The question of how life began has never been simple, and a new study from Imperial College London adds both depth and controversy to one of the most difficult problems in science. Published in July 2025, Robert G. Endres’ paper titled The Unreasonable Likelihood of Being: Origin of Life, Terraforming, and AI confronts the mathematical and informational challenges of abiogenesis while placing serious consideration on alternatives such as directed panspermia and terraforming. Rather than dismissing these ideas as speculative fiction, Endres presents them as logically viable scenarios that must be weighed against the staggering improbability of life self-assembling under the chaotic conditions of the early Earth. His analysis, rooted in information theory and algorithmic complexity, suggests that while life could have emerged naturally, the informational requirements are so extreme that outside intervention cannot be excluded.

Endres frames the problem by returning to biology’s oldest riddle: all cells come from other cells, but where did the first cell come from? Either it emerged from chemistry and physics alone or it was seeded from somewhere beyond Earth. The first option, spontaneous abiogenesis, remains the favored explanation within mainstream science, yet its details are mired in uncertainties. The second option, long marginalized but never eliminated, is that life was delivered or triggered by an advanced intelligence. Here Endres positions directed panspermia, first proposed by Nobel laureate Francis Crick and Leslie Orgel in 1973, alongside the concept of terraforming, noting that what once sounded like science fiction is now a real topic of debate within human scientific literature as plans to engineer Mars and Venus are discussed.

The core of the paper is not cultural but computational. Endres applies rate-distortion theory from information science to estimate whether the early Earth had enough informational throughput to cross the threshold from chemistry to biology. He calculates the entropy of the prebiotic chemical soup and compares it with the informational complexity of a minimal protocell. The results reveal an immense gap that only closes if molecules persist long enough, if information accumulates efficiently, and if some form of structured environment biases the process toward retention rather than random dissipation. A chaotic soup alone is too lossy. Without compartments, cycles, or autocatalytic sets, information would degrade before any organized system could stabilize.

The numbers he derives are striking. A minimal protocell might require on the order of a billion bits of information to achieve structural and dynamic sufficiency. By contrast, the background entropy available from the prebiotic environment is vast, but capturing and stabilizing even a fraction of it is extremely inefficient. He describes it as standing in a melting library of ten million books, each self-destructing after 24 hours, while needing to scan a hundred books every second to preserve enough material to construct a manual for building life. The analogy underscores how improbable it is for blind chemistry to accumulate meaningful order before degradation erases the progress.

In an optimistic framing, Endres shows that over half a billion years the required two bits per year could be gathered if processes were persistent and directional enough. But this relies on assumptions that may not hold. If information accumulation followed a random walk, the timescales balloon to cosmic impossibilities, hundreds of trillions of years longer than the age of the universe. Only strongly biased and memory-retaining processes could succeed. This raises a central dilemma: what provided that directionality? If it was not inherent to Earth’s geochemical cycles, then the case for external intervention strengthens.