Hard science fiction is defined by a particular constraint: the scientific speculation must be grounded in actual science, or at least in a coherent extrapolation from it. The physics, the biology, the information theory — these are not decorative. They are the premise, and the story is what happens when the premise is taken seriously. The best hard SF is not a lecture with characters. It is fiction that could not exist without the science, fiction in which the ideas generate the drama rather than illustrating it.
The canonical titles in the genre include works by Hal Clement, Arthur C. Clarke, and Greg Egan, whose Permutation City and Diaspora remain the most philosophically demanding SF published in English. Kim Stanley Robinson's Mars trilogy is hard SF at its most ambitious in scope — terraforming as a problem in atmospheric chemistry, biology, and political economy, all of it grounded in research. Vernor Vinge's A Fire Upon the Deep introduced the concept of Zones of Thought, one of the most elegant hard SF premises of the 1990s. And Andy Weir's The Martian, whatever its literary limitations, is the most rigorous recent example of survival-as-engineering-problem in the genre's history.
What distinguishes the best hard SF from merely competent hard SF is a question worth spending time on. Clement's Mission of Gravity features a planet with extreme gravity at the poles and nearly none at the equator — the entire plot is generated by the physical properties of the setting. Egan's fiction tends to go further: in Diaspora, the posthuman characters are defined by what it means to have a substrate-independent existence, and the plot is a consequence of working out what that actually entails in a universe with the physics of ours. The science is not incidental. It is the character.
Allo's shelves lean toward literary fiction and natural history, but the book that comes closest to the spirit of hard SF in the catalog is Gödel, Escher, Bach by Douglas Hofstadter. It is not science fiction, but it is the kind of book that hard SF aspires to be: a rigorous intellectual structure in which ideas about formal systems, self-reference, and emergence are explored through the interplay of multiple registers — fugues, dialogues, theorems, visual paradoxes. Hofstadter is doing formally what hard SF does narratively: taking a set of real intellectual constraints and finding what space they open up. If you've been reading Egan and want something adjacent in spirit but different in form, this is it.
The tradition's weaknesses are worth naming too. Hard SF has historically been strong on concepts and weak on character — the drama of ideas tends to crowd out the drama of people. The genre has also been shaped by a demographic that has not always been curious about perspectives outside it, which has limited the range of premises it explores. The newer generation of writers working in the tradition — Becky Chambers at the softer end, Arkady Martine and Yoon Ha Lee at a harder formalist end — are correcting both of these. A Memory Called Empire by Martine is not hard SF in the Clement sense, but it takes seriously the epistemic and political problems of empire in a way that earlier hard SF rarely did.
What the genre offers at its best is a particular kind of cognitive pleasure — the satisfaction of understanding how something works, and then of watching the full implications of that understanding be traced through a narrative. It is the feeling of competence extended into imaginary space. The best hard SF makes you feel smarter, not because it flatters you, but because the rigor of the premise requires you to actually follow the argument. That is a rare thing in fiction, and worth seeking out.