What Is Quantum Computing? A Complete Guide from Qubits to the Future

Imagine a maze painted on a school floor. A classical computer can follow a disciplined set of instructions: try this turn, record the dead end, return, and try another. A quantum computer does not magically walk every path, peek at every answer, and announce the exit. Instead, it prepares a carefully controlled wave of possibilities. Quantum operations reshape that wave so that wrong routes tend to cancel and useful routes become more likely to appear when the machine is measured.

That is the central idea behind quantum computing: not unlimited parallel guessing, but the deliberate control of quantum states, amplitudes, correlations, and interference to solve certain kinds of problems.

So, what is quantum computing? It is a model of computation that stores and processes information in controllable quantum systems. Those systems may be tiny electrical circuits, trapped atoms, photons, or other physical objects that obey quantum mechanics. This guide starts with bits and qubits, builds toward algorithms and error correction, then asks what the technology may mean for science, cybersecurity, and Bitcoin.

A quantum computer is not a universally faster laptop. Its advantage depends on the problem, algorithm, hardware quality, and error-correction cost. Classical machines will continue to run most software; quantum processors are more likely to work beside them as specialized accelerators.

In this guide 20 sections

Quantum computing in one simple definition

A useful comparison is a graphics processor. A GPU excels at selected repeated operations but does not replace every component. Similarly, a conventional computer can prepare a problem, send a suitable part to a quantum processor, and interpret the measurements.

The word quantum refers to the rules governing matter and light at small scales. The word computing means using physical processes to transform encoded information. Quantum computing joins the two by making quantum behavior programmable.

Why quantum physics enters computing

In daily life, a switch is on or off and a ball is here or there. This familiar picture matches classical physics and classical information.

At microscopic scales, nature does not always offer definite answers before measurement. Some properties are quantized, occurring in allowed levels. Think of a staircase rather than a ramp. The analogy breaks because a quantum system can also occupy a superposition of allowed states, with phase relationships no staircase possesses.

The precise description uses a quantum state and probability amplitudes. Amplitudes can be positive, negative, or complex numbers. Their squared magnitudes determine measurement probabilities, while their phases determine how alternatives interfere. This is richer than ordinary uncertainty. A shuffled card is already one particular card even when you do not know which; a quantum state can contain coherent alternatives that later reinforce or cancel.

Engineers exploit physical systems whose states can be isolated, controlled, and measured. Energy levels in ions, microwave states in superconducting circuits, paths or polarizations of photons, and electron spins in silicon can all carry information. NIST’s overview emphasizes that many quantum systems can serve as qubits, while each implementation brings different tradeoffs in lifetime, speed, manufacture, and control (NIST).

Classical bits versus qubits

A classical bit has one value at a time: 0 or 1. Its physical realization might be a voltage, magnetic orientation, or transistor state. Reliable bits are the foundation of modern computing because they can be copied, checked, and stored with extraordinarily low error rates.

A qubit has two reference states written |0⟩ and |1⟩, called basis states. Its general pure state is written as α|0⟩ + β|1⟩. The symbols α and β are amplitudes, and their squared magnitudes add to one. If |α|² is 0.7 and |β|² is 0.3, a measurement in that basis returns 0 about 70 percent of the time and 1 about 30 percent over many identical preparations. One measurement produces one classical outcome; it does not print α, β, or every component of the state.

Feature	Classical bit	Qubit
Basic states	0 or 1	Basis states
General state	One definite value	Amplitudes over basis states
Readout	Reveals the stored bit	Produces a measurement outcome probabilistically
Copying	Can be copied freely	An unknown state cannot be perfectly copied
Correlation	Classical correlation	Can share entangled states
Operations	May be irreversible	Ideal quantum gates are reversible/unitary

For n qubits, a general state requires 2^n amplitudes to describe: two for one qubit, four for two, and eight for three. Yet measurement returns only n classical bits per run. Algorithm design must make that limited readout reveal useful global information. NIST warns that superposition does not permit efficient brute-force search of every answer (NIST).

Quantum superposition: more precise than “0 and 1 at once”

Picture a compass needle that can point in many directions rather than a coin that must show heads or tails. When you ask only “up or down?”, the direction determines the probabilities of those two answers. This helps convey that a qubit can contain more structure than one classical bit.

But the compass is still classical. A quantum superposition is not merely a hidden direction or an ordinary random mixture. It is a quantum state with amplitudes and relative phases associated with possible measurement outcomes. IBM’s learning material notes that saying a qubit is “both states at the same time” can mislead; operationally, the state lets us predict outcome probabilities and interference patterns (IBM Quantum Learning).

If a qubit has equal amplitudes for |0⟩ and |1⟩, repeated preparations and measurements produce an approximately 50–50 distribution. Each measurement yields only 0 or 1 and ends that run’s superposition.

Superposition by itself does not guarantee speed. A machine that prepares every candidate with equal weight and then measures usually obtains a random candidate. A useful algorithm must transform amplitudes so that informative outcomes become more probable.

Quantum interference: the engine behind useful answers

Drop two stones into a pond. Where wave crests meet, the water rises higher: constructive interference. Where a crest meets a trough, they can cancel: destructive interference. Sound waves behave similarly; noise-cancelling headphones create a carefully shifted wave that reduces unwanted sound.

Quantum amplitudes also interfere. Because amplitudes carry phase, two computational paths leading to the same outcome can add or cancel before measurement. A successful algorithm is choreographed to:

• increase amplitudes associated with useful outcomes;
• reduce amplitudes associated with unwanted outcomes; and
• leave a measurement distribution from which the answer can be inferred with high probability.

This is why “the computer tries every answer at once” is a poor stopping point. If all alternatives remain equally likely, nothing has been gained. The algorithm’s real work is designing transformations that make the alternatives interfere in a helpful pattern. IBM’s explanation connects phase to the alignment of waves and identifies interference—not classical randomness—as the source of otherwise unavailable measurement patterns (IBM Quantum Learning).

Quantum entanglement: correlations that belong to the whole

Imagine two sealed boxes prepared together. Opening either reveals a color, and the colors always match. That suggests correlation, but it is not yet entanglement: ordinary boxes could have contained predetermined matching cards all along.

In an entangled state, the complete quantum state cannot be separated into an independent state for each qubit. For the two-qubit state (|00⟩ + |11⟩)/√2, the pair has a well-defined joint description while neither qubit has its own pure state. Measurement outcomes are correlated in ways that cannot be reproduced by a local hidden list of predetermined answers for every possible measurement choice.

Entanglement becomes a computational resource because gates can create global structure across a register. Algorithms use that structure to represent relationships among variables, perform coordinated transformations, and generate interference patterns unavailable to independent qubits.

Entanglement does not enable faster-than-light messaging. The result seen by either observer alone is random. Only later comparison through an ordinary classical channel reveals the correlations. The box analogy therefore captures “connected outcomes” but fails to capture the absence of predetermined local values and the role of measurement basis.

Quantum gates, circuits, and measurement

Quantum programs are commonly expressed as circuits:

1. Initialize. Prepare qubits, often in |0⟩.
2. Apply gates. Use precisely timed microwave pulses, lasers, or optical components to rotate states and change phases.
3. Use controlled gates. Make an operation on one qubit depend coherently on another, which can create entanglement.
4. Measure. Convert selected quantum information into classical outcomes.
5. Repeat. Run the same circuit many times—often called shots—to estimate a probability distribution.
6. Process classically. A conventional computer aggregates results, updates parameters, or decides what circuit to run next.

Ideal quantum gates are unitary: they preserve probability and are reversible. Measurement produces a classical result and discards earlier-state information.

For a two-qubit example, start in |00⟩. A Hadamard on the first qubit creates equal amplitudes for |00⟩ and |10⟩. A controlled-NOT then flips the second qubit when the first is 1, producing an entangled state whose ideal measurements are roughly half 00 and half 11. The example demonstrates a joint state, not a useful application.

The processor is the isolated center of a larger system. Classical electronics generate pulses, lasers address atoms, cryogenic or vacuum equipment maintains the environment, software compiles circuits, and decoders interpret error signals. Quantum computers are hybrid systems from the beginning.

How a quantum computer is built

There is no single physical design called “the quantum computer.” Research groups are developing several architectures, and comparisons require more than one number.

NIST describes ions as comparatively long-lived but slower, superconducting circuits as fast but more fragile, and topological qubits as a direction still seeking definitive evidence (NIST). Neutral atoms, photons, and silicon offer other tradeoffs. No architecture has established universal superiority.

Why quantum computers are so difficult to build

Imagine conducting an orchestra in which every instrument is extraordinarily sensitive to vibration. The musicians must remain isolated from the audience, yet they must also hear the conductor and one another at exact moments. A slightly mistimed cue changes the music, and checking a note can end the performance. That is roughly the control problem facing quantum engineers.

Decoherence is the loss of useful quantum behavior as a system interacts with its environment. Heat, electromagnetic fields, material defects, stray light, vibration, and even high-energy radiation can disturb a state. Control pulses are imperfect. Readout makes mistakes. Neighboring qubits can affect one another through crosstalk. Calibrations drift over hours or days.

Architectures add their own burdens. Superconducting devices need extreme refrigeration. Ion and neutral-atom platforms need vacuum systems and stable lasers. Photonic systems must fight loss. Limited connectivity can force extra swap operations just to bring information together, and every added operation creates another opportunity for error.

The result is limited useful circuit depth: the number of sequential operations a calculation can survive before noise overwhelms its signal. Scaling means delivering control, routing signals, removing heat, calibrating many components, and decoding errors in real time—not merely fabricating more qubits.

Physical qubits, logical qubits, and quantum error correction

A physical qubit is an individual hardware element: one circuit, ion, atom, photon mode, or spin used by the machine. A logical qubit is protected information encoded across many physical qubits with an error-correcting code.

Classical error correction can copy a bit several times and use majority voting. Quantum mechanics forbids making a perfect copy of an arbitrary unknown state—the no-cloning theorem—and direct measurement would disturb the information. Quantum error correction takes a subtler route. It spreads information across an entangled code space, measures carefully chosen error syndromes that reveal what kind of error occurred without revealing the protected logical value, and applies or tracks corrections.

Surface codes arrange data and measurement qubits in a two-dimensional pattern and are prominent because they tolerate local errors and suit some chip layouts. Other families—including color codes, low-density parity-check codes, bosonic codes, and concatenated constructions—offer different tradeoffs in connectivity, rate, gate implementation, and decoding. There is no single universal overhead number.

An error threshold is the physical error rate below which increasing code size can reduce the logical error rate. Google’s Willow work reported exponential error suppression as its surface-code patches grew from 3×3 to 5×5 to 7×7, an important below-threshold result on that device. Google also stressed the large remaining gap between present logical error rates and those required for long algorithms (Google Research). It is a milestone, not a general-purpose fault-tolerant computer.

Fault tolerance extends protection to the entire computation: state preparation, gates, measurement, syndrome extraction, and classical decoding must prevent a small number of faults from spreading into an unrecoverable logical error.

Quantum algorithms: where speedups actually come from

A quantum algorithm needs a problem whose structure can be encoded and manipulated so interference exposes useful global information. Most programs do not acquire a speedup merely by being “run on qubits.”

Shor’s algorithm finds integer factors and discrete logarithms in polynomial time on a sufficiently capable fault-tolerant quantum computer. That threatens widely deployed public-key systems based on those problems, including RSA, Diffie–Hellman, elliptic-curve cryptography, and Bitcoin’s elliptic-curve signatures. The algorithm is proven; the attack-scale hardware is not available today (Shor’s paper).

Grover’s algorithm searches an unstructured space using about the square root of the classical number of oracle queries. This is a quadratic, not exponential, improvement. It motivates larger symmetric keys and changes how we discuss hash preimage security, but deep fault-tolerant circuits and parallel classical alternatives matter in real cost comparisons (Grover’s paper).

Amplitude amplification generalizes Grover’s idea: reflect amplitudes so that marked states become more likely. The quantum Fourier transform extracts periodic structure and is a central ingredient in Shor’s algorithm. Quantum simulation uses one controlled quantum system to model another, matching the physical structure of chemistry and materials more naturally than storing a complete many-body wavefunction classically.

Hybrid variational methods such as VQE and QAOA let a classical optimizer adjust a short quantum circuit. They fit shallower hardware, but noise, measurement cost, difficult optimization, and improving classical methods leave broad practical advantage uncertain. A benchmark need not be a cheaper real-world answer.

Quantum computing applications: promise, evidence, and open questions

Chemistry, materials, and catalysts

The problem: electrons in molecules interact quantum mechanically, and exact state descriptions grow exponentially. How quantum computation might help: represent and evolve those states on quantum hardware, potentially estimating energies, reaction paths, or material properties with less approximation. What remains difficult: chemically useful accuracy requires excellent gates, error correction, efficient state preparation, and algorithms whose total cost beats powerful classical approximations.

This could aid research on catalysts, batteries, fertilizers, carbon capture, or superconductors, not press a “discover material” button. DOE identifies molecules, nuclei, catalysts, and complex materials as quantum-simulation targets (U.S. Department of Energy).

Drug discovery

Drug research also involves biology, toxicity, delivery, manufacturing, and clinical evidence. Quantum chemistry might improve selected binding or reaction calculations, making it a research aid—not a promise of instant medicines.

Logistics and optimization

Routing and scheduling create large combinatorial spaces. Quantum algorithms or hybrid heuristics might exploit structure in some instances, but classical solvers are highly optimized and many proposals lack proven end-to-end advantage on realistic data. “Optimization” is a category, not a guarantee.

Financial modelling and risk

Finance uses sampling, optimization, and differential equations. Amplitude estimation offers theoretical query advantages for some Monte Carlo tasks, but practical value remains unproven once fault-tolerant costs, data loading, assumptions, and classical baselines are included. Quantum hardware cannot make a weak model truthful.

Fundamental science

Quantum simulation may help study many-body physics, high-energy processes, and nuclear systems that are difficult to represent classically. Even when a device does not yet outperform the best supercomputer on a useful calculation, experiments can improve control, measurement, and understanding of quantum matter. This is scientific progress, though it should not be confused with a commercial application.

Machine learning

Quantum machine-learning research studies kernels, sampling, and generative models. Some algorithms have theoretical advantages under restrictive assumptions, but loading classical data may erase them and classical AI keeps improving. “Quantum AI” does not automatically win.

Cryptanalysis and security

Cryptanalysis is the clearest high-impact application because Shor’s algorithm changes the asymptotic difficulty of major public-key problems. Yet the relevant machine must support many reliable logical qubits and a very large number of fault-tolerant operations. Present processors cannot do this. Security work therefore focuses on migration before the threat matures.

The likely model is a hybrid classical–quantum workflow: classical systems prepare, optimize, decode, and verify; quantum processors handle only subproblems that justify their cost.

Advantages of quantum computing

The advantages of quantum computing are specific:

• Natural representation of quantum systems. Quantum hardware can encode states costly to store classically, especially in chemistry and materials.
• Provable speedups for selected algorithms. Shor, Grover, and amplitude estimation improve asymptotic performance under defined assumptions.
• New sampling capabilities. Designed circuits can sample distributions difficult for known classical methods.
• Different problem structure. Phase, interference, and entanglement provide resources absent from ordinary Boolean circuits.
• Hybrid acceleration. A quantum processor may strengthen a classical workflow without replacing it.
• Scientific spillovers. Advances in lasers, cryogenics, fabrication, control, sensing, and decoding can matter beyond computing.

“Quantum advantage” should prompt two questions: on which task, and against which classical method? A tailored benchmark need not prove economic usefulness.

Disadvantages of quantum computing and present limitations

• Extreme sensitivity to noise. Useful states are fragile, and every operation can introduce error.
• Large error-correction overhead. One logical qubit may require many physical qubits and continuous decoding.
• Specialized usefulness. Most workloads have no known quantum advantage, and some “speedups” depend on unrealistic access to data.
• Hardware complexity and cost. Cryogenics, vacuum, lasers, electronics, shielding, and specialist facilities are expensive.
• Difficult programming and verification. Results are statistical, measurement constrains debugging, and hard outputs may resist classical verification.
• Limited practical advantage today. NIST notes that early demonstrations have not yet shown broad useful superiority and that classical methods sometimes catch up (NIST).
• Benchmarking and hype. Raw qubits or proprietary benchmarks can hide errors, connectivity, post-processing, and favorable comparisons.
• Energy and infrastructure. Refrigeration, lasers, controls, and data centers require full-system comparisons.
• Security disruption. Successful large-scale machines would weaken current public-key infrastructure and force long, costly migrations.
• Unequal access. Cost and strategic value may concentrate capability among governments and large firms.
• Schedule uncertainty. Fixed dates ignore unknowns in scaling, codes, algorithms, and manufacturing.

Myths and misunderstandings

Myth: Quantum computers try every answer and reveal all of them. A state can contain many amplitudes, but measurement yields limited classical information. Algorithms must use interference to concentrate useful probability.

Myth: A qubit is just a faster bit. A qubit is a different information carrier with amplitudes, phase, measurement disturbance, and possible entanglement. Its value depends on coherent control, not clock speed alone.

Myth: More qubits always means a better machine. Physical quality, topology, gates, speed, and logical performance determine usefulness.

Myth: Quantum computers will replace phones and laptops. Everyday computing favors cheap, robust classical hardware. Quantum processors are likely to remain specialized accelerators accessed through laboratories or cloud services.

Myth: Quantum computers can currently break all encryption. Public demonstrations are far below the scale and reliability needed to attack modern large public keys. Symmetric cryptography and hashes are also affected differently from factoring- and discrete-log-based systems.

Myth: Entanglement sends usable information faster than light. Local results are random; correlations require ordinary communication to compare.

Myth: Every optimization or AI problem will gain an advantage. Speedups depend on structure, data access, algorithmic proof, hardware cost, and the best classical competitor.

Myth: “Quantum supremacy” means classical computers are obsolete. It describes outperforming classical methods on one task, not replacing ordinary computing.

Where quantum computing stands in 2026

As of September 30, 2026, quantum processors run real circuits, but errors and limited depth constrain them. Below-threshold experiments have reduced logical error as code size increased. That is evidence that protection can scale under specific conditions, not operation of thousands of application-ready logical qubits for billions of gates.

The field is moving from protected memories and operations toward systems of logical qubits. Progress requires logical error rates, gate sets, runtime, decoding, connectivity, and sustained operation—not only physical-qubit totals. Cross-architecture counts are especially misleading because different qubits have different gates, speeds, and lifetimes.

Benchmark demonstrations also need context. Google described its Willow random-circuit-sampling result as a major benchmark and its error-correction result as below threshold; the company’s own account distinguishes this progress from the future goal of commercially relevant applications (Google Quantum AI). Company roadmaps are useful statements of intent, not neutral proof that deadlines will be met.

No publicly known quantum computer can currently break modern large-scale public-key cryptography or derive Bitcoin private keys from public keys. There is no evidence-based fixed date for a cryptographically relevant quantum computer.

The future of quantum computing

In the short term, expect better qubits, control, calibration, error detection, compilers, and small logical demonstrations. Cloud access and classical circuit simulation will remain important.

In the medium term, the decisive question is whether teams can scale logical qubits while lowering total logical error, expanding the fault-tolerant gate set, and keeping decoding and control manageable. Modular architectures may connect smaller processors rather than enlarging one monolithic chip. Quantum networks could distribute entanglement or link modules, although networking is a distinct engineering field—not a shortcut around local processor errors.

Longer-term systems may become supercomputing accelerators. Compilers must account for error-correction cost, and transparent standards should reduce dependence on each vendor’s favorite metric. The workforce must span physics, computing, engineering, chemistry, cryptography, and operations.

Cryptographic migration must begin before hardware can run Shor at attack scale because changing protocols, inventories, implementations, and operational workflows can take years. Advance planning is the practical response to an uncertain timeline; a roadmap is not a calendar.

The forecast is conditional. Better qubits may unlock useful simulation; scaling may stall or demand more overhead; new codes and algorithms may change the picture. Preparation is rational because the date is unknown.

Quantum computing and cybersecurity

Symmetric cryptography uses the same secret key, or closely related secrets, for encryption and decryption. Asymmetric cryptography uses a public key and a private key for encryption, key agreement, or digital signatures. The distinction matters because quantum algorithms affect them differently.

Shor’s algorithm can solve the factoring and discrete-logarithm problems underlying RSA, finite-field Diffie–Hellman, and elliptic-curve cryptography on a sufficiently large fault-tolerant machine. That would threaten key exchange and signatures. Grover’s algorithm gives a quadratic search improvement against idealized symmetric keys and hash preimages, usually addressed by choosing adequate key and digest sizes rather than abandoning symmetric cryptography.

The harvest-now-decrypt-later problem precedes an attack-scale machine: an adversary can retain quantum-vulnerable encrypted traffic and hope to decrypt it later. Signatures raise a different risk of future forged authorization.

Post-quantum cryptography (PQC) uses classical computers and mathematical problems believed to resist both classical and quantum attacks. It is not the same as quantum communication. In 2024, NIST finalized FIPS 203 for ML-KEM key establishment, FIPS 204 for ML-DSA signatures, and FIPS 205 for SLH-DSA hash-based signatures (NIST PQC overview, FIPS 203, FIPS 204, FIPS 205).

Migration must begin before a cryptographically relevant machine exists. Organizations need inventories, data-lifetime assessments, interoperable implementations, testing, vendor coordination, and crypto-agile designs. NIST encourages transition now so a multi-year replacement does not become an emergency.

Quantum computing and Bitcoin: the closing case study

Bitcoin makes the cybersecurity issue unusually visible because it is a public, long-lived system with no central operator who can quietly replace every key.

Bitcoin ownership is protected primarily by digital signatures, not by encrypting the blockchain. The ledger is public by design. A private key authorizes a spend, while nodes verify an ECDSA or Schnorr signature using the corresponding public key. Bitcoin also uses hashes for proof of work, commitments, and address constructions.

A sufficiently capable fault-tolerant quantum computer running Shor’s algorithm could solve the elliptic-curve discrete logarithm problem and derive a private key from an exposed secp256k1 public key. That would threaten both ECDSA and Taproot’s Schnorr signatures. No current public quantum computer can perform this attack.

Exposure is uneven. Some output types reveal a public key from the start; some initially reveal a hash and expose the public key when spent; address reuse can leave other coins associated with an already revealed key. An attacker targeting a newly revealed key during a spend would also face a timing race. These distinctions affect priorities, but hiding a public key behind a hash is not a complete long-term migration strategy.

Grover’s algorithm relates differently to Bitcoin’s hashes and mining. It offers an idealized quadratic search speedup, not Shor’s polynomial-time break. Hardware speed, parallelism, circuit cost, block timing, and Bitcoin’s difficulty adjustment all matter. It does not “destroy SHA-256 overnight.”

The hardest Bitcoin question may be coordination. A safer signature still needs review, wallet and hardware support, fee analysis, activation rules, and a migration path. Dormant coins create governance questions: leaving legacy spends valid could expose them, while restrictions could block a rightful owner. Cryptography cannot decide that policy.

For a dated, Bitcoin-specific assessment, read Bitcoin and Quantum Computing in 2026: What Is Real, What Is Still Theoretical, and What Comes Next. The Shor-versus-Grover guide compares the algorithms directly, and the migration analysis follows the protocol and governance work.

What to look for in a quantum computing book

A trustworthy resource should separate current capability from future possibility. It should explain superposition without claiming that measurement reveals every branch, identify interference as the useful mechanism, distinguish physical from logical qubits, and make uncertainty visible rather than hiding it behind a roadmap.

It should also cover benefits and risks together. Chemistry and materials deserve as much care as cryptanalysis. Quantum computing, quantum communication, and quantum sensing should not be blended into one vague technology. Claims should point to evidence, company performance statements should be labeled as such, and an honest explanation should say what remains unproven.

For security, it should treat cryptographic migration as a planning problem. Standards, inventories, implementation, governance, and lead time matter before an attack-scale machine arrives.

Bitcoin’s Quantum Countdown is not a general quantum-computing textbook. It focuses on Bitcoin’s quantum-security questions: how a future fault-tolerant machine could affect Bitcoin’s signatures, why hashing and mining are different questions, and why migration collides with ownership, incentives, and decentralized governance.

Sources

• NIST, Quantum Computing Explained, updated May 28, 2026.
• IBM Quantum Learning, Quantum Mechanics Basics and Basics of Quantum Information.
• U.S. Department of Energy, Quantum Information Science and DOE Explains: Quantum Computing.
• Google Quantum AI, Willow quantum chip and Making quantum error correction work.
• Peter W. Shor, “Polynomial-Time Algorithms for Prime Factorization and Discrete Logarithms on a Quantum Computer”.
• Lov K. Grover, “A Fast Quantum Mechanical Algorithm for Database Search”.
• NIST, Post-Quantum Cryptography, including FIPS 203, FIPS 204, and FIPS 205.
• Bitcoin Improvement Proposals, BIP 340: Schnorr Signatures and BIP 341: Taproot.