Aion: the question that kept getting bigger

Aion started with a question that sounds ridiculous until you try to say exactly what would count as an answer: could I send a chosen message into the past?

The original target was small enough to write down and enormous enough to cause trouble. Thirty-two bits, readable roughly twenty-four hours earlier. A message somebody actually chooses to send. Not a prediction that happens to be right. Not a pattern you can only recognize after the answer arrives. [1]

That is the project I am spending my time on now.

After Chrono and NXTDeal, I wanted to follow a question that held my attention beyond the prospect of packaging another familiar software category. Aion certainly cleared that bar. It also produced a growing collection of separate tools, models and experiments. The attempt to answer one question keeps exposing others worth working on.

It started with the route. It became about the test.

The early work explored geometry, wormholes and the conditions a temporal shortcut might require. But drawing a possible route and producing a readable message are very different jobs. The research kept coming back to the receiver: what could that person actually know, and when could they know it?

The test is severe. Freeze an earlier record. Let someone make an independent choice later. Ask whether the earlier record contains information that distinguishes that choice, without borrowing anything from the future to interpret it.

Once you insist on that, a lot of attractive answers stop being answers. Selecting the histories that fit after the fact does not give an earlier observer a usable message. Putting a dependence on the future into the model does not demonstrate how nature would supply it. A clever decoder cannot rescue a signal you assumed into existence.

The research record includes corrected calculations, abandoned claims and candidate mechanisms that failed their controls. The November 7 audit reports that none of its ten tested candidates passed the frozen-record test. The next proposed direction asks whether a particular final-boundary hypothesis could be tested experimentally. That remains a proposed test, not evidence of a working backward-time channel. [1]

That distinction is part of the work. I want to ask an outrageous question and be able to inspect the answer. Dressing an assumption up as a breakthrough would make the entire exercise worthless.

CavityEDA: the engineering questions inside the big question

Start asking how information might be carried or preserved using light and you quickly acquire a pile of less cinematic problems. Loss. Timing. Noise. Scheduling. What happens when a design grows? Which physical details does a model keep, and which has it thrown away to make the calculation possible?

CavityEDA is the optical quantum-design branch of that work. It provides software for exploring cavity-based architectures, running modeled circuits and examining the resource costs before anybody builds hardware. Different state representations suit different kinds of circuits; choosing the right one is part of making exploration practical. [2]

I like that this has a job of its own. Someone can investigate a photonic design without sharing the original ambition that led me toward it. Its value has to come from useful models and comparisons. Whether a chosen message can reach yesterday is a separate question.

The same standard applies to performance. A fast reduced model is interesting because of what it lets you explore. It still needs to be clear about the physics it leaves out. A large speed number does not get to erase that distinction.

Vault: what survives, who opens it, and what happens when things go wrong?

Thinking about memory pushed the work toward preservation and recovery. Keeping a quantum state alive, retaining ordinary information and deciding who can open an archive are different problems. Separating them creates room for an application with a more immediate purpose.

AION Vault carries that branch forward. The work spans the earlier Quantum Time Vault research and a more concrete encrypted-file workflow: controlled opening, trusted-device approval, recovery and changing the protection around information after access. Its current presentation also explores a simulated black-hole pair as part of the lock model. The files themselves remain encrypted data; the simulation is not a physical container swallowing anybody's documents. [3]

That is the kind of strange idea I enjoy pulling apart. What does the metaphor contribute? What does the actual security mechanism do? Where does the authority to open the file live? Those questions need real answers before anyone should trust the result with something important.

SecureLink: delivery is only part of communication

AION SecureLink follows another branch: getting a message across and determining what happened to it on the way.

There are currently three distinct tracks in its public record: a reported simulation software package, an earlier protocol experiment and a proposed TLS session design. They are not interchangeable achievements. The package's reported tests do not resolve the earlier experiment's problems, and a proposed session design is not an implemented service. [4]

The interesting work is in separating delivery, checking and trust so their assumptions can be examined. Did the message arrive? What can the checking layer actually detect? What does noise do to that claim? A successful simulation only answers the question its model really asked.

I want the project to earn its claims as it develops. “Secure” in a name is not a force field.

Hive: a continuing life behind the conversation

AION Hive opens a different line of inquiry: persistent agents with bodies, needs, memory, attention and relationships inside a simulated world.

What interests me is what continues between replies. An agent encounters something. Its internal state changes. That experience influences a later choice. Attention has to be shared between competing needs. Relationships acquire a history. The work becomes a system you can perturb and compare, rather than a conversation you judge by how convincingly it talks about itself.

The current Hive presentation introduces Analyst and Empath, along with experiments around embodied affect, shared attention and a model of attention itself. The ambition includes consciousness research. The reported behaviors do not establish consciousness, and fluent language is not a shortcut around that problem. [5]

This branch does not depend on making backward communication work. It belongs here because Aion has become a place to investigate information and intelligence as well as time. There is plenty to examine before anybody gets to declare that they have manufactured a mind.

The Lab is where the side questions accumulate

AION Lab organizes the research; AION Studio runs local experiments. The public Lab connects the research story to an experiment ledger, models and a backlog of possible applications. [6]

Some directions grow directly from the main work: tools for auditing proposed mechanisms, time-controlled release for vaults, and hosted simulation jobs. Others take a method exposed by the research and ask whether it has a use elsewhere.

The named discovery concepts include clinical-trial allocation, credit-risk calibration and scientific evidence integrity. There are also proposals around manufacturing anomalies, asymmetric quantum-error mitigation, autonomous-vehicle safety audits, multi-agent learning, device authentication through Bell-correlation tests, error-correction optimization and causal consistency in smart contracts. These are candidate directions in the registry, not ten finished products. [6]

I do not want a list of names to become a trophy cabinet. Each idea needs a specific problem, something to compare against and a test that can disappoint it. Some will go nowhere. Good. Finding that out is more useful than giving every sketch a logo and pretending I have founded another company.

This is where I want the time to go

Aion has enough uncertainty to keep me interested and enough concrete work to keep that interest honest. There are models to inspect, experiments to design, interfaces to build and assumptions to break. A failed route can still leave behind a useful method or expose a better question.

I am not going to turn that into a motivational lesson about following your passion. I have spent enough time around people who can convert any experience into a LinkedIn carousel.

I wanted to stop rebuilding things I already understood well enough to see the next several years of maintenance coming. Now I have a question I cannot casually answer, several related projects with their own reasons to exist, and a public place to show the work as it changes.

That is a much better use of my curiosity.