Programming
Programming news
What is happening in reversible and zero-energy computing — chips that recover their own switching energy, the field's slow walk from the lab to the data centre, and the people who started it all. The ten most recent stories are below — browse the full archive by year.
Articles
Reversible Computation 2026 convenes in Torino
The field's main annual conference returns on 9–10 July, spanning reversible languages, circuits, quantum computing and the thermodynamics of computation.
ProgrammingVaire moves to its second reversible chip
After demonstrating net energy recovery on silicon in 2025, the startup turns to a second chip aimed at logic and competitive performance, with a sharper prototype due in 2026.
ProgrammingSuperconducting circuits make momentum a resource for reversible logic
Coupled quantum-flux-parametron gates use a momentum-driven erasure-flip operation that beats conventional metastable switching on speed and fidelity at no extra energy cost.
ProgrammingCommunications of the ACM profiles reversible computing as an energy fix for AI
A CACM news feature explains how reversible processors that uncompute their operations, instead of erasing data, could rein in the electricity demand of large AI models.
ProgrammingIEEE Rebooting Computing puts reversible and adiabatic computing centre stage
The tenth ICRC convened in San Diego, listing reversible, adiabatic, ballistic and cryogenic computing among the energy-efficient paradigms it exists to advance.
ProgrammingNew semantics treats reversible computation as ordinary causal computation
Aubert and Krivine extend causal models of concurrency to reversible systems, showing reversibility amounts to a clean swap between the roles of conflict and causality.
ProgrammingA classic matroid problem is shown to fit in catalytic logspace
Linear matroid intersection — solvable for fifty years but never tightly classified — is placed in catalytic logspace, a model that borrows a full memory and must restore it exactly.
ProgrammingPhysicists propose chips that relax the constraints reversible logic removes
A multi-institution team argues that chips harnessing intrinsic physical dynamics — rather than enforcing statelessness and unidirectionality — could break the energy limits facing AI hardware.
ProgrammingReversible Computation 2025 proceedings published after Odense conference
The 17th International Conference on Reversible Computation gathered researchers in Odense, with Springer publishing peer-reviewed proceedings on reversible languages, circuits and semantics.
ProgrammingHow AI researchers could save energy by computing backwards
A Quanta feature argues that reversible computing — running calculations backwards instead of erasing bits — could sharply cut the heat and energy that AI workloads waste.
From the web
Selected reading and talks on reversible, invertible and bidirectional computation from around the web.
Spine: a bidirectional, self-rewriting live language ↗
A LIVE 2025 talk on Spine, a declarative language where editing an expression's output flows backward to its inputs under a "conservation of information" rule, and a reverse parser rewrites the source to match — closing the loop between a running app and its code.
Catalytic computing taps the full power of a full hard drive ↗
Quanta on catalytic computing: a full memory can still be borrowed as scratch space, as long as every borrowed bit is restored to its original value by the end — a reversible discipline that recovers usable computation from apparently occupied storage. Hacker News discussion.
Inverting the Xorshift128+ random number generator ↗
A worked exercise in program inversion: Scott Contini reconstructs the internal state of JavaScript's Math.random() generator from its outputs in about 2²⁶ operations by running its XOR-and-shift logic backwards, instead of brute-forcing 2¹²⁸ states.
BlogOn Dualities ↗
Lucas DiCioccio on the dualities that run through software — producer/consumer, sum/product, the De Morgan and distributive symmetries — and how recognising these reversible structural equivalences lets architectures be flattened and refactored predictably.