What changed in each release. Old versions stay available at all versions.
The first update. Five backends instead of three, a real factoring algorithm, a standard library, a debugger, a way out to real hardware, and a way in from the rest of the quantum world.
qsparse.py, backend sparse). The dense statevector caps out around 20 qubits (a 25-qubit state already costs 51.8 seconds and 268 MB to build; 30 qubits needs ~8 GiB of pointers alone). The sparse backend stores only nonzero amplitudes, so programs whose state stays sparse go much further: a 50-qubit GHZ state builds and measures exactly. Dense 20-qubit GHZ: 9.3s. Sparse 20-qubit GHZ: 0.0000s. Use it with backend sparse as the first line of a program, or qinterp.py --backend sparse program.qc. Honest limit: it only wins when the state stays sparse — a dense superposition over all 2^n basis states is not helped.qshor.py). Real quantum order-finding via phase estimation on the exact statevector engine, then classical continued fractions to the factors. Factors 15 → (3,5), 21 → (3,7), 35 → (5,7), the largest on 16 qubits, order 12 found in 1 shot. Run it: python3 examples/shor_demo.py (about 3 seconds for all three). This is a simulator demo on small numbers — it is not a threat to RSA, and it claims no wall-clock speedup.qlib.py). One-line calls for the classic algorithms: grover_search, qft / iqft, phase_estimation, deutsch_jozsa, bernstein_vazirani, vqe_simple. Verified numbers: Grover finds 1 in 16 in 3 oracle calls (success probability 0.9613, vs 8.5 classical calls); QFT→IQFT round-trip error 4.44e-16; T-gate phase 0.125 exact; Deutsch–Jozsa and Bernstein–Vazirani in one query each (secret "1010"); VQE ground energies −1.118034 and −3.000000, exact. Examples: examples/qlib_grover.qc, examples/qlib_demo.py.qqasm.py). Export Q-Code circuits to OpenQASM 2.0 and import QASM circuits back. Supports h/x/y/z/s/t/sdg/tdg/id/rx/ry/rz/p/u1/u/cx/cz/ch/ccx/cu1, measure, reset, barriers, user gate definitions, and angle expressions in multiples of pi. A Bell pair round-trips exactly; a hand-written QASM Bell run gives 97×00, 103×11, 0×01/10 over 200 shots.qibm.py, backend ibm). OPTIONAL. A user with their own IBM Quantum free-tier token can run Q-Code programs on real hardware. Inert by default: no token means a clear error and no network access. The token is read only from the QCODE_IBM_TOKEN environment variable at runtime — never a command-line flag, never written to disk, never printed. No IBM account was created for this release and no real-hardware job was run. Free tier is 10 minutes of QPU time per 28-day rolling window (confirm on IBM's pricing page — terms change). Setup steps: docs/ibm-backend.md. show peeks are refused on this backend (each peek would be a separate hardware job); use ask.qdebug.py). Step through a program showing the quantum state at each step, with breakpoints and step limits. Works on the field, state, sparse, and quantum backends. python3 qdebug.py program.qc [--backend state] [--break 12,20] [--limit N]. A Bell program traces 7 steps, 50/50 amplitudes, collapse to |00⟩ on ask. The debugger does not support the IBM backend on purpose — debug on Aer, then run on IBM, so a debug session can never submit a hardware job by accident.examples/killer_demo.qc, examples/killer_demo.py). The wave-memory showcase. The secret "WAVES REMEMBER" (112 bits) becomes a superposition of standing-wave modes; one point of the wave is a bare float that holds nothing; reading from only half the points reconstructs the message exactly. The message was never anywhere. It was everywhere.qcode-lang 1.1.0). pyproject.toml, wheel + sdist in dist/, verified clean install in a fresh venv with no network and no Qiskit. pip install qcode-lang (field/state/sparse, no dependencies), pip install "qcode-lang[qiskit]" (quantum backend), pip install "qcode-lang[ibm]" (IBM backend). PyPI publishing is prepared but NOT done yet — the cloud site stays the distribution channel for now. The name qcode-lang is free on PyPI (checked 2026-11-05; qcode itself is taken by another author).376 of 380 tests pass across 42 suites. The only failures are the same 4 known fixture-state failures in the parked vault module (test_qvault.py), unchanged from v1.0 — the vault is parked and out of scope for this release.
The public release. The language, three backends (field, state, quantum), wave memory, Q-RAM + Grover, GhostDB, BB84, compression, superdense coding, photon gates, many worlds, shared QRAM, the HTTP server, and the synthesizer. 276 tests, 34 suites, all green. Documentary: YouTube.