The Long Way Around

For most of my professional life, I assumed I had a collection of unrelated careers.

Over the years I worked in politics, professional kitchens, construction, technical education, developer advocacy, and technology leadership. More recently, I’ve found myself spending time on institutional memory, provenance, AI architecture, museums, archives, and historical research.

Looking at that list on paper, it feels random. A career counselor might call it a lack of focus. An Applicant Tracking System would almost certainly struggle to figure out what box to put me in.

For a long time, I viewed it the same way. Every career change felt like starting over. Every transition came with the uncomfortable feeling that everyone else had chosen a lane and stayed in it while I was wandering between industries.

It wasn’t until recently, while working on the Sovereign Systems Specification and updating my personal website, that I began to notice something unexpected.

The industries had changed.

The questions had not.

The Same Questions in Different Places

When I worked in politics, information was everything. Every statement, every policy position, and every talking point ultimately came down to a simple question:

“According to whom?”

When I worked in professional kitchens, the same question appeared in a different form. Recipes, inventory, supplier relationships, food safety procedures, and training all depended on knowledge being documented, shared, and trusted. Making the same “house ranch” recipe from memory isn’t the same as having it written in a procedure.

Construction wasn’t much different. Plans, permits, change orders, inspections, and customer agreements all relied on accurate information and a clear understanding of where that information came from. A builder who’s working from memory instead of the stamped plans builds the wrong room. That’s not a rounding error — that’s a tear-out.

Technology brought the same challenges into a new domain. Documentation, system architecture, databases, APIs, observability, and developer education all revolve around helping people understand complex systems and trust the information they are using to make decisions.

More recently, my interests have expanded into museums, archives, historical research, and AI systems. Yet even there, the same themes continue to emerge.

The questions kept reappearing in different forms:

  • How do people store knowledge?
  • How do they trust knowledge?
  • How do they lose knowledge?
  • How do they pass knowledge to the next generation?

The technology changes. The industries change. The underlying questions remain remarkably consistent.

A Phrase That Refused to Stay in One Project

While writing the Sovereign Systems Specification, I coined a phrase that I initially thought was simply a good line:

Information without provenance is just gossip.

At first, it was intended as a commentary on AI systems. Large Language Models are increasingly capable of producing convincing answers, but confidence is not the same thing as evidence. If an answer cannot be traced back to a source, its reliability becomes difficult to evaluate.

The more I thought about it, however, the more I realized the phrase applied far beyond AI. Every field where trust matters has a provenance problem.

The phrase kept showing up because the principle kept showing up.

Eventually I stopped thinking of it as an AI concept and started viewing it as a general truth.

Maybe It Wasn’t Several Careers

For years I looked at my resume and saw a collection of disconnected experiences.

Politics, culinary arts, construction, technology, education, and research.

The assumption was that these represented different chapters of my life.

What I’m beginning to suspect is that they were all chapters in the same story.

The industries were different. The tools were different. The job titles were different.

What remained constant was an interest in understanding how knowledge is created, organized, trusted, preserved, and shared.

Seen through that lens, the transitions no longer look quite so random.

Politics was about information and trust.

Kitchens were about process and knowledge transfer.

Construction was about documentation and accountability.

Technology was about systems and understanding.

Museums and archives are about preservation.

AI is forcing us to revisit all of those questions at scale.

The Questions That Follow Us

One of the unexpected benefits of getting older is that you eventually accumulate enough experiences to identify patterns that were invisible while you were living through them.

In your twenties and thirties, careers often feel like a sequence of decisions.

In your forties and fifties, they sometimes start to look more like a sequence of questions.

The jobs change.

The industries change.

The technologies change.

The questions worth asking tend to remain remarkably consistent.

Looking back, I don’t think I’ve spent thirty-five years working in a series of unrelated professions.

I think I’ve spent thirty-five years exploring the same problem from different angles.

And perhaps that’s the lesson hidden inside a long and winding career:

The most important thing you carry from one job to the next isn’t a title, a skill, or a technology.

It’s the set of questions you never stop asking.

That’s the foundation the Sovereign Systems work is built on.

<>
That’s the foundation the Sovereign Systems work is built on.

Facebooktwitterredditlinkedinmail

The Context Compression Pattern

Pattern Defined

Precise Definition: Context Compression is an inference pattern that utilizes
a specialized “selector” model or a ranker to distill large volumes of retrieved
data into its most salient semantic components, removing redundant or irrelevant
tokens before the final inference pass.

Problem Being Solved

We are currently fighting the “Lost in the Middle” phenomenon. Even with massive
token windows, LLM performance degrades significantly when relevant information is
buried deep within a context block; more data often leads to less accuracy.

For a Director of Engineering, this is a direct threat to the
Sovereign Vault’s
integrity. Every irrelevant token passed to the model is a potential point of
failure for privacy airlocks and data governance. As established with the
Sovereign Redactor,
minimizing the noise isn’t just about saving money—it is about shrinking the
surface area for hallucinations and privacy leaks.

Use Case

Consider an Archival Intelligence
system processing 1880s shipping ledgers. A single query about “cargo weights in
1884” might pull 20 pages of scanned text. Most of those pages contain sailor
names and weather reports that have no bearing on the weight data.

Without compression, the model has to “read” the entire ledger, leading to high
costs and potential confusion. With the Context Compression pattern, a smaller,
faster ranker identifies the specific sentences regarding “tonnage” and “cargo,”
passing only those 200 relevant words to the high-reasoning model. The Forensic
Auditor gets a precise answer in half the time.

Solution

The pattern typically follows a three-step pipeline:

  1. Retrieve: Fetch the top documents using standard RAG.
  2. Compress: Use a technique like LongLLMLingua (a token-pruning method
    developed by Microsoft Research) or a Cross-Encoder to rank and prune tokens.
  3. Synthesize: Pass the condensed, high-signal prompt to the final model.
flowchart LR
    A([User Query]) --> B[RAG Retrieval\nTop N Documents]
    B --> C[Compression Layer\nLongLLMLingua /\nCross-Encoder]
    C --> D[High-Signal\nCondensed Prompt]
    D --> E([Frontier Model\nSynthesis])

_The tree-step compression pipeline: retrieve broadly, compress precisely, synthesize confidently.

In an MCP or FastAPI-based system, this happens at the “Glue Code” layer, where
you programmatically filter the retrieval results before they hit the LLM’s prompt
window.

Trade-Offs

The trade-off is Latency in the Retrieval Step vs. Reliability in the Synthesis
Step. Adding a compression layer adds a few hundred milliseconds to your
pipeline, but it significantly reduces the final generation time and token cost.

From a leadership perspective, the risk is Over-Pruning. Tuning the “compression
ratio” to ensure the Forensic Auditor doesn’t lose critical edge cases is a new
engineering requirement—one that takes place in those two extra sprint cycles we
discussed in the series opener.

Summary

Context Compression is the difference between handing a researcher a stack of 100
books and handing them a one-page summary of the relevant chapters. It ensures
that your high-reasoning models only see what matters.

Next Up

In two weeks, we go deep on the Hybrid Retrieval Pattern and explore why your data needs a
map, not just a list.

Inference Pattern Series

Facebooktwitterredditlinkedinmail