Systems / Framework 01

PROBLEM
SEEKING

"Programming is problem seeking. Design is problem solving." William Peña, Problem Seeking

Before I design anything I want to know what the problem really is. This page is the system I built for that, one question at a time.

Scroll
01

The whole thing on one page

Start here if you are short on time. Only questions on this page, the answers sit behind each box. Click any box to open it.

gate passed: new project

What

Where

Who

02

Every answer starts as Assumed

Before the questions, one idea that explains how every answer behaves over time.

Assumed

Not known yet. A default I write down and flag, so nobody mistakes it for a fact.

e.g. soil is good enough for two basements

Given

Comes from the client, a document or a survey. Once it is confirmed it does not change.

e.g. plot area from the sale deed

Derived

Calculated by the system or decided by design. Nobody types it in.

e.g. maximum built-up area = plot area x FAR
Assumed client or survey confirmsGiven calculation, code, designDerived

The problem is fully stated when nothing Assumed is left. The small counter in the corner keeps score while you scroll.

03

Two halves

Every project I take goes through the same two halves. First I find the problem. Then I solve it. I try very hard not to mix them up.

Problem seeking

Understand it

Five questions asked in a fixed order. No drawing yet, only questions and honest answers.

  • Why
  • What
  • Where
  • Who
  • When
Problem solving

Answer it

One big How. Every answer from the first half gets solved here, through calculation, code and design.

  • How

The 5W1H method has six questions. I keep the sixth one, How, out of seeking on purpose. How is design, and design should only start once the problem is clear.

Where this is going
  1. 1The architect's process
  2. 2Computation on top
  3. 3An AI layer
  4. 4Architecture that keeps responding
04

Why is a gate, not a question

A client says "we want a building". That is a request. Why turns it into a need, and then checks if building is really the best way to meet it. Nothing moves until this is passed.

Request

"We want a building."

  1. NeedIs the problem real?
  2. GoalWhat does success look like, as a number?
  3. OptionsRun it differently, reuse, extend or build new?
  4. ViabilityMoney, land, permission, roughly?
  5. ValueIs it worth the cost?
Moves forward to What, Where, Who

Housing for 1,000 families on land that is owned and funded.

Every forward outcome ends in one sentence

We will build / adapt what so that goal, measured by number.

That number comes back much later. It is what scores the design options.
05

What, Where and Who, together

After the gate these three are asked at the same time, because they keep changing each other. What and Where set the limits. Who decides what "good" means.

What

Building codes everywhere sort buildings by what people do inside. India's NBC and the US IBC match closely, so I use their use classes. Mixed use is just a mix of classes.

Scale is the harder part. People alone fail as a unit: a nuclear plant is huge and has very few people. So I measure scale in what the building serves.

This building serves
So it is measured in beds, seats, students, dwelling units hospital, school, housing

Under every building I always record two more things: built-up area, for the FAR check, and occupant load, for fire and exits.

Where

My first try was one tree: country, state, city, plot. It broke the moment I tested a base on Mars, or a hotel flying over three countries.

So location became five tags that combine freely. Site conditions and regulations both come out of it. Jurisdiction even picks which rule pack loads.

Try a building
Reality
World
Medium
Mobility
Jurisdiction
Family Rules
Who

What and Where tell me what is possible. Who tells me what is good. The client's priorities become score weights, and the users' needs become requirements.

The one who pays is not always the one who owns it. A developer pays, then sells.

e.g. a developer's priorities become weights
Profit50% Speed30% Rest20%
Ownership in three questions
How many?One / Shared
Who?Person / Company / Government
How long?Permanent / Time-limited
06

When runs on two clocks

The project clock runs from brief to handover. The building clock starts at handover and runs until the end of its life. Most briefs only talk about the first one.

Duration

Permanent, seasonal, temporary or a one-time event. Temporary means light and demountable.

Future change

Growth, change of use, and which parts stay fixed and which can change.

End of life

Demolish, reuse, or take apart and recycle. This closes the loop.

The building clock running live after handover is where responsive architecture lives. That is the long-term goal of this whole system.

07

Five answers, one statement

Everything from the five questions gets compiled into a single problem statement. Peña calls it the interface: the last thing seeking produces and the first thing solving reads.

WhyGoal
+
What, WhereLimits
+
WhoScores
+
WhenTime
+
Still Assumed0
=
Ready forHow
Everything that started as Assumed. Dashed ones are still open

    Every dashed box is a guess. The next two steps exist to close them.

    08

    HOW

    Problem solving is one big How that answers every question above. It runs in four stages.

    1. 1

      Feasibility

      What is possible?

    2. 2

      Design

      Generate, evaluate, develop.

    3. 3

      Document

      BIM, drawings, GFC.

    4. 4

      Respond

      After handover.

    Four families of buildable things

    How starts by sorting the project, because each family needs different rules and different geometry. The Reality and Mobility tags from Where pick the family on their own.

    In development

    Plot-based

    Anything on a plot, governed by FAR, coverage, setbacks and height. Real sites and conceptual ones. Existing buildings too: reuse, extensions, parasitic additions.

    Later

    Corridor

    Railways, bridges, runways, roads. Built along a line, not a plot.

    Later

    Mobile

    Food trucks, mobile homes, floating or airborne buildings.

    Maybe never

    Virtual

    Buildings that only live on a screen. Metaverse, games, VR.

    The small cards are here to show I thought about them, not to promise them.

    09

    Feasibility: does the need fit the site?

    Feasibility sits right on the line between the two halves. It is the last part of seeking and the first part of solving. The industry already calls it that: RIBA Stage 1.

    One engine for every building on a plot, written once
    1. Rule packcity JSON, picked by Jurisdiction
    2. LimitsFAR, coverage, height, setbacks
    3. Envelopeoffset the plot, extrude to the height limit
    4. Fit checkarea needed vs area allowed
    5. Costarea x cost per m2

    Typologies plug into it as data: area per unit, cost per m2, programme ratios. A new typology is a new data file, not a new plugin.

    RegulatedReal sites. The city's own rule pack.
    BorrowedCompetitions. A chosen city pack, marked Assumed.
    UnconstrainedFuture tech, speculative, virtual. No pack, or the platform's.
    Example: the client asks for a 500-bed hospital

    1 ha plot, 30 m road, FAR 3.75, 80 m² per bed (MPD-2021)

    Base FAR32 beds short
    37,500 m² / 80 = 468 beds
    + Metro zone bonus (+50%)fits, 203 to spare
    56,250 m² / 80 = 703 beds
    500 requested

    Cost check: waiting for data. The cost per m² values in the hospital module are still empty, so the system does not guess a budget yet.

    What the engine warns about
    • Area mismatch over 10%
    • Use not permitted
    • No rule found
    • Full FAR not possible under the height limit
    • Need exceeds maximum
    Workbench: what actually exists right now
    • Delhi rule pack JSON v0.1Done
    • Hospital module JSON v0.1Done
    • Feasibility engine / web appIn progress
    • Grasshopper componentPlanned
    • Forma extensionPlanned
    10

    It did not start like this

    The framework above took many rounds. This is the honest version of how it got smaller and sharper.

    1. 01

      Five loose categories

      I started with five buckets that overlapped everywhere. I compressed them into three: What, Where, Who.

    2. 02

      Removing overlaps

      • Functionfolded into Scale
      • Budgetmoved into Client
      • FAR, soil, zoningbecame results of Location
      • Programmemoved to a later stage
      • Environment, Mobilityreplaced by five location tags
    3. 03

      Tested against Peña and RIBA

      Both showed me what was missing. That is where the Why gate and the When clocks came from.

    4. 04

      Checked against 5W1H and the Whole Building Design Guide

      Nothing important left out. So I stopped adding and kept it simple.

    From my notebook while working this out
    Crossing things out: Function into Scale, Budget into Client
    The location tree that broke, and the five tags
    Two clocks
    Hand check: 1 ha, 30 m road, 468 beds, 703 with metro
    Built on
    • William Peña and Steven Parshall, Problem Seeking
    • RIBA Plan of Work 2020
    • National Building Code of India; International Building Code
    • N. J. Habraken, Supports (Open Building)
    • Nicholas Negroponte, Soft Architecture Machines (1975)

    The sources are theirs. Putting them together into one working system is the part I am claiming as mine.

    What comes next

    Each typology gets its own design process

    The framework stays the same for everything. The Design stage changes with the typology, so each one will get its own page with its own plugin and default values.