I believe there is a way to write code so structurally sound that edge cases need not exist. And that structure, so translatable, can be internalized by any AI model with a requisite level of intelligence.

I may go as far as saying ... if the code for it cannot be elegant, the thing is not worth building. What I'm after is structural fit.

I don’t need any new software in my life and I don’t have new problems that software can solve. And I’m no longer inspired by software that helps me save a bit more time or money or stress. That stuff is saved enough.

I more fancy something immersive and pure ... where I can research, create, experiment, express from the highest levels of clarity. Software that feels inevitable in my life... this is what I aim to build.

The meaning of elegance

I want every permitted situation, failure included, to follow from the same coherent logic. Say, a system in which rare things are no less structurally accounted for than ordinary ones. A beautifully integrated state.

If I could eliminate the need for special-case reasoning, I would. And then I'd make the overarching structure Flowable between artificial intelligences. This is not prompt engineering.

Most software .... the code does work, but by accumulation; it solves the moment but compromises the structural integrity of the whole. There is little structural fit, little integration, little elegance ... and with AI ... now a reliance on orphaned instructions. This is a second-rate line of attack; it leaves one too dependent on the brute force of super intelligence.

The tools we have today reach so finely into the heart and structure of the software that code and its associated tests should render a lot of documentation obsolete. Few things are worth writing down.

I think writing should preserve what the executable structure cannot express. Such writing, to me, is an artifact of taste, quality, integration, and inspiration. Also, Structure and Flow.

Structure and Flow

In the realm of elegant software, fundamentally, there is Structure, and there is Flow. They are closely related; Flow is the turnover of relationships composing a Structure.

Practically then, one must arrive at a set of Structures and Flows that preserve the invariants through time. These invariants are identified at the outset or discovered as the design matures. What are you looking for ... for it to "just work" or for something more than that? This will inform the shape of your Structures and Flows.

When building software this way, I find myself no longer having to restate the same rules as local checks and repairs; they are already embodied in how the system is constructed. It is more about creating the environment wherein Structure can Flow in and out, seamlessly, and with limited unforeseen side effects.

Here:

• $I$ is a capital letter chosen to stand for invariant; it means that the current Structure satisfies the condition we require every reachable Structure to preserve.
• $s$ is the Structure, $f$ is the Flow
• $\delta(s,f)$ is the next Structure after applying the Flow

So it follows that the invariant I am after must be protected through Structure and Flow. This is how I aim to keep my own software (PerillaCove) immune to deadweight accumulation known as features.

Possibilities naturally cluster, but within each cluster is not a possible but an inevitable Structure; this is applicable to life as much as it is to software.

Finally, with anything I add, the goal is not merely to preserve the existing Structure, but to further crystallize the elegance I am after. This is my kind of growth.

The hash map

Just last Saturday (10/10/2026), it dawned on me: it’s high time I harness the hash map again. I loved the hash map since I was 22 years old. But amidst building like a mad man with agents, I had regrettably (and foolishly) forgotten about it.

The hash map was the oldest means through which I'd give my codebase structural soundness. Simply, I'd have it store relationships that need repeat access. Call it an index, a cache, whatever else; those are just labels.

The essence does not change, and that is this: the system I build must, based on stable identity, give direct access to key relationships it frequently needs.

Let there be a single source of truth. Fundamentally, this is about Structure. We can always derive transient views from Structure and apply them in the directions in which the system wants to navigate them. These derived views need a clear lifetime, and when their underlying truth changes, they must be updated, rebuilt, or discarded. Their correctness follows from that relationship to the source.

Efficiency then follows from elegance: the system supports the questions it needs to answer by matter of Structure.

Timeless testing

In the realm of testing software, there are a few invariants I have found that stand the test of time.

1. The suite should be approached with the rigour of a mathematical argument.
2. A passing result cannot certify a test's soundness.
3. A test that has never failed may not exercise the behaviour it claims to protect.
4. Meaningful behaviour across the whole system, including how its parts interact, must be tested.
5. Gaps lie in the assumptions and the omissions.

It is about having the confidence to continuously add and remove code while keeping the system elegant, integrated, and in a shippable state at all times. Speed, then, comes as a side effect.

I love to find the bridge between seemingly-distinct islands. But it must remain testable ... for finding some conceivable connection is not enough. This is because ... given enough freedom, I can tell a story connecting almost anything to anything.

Therefore, if a pathway carries an offshoot, disrupting that pathway should affect the offshoot in the predicted way. This connects back to my own testing principles ... a passing result does not establish that a test is sound, and interactions across the system matter!

And tying this back to Flow ... a connection then, in super practical terms, can be thought of as the relationship between what the inputs and outputs of a function. Plain and simple.

Moat or die trying

What on Earth do I need new software for, if not to expedite my quest to learn the secrets of the universe? That’s about the only objective now.

What is the metric by which you measure your software to be better? Is it faster, cheaper, bigger ... things of that sort? I think the entire point of AI is to stop building this type of software. The data we’ve accumulated, AI has been trained on all of it to now become your genie in a bottle, along with the six senses you already have.

Every company will be dying, literally, to defend its moat. A moat, by definition, is a barrier … something defensive. Corporations must have it because of the game they are playing.

It’s natural for an organization to think and hope that its data is its secret sauce … whether true or not. And an organization has expectations and threats. In response, it creates metric-driven goals, and reinforcement learning within private context fits this well.

However, companies that spend tokens to gain so-called productivity in the form of files, worksheets, and process will be spinning in circles and they won't realize it. AI accelerates the iteration of transient form while never touching the invariant, further confusing the confused.

Having said this, the AIs are very capable. The product of their thinking can be tapped into for instant use provided that thinking was kicked off on the right foot, because beware, for an LLM can rapidly implement features.

But if there’s no unifying vision, no integrating force, no structurally-sound steering, the agent basically functions as a steroid, the code equivalent of neglecting the connective tissue while hypertrophying muscle, putting the system on a devastating collision course.

Lastly: while you can copy logic and code, you can't copy structural fit of a brand as a Whole ... the Whole that makes you feel a particular way.

It’s only you who stands a chance

AI is way more of a gift to the individual than it is to the organization.

The thing about AI is it allows the elegance of code and the elegance of software to go hand-in-hand, if one is passionate about ensuring such things. But many teams do not truly care about what they are building. So AI magnifies this indifference. It will blow up.

I can orchestrate my own integration of islands with AI doing the hard labour. Still, I do think knowing the ins and outs of the architecture, especially since I care about what I am building, is vital. All of it is my world, and I am the game designer.

The enterprise is busy attempting to produce commodities that have a hundred other producers and whose equilibrium profit is zero, all while being subject to ‘enterprise pricing’ aka predatory plans suited specifically for their bloat of which there is no cap. It says "we are hiring" but has no idea what it is truly hiring for.

I, unlike a conglomerate, at least stand a chance at correctly determining what is worth building and what is not.

Seeing my life for what it is and creating a system, integrated across domains, to attain peak existence ... this is the most viable option. Now that a grunt part of software is solved, I'd rather not waste precious seconds doing and building things that don’t inspire me.

We are not there yet, but on the way to cheap and abundant superintelligence ... let the highest model using my highest ideals make integrated carvings in the code that are then dotted and crossed by the lightspeed lightweight.

Emptiness

Today you are robbed of Emptiness, that is, the space that allows for specific emergences to orient your life in a uniquely correct way. What you are instead filled with is a bubble called the ‘feed’ that expands to block all serendipity, yet it presents as Emptiness through its lack of substance.

There is no World in a feed. What there is ... is a force acting on you ... to get you to think and act like a corporation not a Human. They’ve got you internalizing the corporation. They’ve got you asking questions like ‘what is my edge’, instead of looking to attain the most integrated form of yourself this very second.

In the end, however, there are no secrets capable of being held except for the pellucid space forming the cytoplasm of your soul.

The reality is we didn’t need most of traditional software anyway; it was largely an extension of the corporate world that is itself useless and anti-nature. People are out here transforming data that never should’ve been created in the first place. And the mega-funded hyper-scaled technocrats have goaded you into becoming a databrain, when all you ever wanted to do with software was craft a story and make a movie.

Of course, you could have it your way. Nature’s logic is neither explicit nor linear but rather emergent and asynchronous. And this is how software can be too. A system that holds an essence and lets you collide with it in ways neither of you fully predicted. The way a bird surprises you with the exact curve it takes while still, unmistakably, being a bird. It could really be like that.

I’m coming to find that building PerillaCove may be the most effective way to convey the totality of the world inside of me. I can enter it again and again and still create and discover new connections between seemingly-distant islands.

Agency for me comes from orienting my life toward the Way, you know, nature. Attempting to understand it is itself a handful ... but it demands all of me and that's what makes life exhausting in a good way.