Haute Lumière · The reading

The foundation is not the easy part, it is the part everything later inherits

Foundations are not the simple version of a subject; they are the decisions that quietly govern every decision made after them.

She has read this section before. The second reading is where the terms stop looking obvious.

THE WORD ITSELF

Foundations is usually read as a synonym for beginner — the shallow end of a subject, the part to be got through on the way to the real material. That reading is a mistake with consequences. A foundation is the set of terms, assumptions and methods a field settles so that it can spend its attention somewhere else. Once settled, it goes quiet, and the quiet is exactly where its power comes from. Everything built afterwards inherits its shape, whether or not anybody still remembers the choosing.

The volume defines the word carefully: the underlying principles, concepts and base elements that serve as the groundwork for a particular field or subject. That reads like a dictionary entry until the emphasis falls on groundwork. What is laid down first does not simply come before the rest; it constrains it. Almost every part of a building can be renovated later, and the single exception is the part it is standing on.

What a field settles early, it stops seeing. What it stops seeing, it cannot correct.

Notice how often an ambitious piece of work fails for an unambitious reason. Two teams using one word for two things. A method adopted because it was the method available, with no record of the reasoning. A definition of success written after the result was already in hand. None of these are sophisticated failures. They are foundational ones, arriving late and dressed as complexity, which is why they are so often treated at the wrong level and so rarely cured.

The same misreading shows up in how foundations get resourced. Work at this layer produces nothing visible on the day it is done — no feature, no report, no result anybody can hold up — so it competes badly against work that does. That is a problem of accounting rather than a problem of value. The return on a foundational decision arrives distributed across everything built afterwards, which makes it easy to underfund and impossible to skip.

This is why the volume moves in the order it does — from definitions to theory, from theory to method, from method to application, and only then to what comes next. The order is not a matter of taste or tidiness. Each layer can only be repaired from the one beneath it, and a problem addressed at the wrong layer returns wearing a different name and a longer schedule.

There is a practical test for whether something is foundational, and it takes about a minute. Ask what would have to be rebuilt if this turned out to be wrong. If the answer is one report, it is a detail and can be handled as one. If the answer is every report, every process and every measure produced since it was adopted, the thing is foundational, and it deserves the kind of attention the word almost never receives.

FIRST PRINCIPLES

The volume's first instruction to a practitioner is to become familiar with the core principles of the framework so that they can be applied effectively. The word carrying the weight is effectively. Anyone can apply a procedure; a procedure is written precisely so that it can be followed without understanding. Applying it effectively means knowing what it is for, which is a different kind of knowledge and is acquired by a different route.

The distinction is between the procedure and the principle. A procedure is a sequence that worked under a set of conditions nobody wrote down. A principle explains why it worked, which means it also says when it will not. A practitioner holding only the procedure performs well until the conditions change, and then has nothing — not even a way of noticing that anything has changed.

This is why theoretical underpinnings get a chapter of their own rather than being folded into method. Theory has a reputation as the part that can be skipped by anyone in a hurry. In practice it is the part that makes improvisation safe. Someone who understands why a step exists can drop it in the one case where it does not apply and keep the result intact. Someone who does not will drop it everywhere or nowhere, and both of those are wrong in a way that is hard to detect from the outside.

A procedure tells you what to do. A principle tells you when to stop doing it.

There is a cost here and the volume does not disguise it: principles are slower to learn than steps. A team can be taught a checklist in an afternoon and will produce something usable by the evening. Teaching the reasoning takes longer and produces nothing at all on the first day. The trade is not between fast and slow. It is between a capability that ends at the edge of the worked examples and one that does not.

It also changes what a beginner is. On this account, the difference between a novice and an expert is not the number of techniques held but the depth at which they are held. Two people can perform identical work for a year and stand in entirely different positions at the end of it, because one was accumulating instances and the other was accumulating reasons. Only the second one can teach it, and only the second one can tell when the method has stopped fitting the situation.

The volume's phrasing puts the responsibility on the practitioner rather than on the document, and that is the right place for it. No framework can be written clearly enough to be used well by someone who does not grasp what it is for. Every attempt to write one produces the same artefact: a document long enough that nobody reads it and precise enough that nobody feels able to act without it.

THE SAME WORDS

The volume asks for consistency in terminology, structure and process across projects. Consistency reads as a low virtue, the sort of thing a standards document requests and nobody actually defends. It is worth defending, because the alternative is not freedom. The alternative is a portfolio of projects whose results cannot be placed beside each other.

Take terminology first. When two groups use one word for two things, nothing announces the problem. Both write coherent documents. Both cite the other. Statements travel between them and change meaning silently in transit, and the fault surfaces months later, at the exact point where a decision turns on the difference. A shared definition costs one conversation. Its absence is paid for repeatedly, in a currency nobody is tracking.

Structure is the same argument at a larger scale. When projects are organised differently, every comparison requires translation, and translation is where evidence quietly degrades. Consistent structure is what allows the second project to be read by someone who read the first without a guide. It is, in the plainest sense, the infrastructure of learning across projects rather than only within them.

A vocabulary is not a formality. It is the cheapest instrument a group of people can share.

Process consistency has a further benefit that is easy to miss: it makes variation visible. If every project runs differently, a difference in outcome tells nobody anything, because there are too many candidate explanations and no way to choose between them. Hold the process steady and a difference in outcome becomes information. This is the reason an experiment holds its variables constant, applied to an organisation rather than to a laboratory bench.

There is an honest cost to consistency and it should be stated. A shared vocabulary is a constraint, and constraints are felt hardest by whoever sits furthest from where the terms were set. The answer is not to abandon the standard but to keep a route by which it can be revised — a definition that cannot be challenged stops being a tool and becomes an inheritance. Held that way, consistency is not a rule imposed once; it is something maintained, which is more work and worth the work.

None of this is an argument for uniformity of thought. The vocabulary is fixed so that the arguments can get sharper, not softer. Two people who disagree in the same terms are having a disagreement that can be settled by evidence. Two people who disagree in different terms are usually not having a disagreement at all, and will discover this only after considerable expense.

Nothing on the page was chosen quickly. The order of the chapters is itself part of the argument.

WHO IS ASKED

The instruction is to involve key stakeholders early in the process, to gather insights, and to ensure alignment with objectives. Read once, it sounds like courtesy, and courtesy is easy to defer. It is not courtesy. Early involvement is the only reliable method for obtaining a category of information that exists nowhere in the documentation and cannot be derived from it.

The people closest to the work hold the exceptions. They know which step is skipped in practice and why the skipping has been harmless for three years. They know which figure is entered by hand because the system produces it a day late, and which client is handled differently for reasons that predate everyone now in the room. None of this is written down anywhere. A plan built without it is a plan built on a description of the work rather than on the work.

Timing is very nearly the whole of it. Consultation after a design is finished is not consultation; it is ratification with extra steps, and everyone in the room recognises the shape of it within a minute. The information that would have improved the design arrives when the design can no longer absorb it, which converts a useful objection into an obstruction. That conversion is entirely an artefact of when the question was asked.

Asked early, an objection is information. Asked late, the same sentence is resistance.

Alignment with objectives is the second half of the instruction and it fails in a particular way. Groups routinely agree on a goal at a level of abstraction where agreement is free — better, faster, more coherent — and discover much later that they had meant incompatible things by it. Early engagement, conducted honestly, forces the objective down to a level of detail at which disagreement can actually appear. A disagreement that surfaces in the first week is a design input. The same disagreement in the ninth month is a rebuild.

One failure mode is worth naming, because good intentions produce it reliably: asking everyone about everything. Wide consultation with no question attached gathers opinion rather than information, spends the attention of exactly the people whose attention is scarcest, and teaches them that being asked means nothing in particular. The useful version is narrow. Name the decision, work out who holds evidence bearing on it, and ask those people that.

The practical form of this is unglamorous. Name who is affected. Ask them before the shape is fixed. Write down what they said in their own words rather than in the plan's words. Then show them what changed as a result. The last step is what makes the second round honest, because people answer carefully once they have seen a previous answer alter something real.

THE SECOND DRAFT

The volume asks for an iterative approach that allows for feedback and adjustment throughout the implementation process, rather than at the end of it. The difference between throughout and at the end is the entire content of the instruction, and it is almost always lost in the summary.

A framework applied once is a hypothesis with a sample of one. Applied repeatedly, with each application observed and the observations fed back, it becomes something much closer to knowledge. The difference is not the number of applications. It is whether a return path exists — whether what is learned during the third application can reach the fourth before the fourth begins.

Iteration has to be designed, and this is where the idea most often evaporates. A team that intends to iterate but names no moment for it does not iterate; it drifts. Drift and iteration look similar from the inside and are opposites. Drift is change with no record of why, which means it cannot be undone, cannot be defended, and cannot be taught to anyone who was not present.

Change with a reason attached is iteration. The same change without one is drift.

There is also a discipline about what iteration is for. It is not a licence to keep adjusting until the result stops attracting comment. Each round should answer a question that was genuinely open, and the question should be written down before the round begins. Without that, the process quietly becomes a search for a version nobody objects to, which is a different objective, usually a worse one, and difficult to abandon once it has started.

It is worth noticing what this asks of a schedule. A plan with no room after the first delivery cannot iterate, whatever anyone intends, because the only available response to what is learned is to note it down for next time. Next time rarely arrives with the same people and never with the same detail. Reserving capacity for the second pass is the scheduling decision that determines whether an approach is iterative in fact or only in description.

The reason the volume attaches iteration to implementation rather than to design is that implementation is where the untested assumptions are kept. A design can be reviewed indefinitely without generating new information, because the review is conducted by people who hold the same assumptions that produced the design. Contact with the actual work is what supplies the surprise, and the surprise is the entire point of doing it early.

WHAT IS WRITTEN

Keep thorough documentation of processes, decisions and changes, the volume says, to support transparency and future reference. The word deserving the emphasis is decisions. Most organisations document their processes tolerably well and their decisions hardly at all, and the second omission is the expensive one by a wide margin.

A process can be re-derived by watching the work. A decision cannot. The reasoning that produced it lived in a conversation, under conditions that no longer hold, weighing alternatives nobody recorded. Six months later the decision is still in force and the reasoning is gone, which leaves everybody holding a rule they have no way to evaluate. They can obey it or break it. They cannot assess it.

This is how sensible choices turn into superstitions. The condition that justified a rule disappears; the rule remains, because nobody can demonstrate that it is no longer needed and nobody wants to be the person who removed it. Over time an organisation accumulates a layer of practices followed for reasons no one can state, and from the inside that accumulation is indistinguishable from having standards.

Record the alternative you rejected. It is the only part that cannot be reconstructed later.

The remedy is narrow and cheap. For any decision with weight, write four things: what was decided, what else was considered, what made the difference, and what would have to change for the decision to be revisited. The fourth is the one that gets skipped and the one that pays. It turns a rule into something with a stated expiry condition, which is what allows a later team to retire it without a fight or a ceremony.

A second effect is that a documented decision can be argued with. An undocumented one can only be complied with or resented, and resentment is a poor instrument for locating errors. When the reasoning is visible, a newcomer who thinks it is wrong has something specific to attack, and occasionally the newcomer is right. That is an inexpensive source of correction, and most organisations have it switched off without ever deciding to.

Transparency is the last effect and it is not only for auditors. A person who can see why a thing is done can do it well, adapt it under pressure, and explain it to the next person. A person who cannot is performing a ritual, and will perform it exactly as instructed in the situations where it helps and in the situations where it does harm.

A warm afternoon, one plant, one page open. Understanding tends to arrive at about this pace.

TEACHING IT

Provide adequate training and resources, the volume says, so that users can understand and use the framework effectively. Behind that plain sentence sits a claim worth drawing out: a framework does not exist in a document. It exists in the number of people who can use it well when nobody is watching and nothing is open in front of them.

That changes what adoption means. Adoption is not a decision taken once by whoever is entitled to take such decisions; it is a teaching problem, and it is finished only when the framework can be used unsupervised. By that measure, a framework that has been announced, circulated and referenced in three meetings has not been adopted at all. It has been mentioned.

Training fails in a characteristic way, too. The session runs, the materials are good, the room follows along — and understanding actually forms later, at the first real encounter, which is precisely the moment when nobody is available to ask. Support after training is not a courtesy line in a budget. It is the point in the process where the learning happens. Cutting it does not save the cost of the training; it wastes it.

A framework lives in the people who can use it with the document closed, and nowhere else.

The volume's own list of what could grow around the material follows from this, and it reads better as an argument than as a wish list: workbooks with exercises and reflections, an application for learning away from a desk, integrations with the systems people already work inside, an online community where questions are asked in public, and specialised courses for particular depths and interests. Each is a different answer to one question — where does understanding actually form, and what would it take to be there when it does.

The community item deserves its own note. A question asked in public is answered once and read many times, and the answers accumulate into a record of where the material is genuinely hard rather than where its authors expected it to be hard. That record is not obtainable any other way. It is the most honest feedback a framework ever receives, because nobody has ever asked a question in order to be polite.

The progress-tracking idea has the same character. Its value is not motivational decoration; it is that a learner who can see where their understanding is thin can direct their own attention, which is the capability the whole exercise is trying to produce. A framework whose users can diagnose themselves needs far less support than one whose users can only report that something is not working.

THE MEASURE

Establish clear metrics to evaluate the effectiveness of the framework, and adjust as needed based on performance data. This is the instruction most likely to be nodded at and least likely to be carried out, because it is the one that makes failure visible to people who did not previously have to see it.

A framework with no measure occupies a comfortable position. It cannot be shown to be failing, so it can always be defended, and because it can always be defended it never improves. The measure is what converts an opinion about the work into a claim that could turn out to be wrong. That is uncomfortable by design, and the discomfort is the mechanism rather than a side effect of it.

Choosing the measure is the hard part, and the volume is right to ask for it explicitly and in advance. A metric selected after the results are in hand is not a metric; it is a summary chosen for its flattering shape. Stated beforehand, a measure commits the work to a definition of success it will then have to live with, and that commitment is the only condition under which the resulting evidence means anything at all.

A measure chosen after the result is not a measure. It is a description with an opinion.

The known hazard is that the measure becomes the goal and the work bends towards the number. It is real, and it is not an argument against measuring. It is an argument for choosing measures that sit close to the thing actually valued, for holding more than one at a time, and for treating every measure as provisional. The instruction to adjust applies to the metrics as fully as it applies to the framework; a measure that has stopped corresponding to what matters should be retired with the same seriousness with which it was adopted.

There is also the question of what is not being measured, which is where confident numbers do their damage. Every measure has a denominator, and the part of the work that falls outside it does not stop existing — it stops being visible, which is worse, because absence of evidence now arrives looking like evidence. Stating plainly what a metric does not cover is a small habit with a large effect, since it keeps the number in proportion to the claim it can support.

Then there is the part that rarely happens: acting on what comes back. Performance data that changes nothing is a ritual with a spreadsheet attached. The decision about what will be done at each possible result is best taken before the data arrives, while nobody yet has a stake in which result it turns out to be.

BENDING NOT BREAKING

Be open to adapting the framework to suit specific project needs while maintaining its core principles. Two instructions inside one sentence, pulling against each other, and the tension between them is the useful part rather than a flaw in the drafting.

A framework that cannot adapt gets abandoned. Not argued with — abandoned, quietly, by people who have work to finish and a method that does not fit the work in front of them. The formal version stays on the intranet while the actual version lives in habits nobody has written down. That is the worst of both outcomes: the full cost of having a standard, and none of the benefit.

A framework that adapts without a core has a different and slower fate. Each adaptation is locally reasonable; the sequence of them is not. After enough rounds the name is attached to whatever people were already doing, and the framework has become a label rather than a method. Nothing dramatic occurs at any single step, which is exactly why the drift is so rarely caught in time.

Name what cannot change, and everything else becomes safe to change.

It helps to be specific about what core means here. Core is not the part hardest to change, nor the part written first, nor the part with the most senior author. It is the part whose removal would make the rest incoherent — the assumptions that the other elements exist as answers to. Tested that way, most frameworks have a smaller core than their documentation implies, and discovering this is liberating rather than deflating: most of the surface turns out to be adaptable in good conscience.

Which returns to the first two movements and shows why they had to come first. The line between core and local can only be drawn by someone who knows what each element is for. A practitioner holding principles can say: this step exists to catch a particular failure, that failure cannot occur in this setting, the step goes. A practitioner holding only the procedure has no basis for that judgement and must either preserve everything or negotiate each item on grounds of convenience, which is how a method becomes a preference.

A workable form of the rule is to write the core down as a short list and require every adaptation to state which item it is preserving. That is a light constraint and it does two things at once. It obliges the adapter to have a reason, and it produces a running record of how the framework is actually being used in the field — which is, in the end, the raw material for the next revision of it.

WHAT PROSPERITY MEANS

The volume states its position on prosperity without hedging: prosperity is not merely material wealth but also personal growth, community well-being and sustainable practice, and the extent to which it is realised depends on commitment to those foundational values. That is easy to agree with and worth slowing down over, because as written it is a definition rather than a sentiment, and definitions in this book have consequences.

A definition of this kind has a job to do. It decides what counts as success. What counts as success decides what gets measured, and by the argument of the previous movement, what gets measured decides what gets done. Defining prosperity as material wealth alone is not a neutral simplification made for convenience; it is a choice that sets the measures, and the measures then run the work for years. Widening the definition widens what the framework is permitted to notice.

Ask what a group counts as prosperity and you have already learned what it will build.

This is also where the volume places itself inside something larger. Its stated role is as a foundational element supporting the growth of interconnected components — sustainability, innovation, collaborative ventures — across sectors usually treated as separate concerns. Read against the first movement, that is a strong claim rather than a modest one. It says that the shared layer beneath several fields is where a correction travels furthest, because a change made in the foundation reaches every structure standing on it.

The volume is equally direct about where its account diverges from the ordinary one. The divergence runs between how resources are applied and how environments are augmented — between the how of a practice and the what of its outcomes, between choices of place and the enhancements built into that place. Those are separable questions that habitually get asked as one, and separating them is what allows a practice to be examined on its own terms without its results being quietly assumed in advance.

Cross-pollination, the volume's own word for it, follows from the same position. Integrating different approaches produces insight that neither approach produces alone, and it only works where there is a layer at which the two can meet. That layer is the foundational one: shared terms, shared structure, shared standards of evidence. Without it, integration is a conversation between two vocabularies with no exchange rate, conducted with enthusiasm and settling nothing.

It is fair to ask what commitment to those values amounts to in practice, since commitment is a word that survives almost any quantity of inaction. The answer implied by the volume's own argument is unsentimental. It means the definition is written into the measures, the measures are stated before the results arrive, and what comes back is acted on even when it is inconvenient. Values that never reach a metric stay decorative. Values that do will occasionally cost something, and that cost is the only available evidence that they were real.

So the practical answer to why any of this matters. Foundations are not the preliminary stretch of a subject, to be cleared before the interesting work starts. They are the part that decides what the interesting work will be able to do, what will count as its success, and whether anything learned in one place can be carried to another. The reading is free and nothing is held back. The argument it makes is a single one: the layer everybody stops looking at is the layer worth looking at first.

Free to read

Free to read, and free to hear. Every chapter of every book in this house, and every narration of it, is open to anybody. No account, no card, nothing to cancel.

What is in it

• Volume I: Foundationsopen this in the house search

Advanced problems are rare. Most late failures are early decisions arriving exactly on time.

What is never written down does not vanish. It hardens into custom and stops being questioned.

A method that cannot be measured cannot be improved, only defended.

Consistency is not bureaucracy. It is the condition under which two results can be compared.

Keep looking

Every phrase on this page opens into the house search. The shelf holds The Developmental Canon and six other shelves, and the reading is free.