Introduction
Welcome to the frontier. The year is 2026, we’re now closer to 2100 than to 1950, and closer to 2050 than to 2000.
Thresholds have been crossed, and the age of AI has begun. What’s now possible is steadily changing. Individuals and organisations are figuring out their position on AI, what’s going on, what matters, how to best leverage this technology and its failure modes.
The software and product development lifecycle is being disrupted. Efficient highly leveraged teams, seems to be the received wisdom for engineering leaders across organisations.
Aye, the act of coding has been commoditised.
The value of being able to close off issues through translating requirements into syntax has plummeted. While the art of programming computers, that is articulating what you want to achieve, and how, combined with product and design first thinking is more leveraged than ever.
A vast unexplored continent awaits. It’s not clear what the fog of war will reveal, we’re still early, but how tech teams work today is changing into something completely different even from a few years ago.
What it means to develop products, and what products will become valuable in this new land is still an open question. Yet human desire is infinite, when you remove one constraint we reliably find more. When it becomes possible to do more, we’ll do more.
At the same time software development is opening up to more and more people, today for rapid prototyping and experimentation, tomorrow for shipping features and products within expertly harnessed codebases.
In this world, it’s those who have the judgement on what is worth building, how best to build it, and how to maintain and evolve it over long time horizons, that are the most equipped to navigate the frontier.
Navigating these lands calls for those who can grapple ambiguity, identify and take aim at worthy goals, take cognitive ownership, and who, despite as more and more thinking gets outsourced, do not capitulate to outsourcing their understanding.
AI can be queried, but querying is not the same as answerability. The product engineer is needed because cheaper outputs don’t translate to better outcomes.
Engineers who can go beyond, and think and act like product managers, and push for user-centric designs when creating specs and constraints have always been highly sought after.
These abilities can be trained and developed. But there is a price to pay for everything you gain, and technology compounds both good and bad outcomes with compound interest.
Every time you accept a new technology, you don’t just get something for free, it’s a deal. You get something, but you must give something away in exchange. Where in the excitement and hype you will rarely be told explicitly what you are giving away.
To help us navigate this new deal, in this post we’ll explore a set of principles for effective product engineers.
It’s your principles that structure your contact with the unknown, how you learn, adapt, and orient yourself in an ocean of uncertainty.
These are load-bearing principles that are key to managing both yourself and intelligent machines. And at the same time, are also at the most risk of being undermined in the coming world.
These are pillars for a new wave of applied AI product engineers, the new pioneers and the builders of the new world.
May they help guide you when walking the path along the jagged frontier.
Humility
Honour those who seek the truth. Run from those who have claimed to have found it
It’s a hard lesson to learn.
The first move when navigating the unknown is orienting yourself. To do this you need accurate models of what you do and don’t know. What is and isn’t possible. What you can, and can’t control.
This requires humility. The first pillar, and the foundation upon which learning takes place.
Hubris - the opposite of humility, is the most expensive trait one can hold.
It’ll keep you out of touch with reality, and if you’re lucky enough to realise it, you may be already out of time.
When you’re offered short term pain to your ego or sense of self in exchange for long term learning and growth - take learning every time.
No matter how painful it may be to face the mistakes you’ve made, the error of your ways, or how far you now have to back-track, only to begin again. You are far better off to face it.
This is the difference between long term chronic pain, versus taking the hard sharp pain needed to be free of it.
Humility is the precursor to resilience and is the practice of losing well.
It’s not to be confused with shrinking away from challenges and thinking that you’re not worthy. It is the courage to continually feel like a beginner once again, over and over.
It’s the foundation for being able to adapt, reassess and iterate, taking what you’ve learnt into the next attempt.
The world is not deterministic, which means you’ll often be without a map.
Aye, you must learn to proceed without certainty.
This feels hard because we seek comfort in being able to predict how things will go, and to be able to commit to a particular way of being.
If we accept the only constant as change, it means that at some point our preferred way of thinking and being may no longer be possible.
Accepting what is, can sometimes be incredibly painful. Change happens fast, and there are times where we may have no choice but to simply move forward without closure.
This ability to adapt, from one way of being to another, is the key to navigating the non-deterministic nature of the unknown.
Anything that adapts and learns needs three things: a desired state, a measure of the gap between where it is, and where it wants to be, and an action that closes it.
You need to know what the goal is, how far away the goal really is, how risky your current plan is, what direction to take the next concrete step, and what it’s going to cost you.
Humility is ground you can stand to accurately assess how far this journey actually is, not how far you wish it were. Or in the age of Dunning Kruger at scale, how far you are being told it is.
It’s the foundation to think on your feet, and to push or pivot as needed. To not be attached to your idea of a solution, and to be able to strategically let go as needed.
In the times you are not sure of what to say, it allows you to ask questions rather than statements.
And so, if humility is about accurately seeing and accepting reality, no matter how overwhelming, painful, or uncertain. Agency then, despite all that, is all about not accepting your fate.
Agency
The training is nothing!! Will is everything! The will to act.
Too much humility can result in an insidious form of procrastination. Collecting more knowledge, preparing just a bit more before feeling ready to take the next step.
This is another form of self-protection. Fear of spending energy on action that might not pan out. It protects our sense of self if in case things don’t go well.
The counter to that is agency - a bias for action and clear thinking.
It comes from understanding that beyond a certain threshold, more knowledge and preparation isn’t needed. And that threshold is surprisingly low.
It’s the wisdom knowing your plan will likely go out the window the second you get punched in the face. Doing the work is part of figuring out what you should be doing, and taking action will produce the information you need to move forward.
There’s an important difference to highlight between simply creating motion and effective action that moves you forward.
Motion is when you’re planning, strategising, researching, spinning off busywork.
Research and planning is required, but we’re attracted to excess motion because it gives us the illusion of progress without the risk of failure.
Excess motion can make it hard to know where to aim at, and can leave you paralyzed.
An inch of movement is better than a mile of intentions. The action that moves you forward is much more difficult than motion, even with military grade intelligent AI systems at your disposal. It requires deep focus and embracing FOMO.
We can understand agency by looking at its opposite - something like the feeling of burnout or depression.
In this state you feel paradoxically more grounded in reality, which often means overwhelmed, because reality is complex and can be harsh.
You see all the problems with unusual clarity and the obvious negative side of things. This can lead to a sense of low self-efficacy, where the delta between the current state, and the end ideal feels insurmountable, leading to the feeling that you have no real recourse to affect anything.
Cultivating agency is the bulwark against this state. It’s the wisdom to start moving, stop overthinking, and the belief that you’ll figure it out as you go because you’re willing to learn and adapt.
Low agency is “I can’t because X”.
High agency is “I could if…”
Both perspectives lead to very different experiences and outcomes.
Autonomy
Knowing is not enough, we must apply. Willing is not enough, we must do.
Autonomy is the cultivated capacity of deliberating and deciding well.
It’s the ability to act on our own independent judgment. It comes from lived experience and pursuing mastery.
Pure agency without autonomy is hollow, empty execution without heart or aim. Both are necessary, autonomy without agency is impotence: a clear vision of what matters, with no will, or power to pursue it.
Engineers, designers and product managers are all landing on the same conclusion, that their value in this new world is their lived experience and judgment.
The key question then is what does it mean to be able to deliberate well, and to identify which problems are worth solving, and what solutions are actually good?
Effective product engineers must be masters of discernment and long-term vision, who can simplify and get to the core of things.
It requires deep understanding, attention, imagination, balancing urgency, patience and inhibition, and the ability to accurately assess and accept trade-offs. All while dealing with large numbers of conflicting pressures that the real world presents us.
The difference comes down to sensing and detecting trajectories. Trajectories are hard to see from inside the structures where they have been built. Especially when at each step the help offered by AI seems reasonable, each individual decision makes sense in isolation, and where each snapshot along the way looks okay.
Added to this, each small delegation of judgment seems harmless, but accumulated, it compounds, habituates you away from mastery, and toward dependence.
If we confuse the ability to easily produce code with building products, the accumulation of reasonable features no one said no to, make the art of simplification much harder.
The tacit knowledge needed to navigate this new environment only comes through the pursuit of mastery, which is what affords this capacity for self direction to the right places.
Without conscious cultivation of our human autonomy, our imagination of what might be good is just the mean of the training data.
The products we build rapidly incur comprehension debt, building products you can’t really explain, codebases you can’t navigate, decisions you can’t explain, systems you can’t evolve, resulting in cognitive surrender and AI brain fry.
Where one knows where things are going, and who’s in charge.
Get the fundamentals down and the level of everything you do will rise.
In “Thinking Fast and Slow” Daniel Kahneman describes two modes of thought:
System 1 is fast automatic pattern recognition. It associates and perceives new information in line with existing internalised patterns.
System 2 is the slow effortful, conscious thinking. You labour through something difficult, consciously, painfully, and over time it becomes instinctual - tacit knowledge.
Mastery is the process of moving hard-won system 2 understanding up into system 1 muscle memory and instinct.
As each new group of skills mature as we turn ideas into experiments and knowledge into action, transmuting it into skill, we build more layers on top of them.
As time goes on, the layers below become increasingly distant, when we try to articulate them we find it difficult, so we end up using terms like “smell”, and “it’s just a feeling”.
This is the hard won lived experience that is hard to articulate, and forms the basis of taste and good judgement.
This lived experience is not in the training data, and is valuable because it affords you the ability to know what’s possible, what to safely ignore, how to set proper constraints, and how to best draw on the latent potential of the foundations beneath you.
This is important because agents move the work but don’t make it disappear. It allows you to navigate from first principles - to be able to break down problems and systems simply. This requires being intimately familiar with the fundamentals, up and down the stack.
For understanding any large and complex thing, we start by knowing how each separate part works. We then must know how each part interacts with those to which it is connected.
Reality has layers, where the top layers cannot violate the foundations underneath. At the base layer are the known laws of physics, which provide the constraints to chemistry, which provide the constraints of biology, to psychology, then sociology, and culture and so on.
These higher level layers produce unique emergent phenomena that aren’t well explained by the layers below.
For example chemistry alone cannot be used as a tool to predict what will happen in human culture. It’s also not very useful to use physics as a framework or language to explain the layers above.
LLMs in their current form don’t seem to be a new layer of abstraction.
The key to harnessing their power is through the ability to articulate deep technical knowledge simply. The same language you would use to describe a problem space within teams pre-AI.
The frontier also looks jagged, so without deep mastery you may not realise when you have crossed into territory where the AI is unreliable.
The only way the killer app of AI, coding agents, have succeeded so far, is by drawing upon the tremendous latent potential of the foundation that it runs on top of. Simple, well-constructed, primitives like file-systems, unix commands, and pre-built components and frameworks.
AI in its current form has incredible potential. It has studied all available human knowledge on the internet, but does not understand it, it needs a master to guide it towards real world outcomes.
Understand this - the fundamentals are not a one-time stage you pass through at the beginning of your journey.
Masters never move past the fundamentals, they continually refine them, and as a result, see and create new possibilities by taking them to the limits of their vast potential.
Alignment
The Sailor cannot see the North, but knows the Needle can.
Life is a navigation problem. Too much information. Too many options and paths to take.
In the age of cheap execution, the question is - out of all the things we could do, all the places we could go, how do we choose?
Alignment is about working to establish clear, worthy goals.
When grappling with ambiguity, the first position a product engineer needs to fight for is clarity.
Without clarity, you cannot have simplicity, and without simplicity it’s very hard to effectively achieve your aims.
Agency is important here, because clarity usually doesn’t come before taking action but comes as a result of taking action.
Speed is afforded by clarity. Once you’ve secured a position of clarity you can now move fast effectively.
A significant chunk of work at large organisations is working to gain this clarity, and aligning everyone to be aiming at the same goals and ensuring the hierarchy of sub-goals that form underneath are being completed.
Alignment then is really about deciding what not to do. Start by saying no to 10,000 things and define, for the time being, what narrow path you will walk that will achieve the goal.