being wrong is not a personality crisis.

That sentence hides a few practical rules.

First, treat ruin as a different category from loss. A twenty percent drawdown is information. A position that, if it went to zero, would force you to sell the rest of the portfolio, change your plans, or make frightened decisions is not a holding. It is a hostage situation. The whole point of a risky asset class is that some tickets will be worth zero. Plan for zeros the way a ship plans for holes.

Second, correlation is a sneak. Five tokens that all rise when liquidity is cheap and all fall when it is not are not five bets. They are one bet in five costumes. If every coin in the wallet is a leveraged cousin of the same risk-on wave, the portfolio is concentrated even when the tickers look diverse. A privacy coin, an oracle network, and an exchange token may still move together in a panic. They will not always fail for the same reason. That difference is worth something, but it is not a seatbelt.

Third, cash and boredom are tools. The ability to do nothing is underestimated in a culture that treats every dip as a quiz. If a position is sized so that you need it to recover for your plan to work, you will start reading the news like a prayer book. That is how people add to losers to “get back to even,” which is not analysis. It is mood.

Fourth, time is part of size. A token whose thesis takes years - a new market for machine intelligence, a slow migration of finance onto public rails, a long fight over whether private money is allowed to exist - cannot be sized like a weekend trade. If you cannot hold it through a season where the chart looks like a mistake, you do not have a thesis. You have a wish with a stopwatch.

This filter also forces a kinder kind of humility. The first three steps can make you feel like an analyst. The fourth reminds you that the future is not required to respect your notes. You would rather own a smaller amount of something you understand than a large amount of something you needed to be large so the story would feel worth the work. The work is supposed to protect you. It is not supposed to demand a monument.

None of this is a promise that small positions cannot hurt. They can. It is a promise to keep hurt in the realm of “that was expensive schooling” rather than “that was the year everything else had to be rearranged.”

---

What this process will not do

It will not make you early to every winner. Some of the best-performing tokens of any year fail these filters on purpose. They have no niche except attention. Their technology is a costume. Their economics are a transfer from the impatient to the slightly less impatient. Then the costume works long enough to make the filters look prudish.

That is not a reason to abandon the filters. It is a reason to be clear about what they optimize for. You are not trying to catch every rocket. You are trying to own a few machines whose job you can explain, whose workings you can inspect, whose demand you can imagine in a quiet market, and whose failure you can survive.

The process also will not remove judgment. Two honest people can look at the same oracle network and disagree about whether the token captures the value. They can look at the same privacy coin and disagree about whether the political risk is a reason to own it or a reason to avoid it. Filters are not a substitute for taste. They are a way to keep taste from turning into improvisation.

And it will not run itself. The temptation, after writing four numbered steps, is to treat them like a stamp on a passport. Real work is recursive. A project that passed last year can fail step three this year if usage was rented. A technology that looked sufficient can look naive after a competitor ships. Sizing that felt conservative can become reckless if the rest of life changed. The process is a habit, not a certificate.

---

Why bother with all this when a ticker is faster

Because crypto, at its best, is not a ticker. It is a set of experiments in whether strangers can coordinate without asking for permission.

That is a larger idea than a portfolio. It is also why sloppy portfolios are a kind of insult to the idea. If the whole point is to replace blind trust with systems you can inspect, then buying a coin because a stranger on a stage was confident is a step backward. You have only changed which stranger you trust.

The four filters are a way to stay on the side of inspection.

Identify the job, or you will own a mascot. Verify the machine, or you will own a brochure. Demand that adoption and economics meet, or you will own a wonderful invention that never needed its token - or a token that never needed its invention. Size the risk so that a wrong call is not the end of your ability to keep looking. Looking is the actual skill. The market will still be there after a mistake. You need to be there too.

What would have to go right for this process to be worth anything is not a single winning pick. It is a quieter result: fewer coins, better reasons, losses that teach instead of wreck, and a portfolio that still makes sense if you explain it out loud to someone who does not care about your capital gains.

If a holding cannot survive that explanation - niche, proof, demand, size - it probably cannot survive a bear market either. The market is not required to be fair. It is very good, over time, at asking what a thing is for. That is the question you are trying to answer before it does.