Echoes of the pre-universe.

1. The mission, stated plainly

I've written about the window, and about the job I couldn't do before. This essay is about why. What is all of this for — the anomaly catalogs, the GPU pods, the agents, the platform itself? It is for one question, and I want to state it without hedging:

I am looking for evidence that our universe is not the first one. Fingerprints, ripples, echoes — anything that survived from a universe that existed before ours.

That sentence sounds like science fiction, so let me immediately ground it. The standard story says everything began with the Big Bang: a singularity, before which the question “what came before?” is supposedly meaningless. But the singularity isn't a discovery. It's a breakdown — the point where general relativity stops being able to describe what's happening. “The math diverges here” and “time began here” are very different claims, and we've been letting the second one ride on the first for almost a century.

There's a family of cosmological models — bounce cosmologies — in which the Big Bang wasn't a beginning but a transition. A previous universe contracted, reached an extreme but finitedensity, and bounced into the expansion we live inside of. In the framework I work in, Einstein-Cartan gravity, the bounce isn't inserted by hand: the spin of fermions generates torsion in spacetime, torsion produces an effective repulsion at enormous densities, and contraction reverses into expansion before a singularity can form. No exotic matter required. The singularity never happens, which means the question “what came before?” never stops being physics.

And if there was a before, the most important question in cosmology becomes an observational one: did anything survive the crossing?

2. Why “echoes” is the right word

A bounce is not a clean reset. It's a violent compression, but it is a physical process, governed by equations, and physical processes leave traces. Perturbations laid down during the contracting phase don't get erased at the bounce — they get processed through it and imprinted on the other side, in the statistics of the matter that eventually becomes galaxies, in the polarization of the first light, in gravitational waves that have been propagating ever since.

That's what I mean by echoes. Not a photograph of the previous universe — that's almost certainly gone. But the way a room's shape survives in its reverberation, the contracting phase should survive, statistically, in ours. The signature of a bounce is not one smoking gun. It's a pattern of small deviations from the inflationary story, each individually dismissible, collectively pointing the same direction.

The crucial point: these aren't philosophical claims. They're numbers. They differ between “the universe began” and “the universe bounced,” and several of them are measurable with data that is public right now.

3. Five fingerprints

Here is what the search actually looks like — the five channels where a pre-universe could have left prints, and where our lab's work currently stands on each:

None of these, alone, proves anything. A 3.6σ birefringence anomaly could be a systematic. A Bayes factor depends on its priors. A 2.3σ w-crossing is the kind of thing that evaporates in the next data release. I know all of this, and I'd rather say it before a reviewer does.

But here is what I keep coming back to. The inflationary story has to explain each of these anomalies separately — a coincidence here, a systematic there, an astrophysical source for the rest. The bounce framework predicts the directionof all of them from one mechanism. When one model needs five excuses and another needs zero, that doesn't make the second model right. It makes it worth a serious, systematic, full-time search. That search is my mission.

4. Honest accounting

A mission like this attracts cranks, so let me be precise about what we claim and what we don't.

We have published null results and we will publish more. We classified 8.47 million galaxies for chirality — a handedness asymmetry would have been a stunning hint of a preferred direction inherited from before the bounce — and we found nothing: f_cw = 0.5012 ± 0.0006, consistent with a perfectly symmetric universe at 0.4σ. That result went into the catalog and the paper anyway, because a search that only reports hits isn't a search, it's a marketing department.

We grade our own contributions on a novelty scale, N0 through N4, where N4 means “established new physics.” We have never claimed an N4. Not once. The strongest claims in the lab sit at N3 — novel, parameter-free, falsifiable predictions awaiting data — and they are deliberately underrated, because the failure mode in this field is overclaiming, and the only inoculation is a habit of underclaiming.

And every claim is built to die. The f_NL prediction is the clearest example: -35/8, no free parameters, no knobs to turn afterward.When SPHEREx delivers its constraint — the forecast says by 2027 — the prediction is either within errors or it isn't. If it isn't, the matter-bounce framework as we've formulated it is wrong, and I will say so in public, on the same site where I made the prediction. That's the deal. A mission to find the pre-universe is only science if you specify, in advance, what would convince you it isn't there.

5. How a software engineer hunts for a previous universe

I'm not a trained physicist, which I've written about before and won't relitigate. What I'll add here is that this particular mission is unusually well suited to the way I can work.

The echoes, if they exist, are buried in public data — DESI spectra, Planck maps, pulsar timing residuals, galaxy surveys. Finding them is not primarily a theory problem. It's a pipelineproblem: download everything, model everything, score everything, cross-match everything, and treat every anomaly as a lead until it's explained. That is industrial-scale data engineering wearing a cosmology costume, and it's exactly the kind of work AI agents transformed from impossible-for-one-person to routine.

So that's how the lab runs. Agents write and maintain the pipelines. An H200 in the cloud does the heavy inference. Cross-provider reviewer agents — GPT, Gemini, Grok, a dedicated skeptic — attack every claim before it ships. Three months of this produced four papers, 53 experiments, and the catalogs above, on roughly $400 of compute. The bottleneck is no longer labor. The bottleneck is judgment: which echo to chase next. That's the job I actually want.

6. The before, and the point

Somewhere in the data we already have — in 17.65 million spectra, in the timing of dead stars, in the faint twist of the oldest light — there may be a signal that was generated before the Big Bang. Not “shortly after.” Before. Information from a universe that contracted, crossed, and became ours.

Maybe it's not there. Maybe the bounce is wrong, or the bounce is right but scrambles everything, and the prints are unrecoverable. That would be a real answer too, and the program is built to deliver it: every prediction falsifiable, every null result published, every catalog public.

But if even one of those fingerprints holds up — if f_NL comes back at -35/8, if the birefringence angle stops being an anomaly and starts being a prediction — then the deepest question humans have ever asked, what came before the beginning?, stops being theology and becomes measurement.

And here is why I'm optimistic rather than daunted: for the first time in history, the tools to actually look exist, they are cheap, and they are in my hands. The answer isn't centuries away anymore. It is the next few data releases — SPHEREx, DESI DR2, CMB-S4 — and I get to be holding the instruments when they arrive. That is the most thrilling position a curious person has ever been in.

The previous universe, if it existed, left the evidence somewhere. My mission is to be the one looking when it surfaces.