Blog

What is actually new in DeFi

Decentralised finance is described either as a reinvention of banking or as a casino with extra steps. Both miss the more interesting question: which parts are actually new mechanisms, and which are familiar arrangements wearing new words?

Genuinely new: automated market makers

Traditional markets match buyers to sellers through an order book. An automated market maker replaces that with a pool of two assets and a formula: the price is set by the ratio between them, and every trade shifts it.

This is a real invention. It lets a market exist for an asset with no professional market makers, no minimum size and no permission. It is also why slippage and impermanent loss exist — both are direct consequences of the mechanism, not bugs.

Genuinely new: flash loans

A flash loan lets you borrow any amount with no collateral, provided you repay it within the same transaction. If you do not, the entire transaction reverts as though it never happened.

Nothing analogous is possible in conventional finance, where settlement is not atomic. It enables useful things — arbitrage, refinancing a position in one step — and has also been the tool of choice for manipulating thinly traded price oracles during exploits. A genuinely novel primitive with genuinely novel failure modes.

Genuinely new: permissionless composability

Any protocol can call any other without an agreement, an integration project or an API key. A lending market can accept another protocol’s receipt token as collateral automatically.

This is the most under-appreciated property and the most double-edged. It produces rapid innovation, and it means a failure in one protocol propagates instantly through everything that depended on it — a systemic risk with no central point that could pause it.

Not new: lending against collateral

Over-collateralised lending is a pawnbroker with a smart contract. The mechanism is ancient; what changed is that liquidation is automatic and merciless, with no discretion and no phone call. That is a real difference in user experience, not a new financial primitive.

Not new: yield that comes from somewhere

Much of what is presented as innovation is a familiar structure with the safeguards removed. Returns paid in a newly minted token are dilution, not income. Returns paid from later deposits have a very old name. Our guide on yield farming and the truth about APY works through where the numbers actually come from.

Not new: leverage

Leverage is the oldest amplifier in finance. Wrapping it in a token does not change what it does, and in DeFi the liquidation is faster and less forgiving than a broker’s would be.

How to tell the difference

Ask what the mechanism actually does that could not be done before, and what the yield is paid from. If the answer to the first is “the same thing, without a licence” and to the second is “the token it prints”, you are looking at repackaging.

That is not automatically a reason to avoid something — but knowing which you are holding is the difference between an informed decision and a hopeful one. Start with what is DeFi for the risk model.

7 August 2026
How to follow Ethereum development without the noise
Ethereum coverage is dominated by price commentary and speculation about upgrades. The underlying process is public, well documented and genuinely readable — often more so than the articles about it.

Where changes actually happen

Protocol changes go through Ethereum Improvement Proposals. An EIP sets out a specific change, its motivation, its technical specification and its backwards-compatibility implications, in a standard format. Anyone can read them.

Proposals are discussed on public forums and on recorded core developer calls with published notes. There is no private committee. When you read that “Ethereum is planning” something, the underlying document is nearly always a linkable EIP with a status field telling you how real it is.

Reading an EIP without being a developer

Every EIP opens with an abstract and a motivation section written in prose. Those two sections usually tell a non-developer what the change does and why, before the specification gets technical.

Check the status field. Draft means someone proposed it. Review and Last Call mean it is being seriously considered. Final means it is settled. A great deal of breathless coverage concerns proposals that never left Draft.

What “the roadmap” actually is

Ethereum’s roadmap is a set of directions rather than a schedule: scaling through data availability and layer twos, reducing the cost of running a node, improving censorship resistance and finality.

Timelines slip routinely and by long margins. This is normal for consensus-driven infrastructure and is not evidence of failure — but it does mean any article citing a firm date deserves scepticism, especially if the date is being used to justify a price target.

Sources worth your time
- The EIP repository — the actual proposals, with status.
- Core developer call notes — what was discussed and decided, published after each call.
- The official documentation — maintained, versioned, and clearer than most explainers.
- Client release notes — what shipped, as opposed to what was announced.
What to discount

Coverage that describes an upgrade only in terms of its expected price effect. Anything citing a hard date without linking the decision. Threads explaining what an upgrade “means for holders” without explaining what it changes technically. And anyone presenting a Draft proposal as settled.

Why this is worth the effort

Ethereum is the settlement layer for most of what happens in crypto. Understanding how it changes — slowly, publicly, with argument — is a useful corrective to an industry that presents everything as imminent and revolutionary. Start with Ethereum explained if you want the fundamentals first.
7 August 2026
NFTs, the metaverse and Web3: an honest reckoning
Enough time has passed to assess the 2021 wave honestly rather than defensively. Some of it worked. Most of it did not.

What was promised

A read-write-own internet: users holding their data and identity through wallets rather than platform accounts; creators selling directly to audiences without intermediaries; persistent virtual worlds with portable, user-owned assets.

What actually happened to NFTs

The dominant use was speculation on profile-picture collections. Enormous sums changed hands, and the great majority of those collections subsequently lost nearly all their value and — more importantly — nearly all their liquidity. Many cannot be sold at any price today.

Two structural problems became clear. Most NFTs point at artwork hosted on ordinary servers, so the token survives while the image can vanish. And ownership of the token rarely conveys copyright, which surprised a great many buyers who believed otherwise.

What survived is narrower and quieter: ticketing, credentials, provenance records, and some genuine creator-audience relationships. Those were always the more plausible applications and got the least attention.

What happened to the metaverse

Virtual land sold for extraordinary sums in worlds that, on inspection, had very few concurrent users. Several of the most-hyped platforms have since reported user numbers that make the valuations difficult to explain.

The underlying idea — persistent shared virtual spaces — is not absurd, and versions of it thrive in gaming. What failed was the specific claim that blockchain-based ownership of virtual property was the missing piece, and that people wanted it.

What happened to Web3

Here the record is genuinely mixed. Wallet-based sign-in works. Decentralised storage networks run. Blockchain naming services function. Tokens really can represent ownership a platform cannot unilaterally revoke.

Against that: a great many “decentralised” applications depend on centralised front-ends, hosting and APIs, so they can be taken offline by the same parties as any website. Governance tokens are frequently concentrated enough that a handful of holders decide outcomes. And most users have shown little appetite for managing their own keys — which is the precondition for the entire model.

What is worth keeping
- Verifiable ownership records for things where provenance genuinely matters.
- Censorship-resistant value transfer, which has clear real-world use for people under capital controls or excluded from banking.
- Permissionless composability — that anyone can build on top without asking.
Those are real and useful. They are also considerably less exciting than what was sold, which is why they attracted less capital and less coverage.

Why we cover this sceptically

Our readers are largely beginners, and beginners arriving now encounter this vocabulary mostly through people trying to sell them something. Presenting a promise as an achievement would not serve you. See our ethics page for what we will and will not publish.
7 August 2026
How to follow crypto markets without checking hourly
Crypto markets never close, which creates a genuine problem: there is always something happening, and almost none of it matters.

Why hourly checking hurts

Short-term price movement is overwhelmingly noise. Checking frequently exposes you to a stream of meaningless fluctuation, and each glance is an invitation to act.

The documented result is worse outcomes. Frequent monitoring correlates with more trading; more trading means more fees, more spread paid, more tax events, and more decisions made from feeling rather than reasoning. It also amplifies loss aversion — watching a position fall in real time makes selling at the bottom considerably more likely than reviewing it weekly would.

A weekly rhythm

Once a week, at a set time, look at:
- Where prices sit in a one-to-two year range, not against last Tuesday. Context, not movement.
- Anything structural: regulatory decisions, a major protocol upgrade, an exchange in difficulty. These move markets in ways patterns do not.
- Your own position sizing. Has one holding grown into an uncomfortable share of the total?
- Whether your original reasoning still holds. Not the price — the reason.
That is genuinely enough for anyone not trading professionally, and professionals have tooling and time you do not.

What is worth reacting to

Very little, and almost none of it is a price move. Things that warrant attention: an exchange you use restricting withdrawals; a protocol you hold funds in disclosing an exploit; a regulatory change affecting your ability to access or report holdings; a project you hold shipping — or abandoning — the thing you bought it for.

Notice these are events, not chart shapes. A 12% daily fall in a volatile asset is ordinary; an exchange delaying withdrawals is not.

Sources worth having

Prefer primary sources: protocol documentation and release notes, exchange status pages, regulator publications, on-chain data you can query yourself. Treat aggregators and social media as pointers to primary sources, never as the source.

Be especially wary of anything with a referral link, and of accounts whose confidence never varies with the difficulty of the question.

The uncomfortable part

A sensible approach to following markets is boring, and boredom is exactly what the industry’s content is designed to prevent. Urgency, constant updates and dramatic framing all exist to keep you engaged, because engagement is the product being sold.

You are allowed to check weekly, and you will probably do better for it.
7 August 2026
Why most crypto price predictions are worthless

Search any coin’s name and you will find price predictions for the next five years, stated to two decimal places. Understanding how that content is produced explains why almost none of it is worth reading.

The incentive problem

Prediction articles exist because they rank and attract clicks, not because anyone can do it. The people publishing them are rarely taking positions on the forecasts, are never scored on accuracy, and face no consequence when a target is missed by an order of magnitude.

A forecast with no accountability is not a forecast. It is a headline.

The tells

No time horizon. “X will reach $Y” without a date cannot be wrong, and a claim that cannot be wrong contains no information.

No assumptions. Real analysis says what would have to be true. If the reasoning is absent, there is nothing to evaluate.

A single number. The future is a distribution, not a point. A range with stated probabilities is at least an honest shape; a precise figure is false precision.

No falsification condition. Ask what would show the call was wrong. If nothing would, it is not analysis.

No track record. Anyone confident in their method publishes their past calls, including the failures. Almost nobody does.

The special case of extrapolation

Many predictions are trend lines extended forward, sometimes on logarithmic axes that make explosive growth look like a straight line. This is not modelling; it is assuming the future resembles the past by construction. The technique produces impressive charts and has no predictive content.

Watch particularly for extrapolations that assume an asset captures some enormous existing market — “if it takes 1% of gold” — with no argument for why that would happen or when. That is a multiplication, not a forecast.

What honest forward-looking analysis looks like

It states a time horizon. It sets out assumptions explicitly. It gives a range rather than a number. It names what would falsify it. And it is revisited afterwards, publicly, including when it was wrong.

That kind of writing exists and is much rarer and much less exciting than the alternative. It also tends to conclude that the range of outcomes is wide, which is both true and commercially useless.

Our position

We do not publish price targets as predictions. Where we discuss possible future prices at all, it is as labelled scenarios with stated assumptions, a stated horizon and a range — set out in our editorial guidelines. Nothing we publish is financial advice.

If you want to understand what moves prices rather than what someone claims they will be, how to read a crypto chart and reading on-chain data are more useful starting points.

7 August 2026
How to read a crypto chart without fooling yourself
Technical analysis attracts strong opinions in both directions. The honest position is narrower than either camp: charts are useful for a specific, limited set of things, and useless for the thing people most want from them.

What a chart actually shows

Price over time, and volume. That is all the data. Everything else — support levels, trend lines, indicators — is interpretation layered on top.

Some of that interpretation has a real basis. Levels where price repeatedly turned are levels many participants are watching, and that shared attention can be self-fulfilling: enough people placing orders at a round number makes that number behave differently. That is a genuine, if circular, effect.

What it cannot do

A chart contains no information about the future. It cannot know that an exchange will fail next week, a regulator will act, or a protocol will be exploited — and those events move crypto prices far more than any pattern.

The strong claim, that patterns reliably predict direction, does not survive contact with the evidence. If a pattern predicted price with useful accuracy, it would be arbitraged away almost immediately.

Patterns that are not there

Humans are extremely good at finding structure in noise, and financial charts are largely noise. Two failure modes are worth naming:

Selection after the fact. Every chart contains dozens of shapes. Identifying the one that “worked” afterwards is not prediction. The honest test is writing the call down beforehand with a time horizon, and keeping score of all of them — including the ones you would rather forget.

Indicator shopping. With enough indicators and enough settings, something will always support the view you already hold. If you are adjusting parameters until the signal agrees with you, you are decorating a decision, not making one.

What charts are genuinely useful for
- Context. Is this price high or low relative to the last two years? Surprisingly many people buy without knowing.
- Liquidity. Volume tells you whether you could exit a position at anything like the displayed price.
- Risk levels. Identifying where you would conclude you were wrong, before you enter, is the most valuable thing a chart offers.
- Volatility. How much this asset routinely moves tells you how large a position you can hold without being forced out.
Notice that three of those four are about risk rather than direction. That is where the actual value sits.

The honest summary

Use charts to understand context, liquidity and risk. Do not use them to forecast, and be suspicious of anyone who does — particularly anyone selling the service. Our price analysis states its time horizon and shows its reasoning for exactly this reason, and none of it is advice.
7 August 2026
Stablecoins explained: three mechanisms, and how each has failed
Stablecoins exist because moving in and out of volatile assets through the banking system is slow and expensive. Rather than converting back to ordinary money, traders park value in a token that is supposed to stay at a dollar.

The word “stable” does a great deal of work in that sentence, and it is worth understanding what is actually holding the peg.

1. Fully reserved

How it works: a company issues tokens and holds matching reserves — cash and short-term government debt — redeeming tokens for dollars on demand.

What you are trusting: that the reserves exist, are genuinely liquid, and that the company will honour redemptions. This is a credit relationship with an issuer, not a property of the token.

How it has failed: reserved stablecoins have briefly traded below a dollar when confidence in the backing wavered — for instance when reserves were held at a bank that failed. They recovered, but holders who sold during the gap did not.

2. Over-collateralised

How it works: users lock crypto worth more than the stablecoins they mint — say $150 of ETH for $100 of stablecoin — with automatic liquidation if the collateral falls too far.

What you are trusting: the smart contracts, the price oracles feeding them, and that liquidations can actually execute during a crash.

How it has failed: during violent falls, network congestion has prevented liquidations from processing in time, leaving positions under-collateralised. The mechanism depends on the chain working precisely when everyone is using it hardest.

3. Algorithmic

How it works: the peg is maintained by supply mechanics — typically minting and burning a paired token — without adequate external backing.

What you are trusting: that market participants keep arbitraging the peg, which requires continued confidence in the paired token.

How it has failed: catastrophically. In 2022 a major algorithmic stablecoin lost its peg and collapsed to near zero within days, destroying tens of billions of dollars of value and taking several lenders with it. The failure mode is reflexive: confidence falls, the paired token falls, which reduces the backing, which reduces confidence further. Once that starts it does not stop.

What this means practically
- Holding a stablecoin is holding a claim on an issuer or a mechanism. It is not cash.
- Yield offered on stablecoin deposits is compensation for risk — someone is lending them out. Ask to whom.
- A stablecoin paying substantially more than others is not a better deal; it is a different risk.
- Diversifying across designs is more useful than diversifying across issuers of the same design.
If you are using stablecoins inside DeFi, read what is DeFi for how these compose — and how a peg failure cascades through everything built on it.
7 August 2026
How to assess an altcoin before you buy it

There are tens of thousands of altcoins and the great majority are worthless. The useful skill is not picking winners; it is discarding obvious losers cheaply, which a short checklist does surprisingly well.

1. Who holds the supply?

Open a block explorer and look at the holder distribution. If a handful of wallets control most of the supply, they can sell into whatever liquidity exists and leave everyone else with a collapsed price. This is the single most predictive check available and takes about a minute.

Watch for supply split across many wallets funded from the same source at the same time — that is one party, not many. And identify which large holders are exchange or locked contracts rather than individuals.

2. What is the unlock schedule?

Compare circulating supply against total supply. If most tokens are locked, find the vesting calendar. Those unlocks are dated sell pressure, and they routinely coincide with sharp falls. A token that looks reasonable on circulating supply can look very different fully diluted.

3. Is there a product, or only a roadmap?

Does the thing exist and do people use it? Not “is there a testnet” or “is it coming in Q3” — is there something working now, with users who are not being paid to be there? A great many projects have sophisticated marketing and no shipped product.

4. Does the token actually capture value?

This is the question most often skipped and it is frequently decisive. If the project succeeds enormously, what forces anyone to buy the token?

Sometimes the answer is real: the token is required to pay fees, or to secure the network, or it has a claim on revenue. Often the answer is nothing — the protocol works fine and the token is a governance badge with no mechanism connecting success to price. In that case the token can fall while the product thrives.

5. What does the chain say?

Reported volume can be manufactured by trading with yourself; active address counts can be inflated by one person with a script. Look instead at whether addresses return over time, whether liquidity depth is proportionate to the claimed market cap, and whether activity correlates with anything real. Our guide on reading on-chain data covers this properly.

6. Who can change the rules?

Check whether the token contract lets an owner mint unlimited supply, pause transfers, blacklist addresses or alter fees. Any of those means the terms are whatever the deployer decides tomorrow.

What this checklist will not do

It will not find you a winner. Passing every question means a project is not obviously broken, which is a much weaker statement than being a good investment. Plenty of well-constructed projects still fail.

What it does reliably is eliminate the large majority of tokens in under ten minutes, which is where most of the avoidable losses are. See also rug pull and how to spot a crypto scam.

7 August 2026
Ethereum explained: smart contracts, gas and staking

If Bitcoin is a ledger for money, Ethereum is a computer that many machines run together. That difference explains almost everything else about it.

Programs on a blockchain

Ethereum runs smart contracts — code deployed to the network that anyone can call and that executes identically on every node. A contract can hold funds, enforce rules and interact with other contracts.

This generality is why the overwhelming majority of crypto applications, from lending protocols to NFT marketplaces, live on Ethereum or on chains that copy its design. It is a platform rather than a product.

Why gas exists

Running code consumes resources, and gas prices that consumption. A simple transfer uses little; a complex contract interaction uses much more. Your fee is roughly gas used × gas price.

Because block space is limited and demand varies, the price is effectively an auction. When the network is busy, users bid higher to be included sooner and fees can rise sharply within minutes. Two consequences catch newcomers out: a small transaction can cost more in fees than it is worth, and a failed transaction still costs gas, because the network performed the work before the failure.

What proof of stake changed

Ethereum moved from mining to proof of stake in 2022. Validators now lock up ETH as collateral instead of burning electricity, and misbehaviour destroys part of that stake.

The headline effect was energy: consumption fell by more than 99%. What it did not do is make fees cheap — fees are a function of block space demand, not of how the chain reaches consensus. That distinction is frequently muddled.

The open criticism is that influence now follows existing holdings, which may concentrate control among large holders and staking services over time.

Where the risk actually is

Ethereum’s base layer has run since 2015 without its ledger being forged. Almost every large loss in its ecosystem has been a smart contract exploit — ordinary software bugs, except that anyone can trigger them and the losses are irreversible.

Being audited reduces that risk without settling it; plenty of audited contracts have been drained. If you intend to use applications rather than just hold ETH, read how to check a smart contract before you interact first.

Layer twos

Most activity is migrating to layer-two networks that batch transactions and settle to Ethereum, which is how fees come down without changing the base layer. They introduce their own trade-offs around who can censor or halt them, and those trade-offs differ substantially between designs — worth understanding before assuming they are all equivalent.

7 August 2026
Bitcoin explained: what it is for, and what it is not

Bitcoin is the oldest cryptocurrency and the one most people encounter first. It is also the one most often described in terms that have little to do with how it works.

The actual proposition

Bitcoin is a payment and settlement network with a fixed money supply that nobody can change. There will only ever be 21 million bitcoin, released on a schedule that halves roughly every four years until issuance stops around 2140. That schedule is enforced by software every participant runs and can verify.

No company operates it. No government issues it. There is no head office to subpoena and no board that can decide to print more. Whether that is valuable depends entirely on whether you think those properties are worth having — but they are real properties, not marketing.

How it is secured

Mining is what makes the ledger expensive to rewrite. Miners run hardware that competes to solve a puzzle; the winner proposes the next block and receives newly created bitcoin plus transaction fees. Because winning costs genuine electricity, an attacker wanting to rewrite history must out-spend the entire honest network, sustained over time.

This is also the source of the energy criticism, and it is worth stating plainly rather than defending reflexively: the energy consumption is not incidental inefficiency that better engineering will remove. The security is the expenditure. Arguments about Bitcoin’s energy use are really arguments about whether what it secures is worth the cost.

What it deliberately does not do

Bitcoin has no meaningful smart contract capability, processes a limited number of transactions per second, and changes very slowly — protocol changes take years and require overwhelming consensus.

Supporters treat this conservatism as the entire point. A monetary network is not somewhere you want rapid experimentation, and every feature added is a new way to fail. Critics point out that this leaves Bitcoin unable to do most of what the wider crypto industry now does. Both observations are correct; they simply value different things.

The “digital gold” claim

Bitcoin is frequently described as digital gold — a scarce asset that holds value when currencies weaken. This is a claim about what it might become, not a settled description of how it behaves.

Observed behaviour has been considerably more volatile than gold, and Bitcoin has repeatedly fallen alongside risk assets during market stress rather than acting as a haven. It may grow into the role. It has not demonstrably occupied it yet, and anyone presenting that as established fact is describing a hope.

What actually matters if you hold it

Two practical things. First, custody: bitcoin held on an exchange is a claim against a company, and several such companies have failed. Second, volatility: falls of 70–80% from peak have happened more than once and taken years to recover.

If you want the mechanics rather than the narrative, start with how blockchains actually work, then how to set up your first crypto wallet.

7 August 2026