The record
How it was got wrong first
Six massing versions, five drawing stages,
and the corrections that made the difference.
The twenty sheets are the result. This is what they cost. Each correction below is filed with the number that found it, because a method that only reports its conclusions is not a method — and because four of these five were caught by arithmetic rather than by taste.
Five corrections
What was wrong, and what found it.
-
v1 → v2
A roof correct for summer, destroying the winter it existed to serve
Version 1 was a pitched roof — 21° south, 38° north, ridge at 8.05 — with a 1.20 m overhang on the south side. It shades beautifully in June. Set the sun to winter solstice at this latitude, 38.9° at noon, and take the soffit at the outer arris of that overhang, 5.262:
5.262 − 1.20 × tan 38.9° = 4.294
The shadow line lands at 4.29 m up the south face. The ground floor runs 0.45 to 3.15, so all of it is in shadow, and so is the first 1.14 m of the upper floor — 45% of that storey too. In a house at 2,300 m whose only heat is a wood stove, the roof was cancelling the sun on the shortest day of the year.
v2 threw the pitched roof away for a flat, inhabited terrace and split the shading into two brows, one at the floor line and one at the wall head. Everything after this is a flat-roofed house.
-
v2 → v4 · the persistent one
The curved plan kept being extruded straight up
The hardest failure to kill, and the one that came back. All the body-thinking goes into the plan, and then the section stays a straight extrusion — which produces a drum. Horizontal bands on a drum read as 1930s seaside Moderne, which is a look, and a look is what this method is supposed to be incapable of producing.
It is now written into the design language as the first thing to police, with the test that catches it: what does the outline do at 1.5 m that it does not do at 4.5 m?
v4 made the section step twice as it rises — the upper floor pulls in 0.35 over the north side, carried on the floor band, and the parapet pulls in a further 0.55 behind a sloped shoulder. The south face was left untouched, because that is the face doing the solar work.
-
withdrawn
The objection that turned out not to exist
The standing objection to a curved plan is that it costs compactness, and compactness is what keeps a house warm. It was raised, and then it was checked: at equal floor area a compact curved plan has roughly 10% less perimeter than the rectangle, not more. The settled outline encloses 70.62 m² with 30.93 m of wall.
The objection was withdrawn, and the curve stopped being a cost. Less wall is less heat loss and less material, and a curved plan has none of the masonry corner junctions that fail first in an earthquake — which at 2,300 m in Nepal is not a footnote.
-
Stage 2 → Stage 3 · the expensive one
A rule that passed at both solstices and failed in between
Stage 2 drew the upper brow at 450 and defended it in writing: at 28° N this shades the glass completely at summer noon and not at all at winter noon — the geometry does the work, not a shutter someone has to remember. True on both counts. Both counts were the wrong two days to check.
Stage 3 tested four dates instead of two. At the equinox, 20 March, the sun is at 62.3° and the 450 brow puts its shadow line at 5.083 against a head of 5.51 — it cuts 427 mm of the upper glass, 32% of it, in the season when heat is still wanted. At 450 the upper floor was losing its head on 28 January, in the coldest month of the year.
The brow came down to 210, and the fix produced a general rule instead of a number: projection = drop ÷ tan (90° − latitude), which is 0.525 at 27.7° N. Both brows now surrender the head within two days of each other in March and give it back together in September. What 210 cost was rain cover, and that is answered separately by an 80 mm throated band — and it is still the live doubt.
-
Stage 3 · sheet 11
Every view of the house, until Stage 3, was a three-quarter orbit
A view nobody will ever have. Sheet 11 is the first drawing made from a position a body can occupy — eye at 1.60 m, on the walked line, projected from the same outline and levels as the plans rather than drawn.
What it found, in its own words: from 21 m the house reads as two horizontal bands and a stone lump, not as a curved plan. The womb shape is invisible from every position a person will ever stand in.
It changed what needed fixing. The design language now ends its method with it — model last, and view it at eye level from the path, at 1.6 m, where a person will stand.
The correction, drawn
The sheet that was wrong, and the sheet that found it.
Stage 2's section on the left of the year; Stage 3's four dates on the right of it. The second drawing is the reason the first one no longer exists.
Stage 2 · sheet 03. The section as it stood, with the upper brow at 450 and the shading tested at the solstices. Superseded.
Sheet 03
Section A–A · north–south
through the throat, the spine and the lantern — with sun at 28° N
1 : 100 at A3
winter noon 38.6° · summer noon 85.4°
The brows are just the slab edges. 350 at the first floor, 450 at the roof — no separate element to build, flash or hold up. At 28° N that geometry gives full sun on the glass at winter noon and full shade at summer noon, which is the whole job.
Winter sun lands on the floor, not the wall. 2.38 m in on the ground floor, 2.06 m upstairs, both starting about 0.6 m inside the glass — which is where the seat and the landing bench are, and why the floor is the mass that stores it.
The throat is 5.72 m of open air. Ground floor to lantern glass. That is the drawing's biggest claim and its biggest risk — see the note on Sheet 06.
Stage 3 · sheet 06. Four dates, both brows, and the arithmetic in the open. 22 December at 38.9°, 1 February at 44.7°, 20 March at 62.3°, 21 June at 85.7°.
Sheet 06
Sun on the glass, four dates
both brows tested at winter solstice, 1 February, equinox and summer solstice — not only at the solstices
1 : 25 at A3
shadow line = soffit − projection × tan (altitude)
| Date, solar noon | Altitude | Ground brow 350 · head + 2.65 | Upper brow 450 · head + 5.51 | Upper brow 210 · adopted |
|---|---|---|---|---|
| 22 Dec | 38.9° | 3.058 · clears 408 mm | 5.577 · clears 67 mm | 5.771 · clears 261 mm |
| 1 Feb | 44.7° | 2.993 · clears 343 mm | 5.494 · cuts 16 mm | 5.732 · clears 222 mm |
| 20 Mar | 62.3° | 2.673 · clears 23 mm | 5.083 · cuts 427 mm — 32 % of the glass | 5.540 · clears 30 mm |
| 21 Jun | 85.7° | no sun on the glass at all | fully shaded | 3.121 · 1.04 below the sill |
The rain answer — a throated drip band
A separate stone band over the upper head only: 120 high, underside + 5.70, projecting 80, throated. Your + 5.66 puts the equinox line at 5.508 — it cuts the head by 2 mm and makes the band, not the brow, the governing element. At + 5.70 it clears by 38, the brow keeps the equinox, and the two change places about a week before it. The band shades nothing on any of the four dates.
| Date | Alt | Brow 210 | Band 80 at 5.70 | Governing line |
|---|---|---|---|---|
| 22 Dec | 38.9° | 5.771 | 5.636 | 5.636 · clears 126 mm (band) |
| 1 Feb | 44.7° | 5.732 | 5.621 | 5.621 · clears 111 mm (band) |
| 20 Mar | 62.3° | 5.540 | 5.548 | 5.540 · clears 30 mm (brow) |
| 21 Jun | 85.7° | below the sill | 4.626 | fully shaded |
The rule carries its latitude
Written properly: projection = drop ÷ tan (90° − φ), where 90° − φ is the equinox noon altitude. At 27.7° N the coefficient is 0.525. Both brows are cut just under it — 0.507 on the ground, 0.488 above — and that shortfall is exactly the 23 and 30 mm of equinox clearance. The 0.51 I gave in Stage 3 was the built average, not the rule.
| Latitude φ | 90° − φ | Coefficient | On 690 / 430 drop |
|---|---|---|---|
| 15° N | 75° | 0.268 | 0.185 / 0.115 |
| 27.7° N — here | 62.3° | 0.525 | 0.362 / 0.226 |
| 40° N | 50° | 0.839 | 0.579 / 0.361 |
| 52° N | 38° | 1.280 | 0.883 / 0.550 |
At 52° N the same rule wants 1.28 × the drop — 880 of oversail on this head, four times ours. The number is not portable; the rule is.
The rule, not the number. Projection = drop ÷ tan (90° − φ). The ground brow already obeyed it — 350 on a 690 drop. The upper did not: 450 on a 430 drop is a far deeper cut, which is why it failed in the shoulder seasons while the ground brow passed by 23 mm.
Where the year turns. The ground brow first touches its head at 63.1° altitude — 24 March; the upper at 64.0° — 26 March. Two days apart, and both give the glass back on the matching September dates. At 450 the upper was losing its head on 28 January, in the coldest month of the year.
What 210 cost, and what it now costs. 240 mm less eaves over the upper glass — answered by the 80 band above. But the answer is a second element on a wall that had none, 5.7 m up on the only elevation anyone sees. That is the live doubt: Sheet 12.
Stage 1 · six versions
The massing, in the order it happened.
Each of these is a model that was built, looked at and moved on from. The notes are the ones written at the time.
-
Stage 1 · metres · Y up
Massing study
- Footprint 9.30 × 7.50 m
- One compact volume, long axis east–west. ~55.4 m² internal per floor, 111 m² over two.
- Height 8.05 m to ridge
- Plinth 0.45 · eaves 5.70 · asymmetric roof, 21° south / 38° north, ridge 1.40 m north of centre.
- Entry — carved, not added
- A 2.60 × 1.50 m recess bitten out of the east end, sheltered by the floor above. No separate porch roof.
N north is −Z; the long overhang faces south
-
Stage 1b · metres · Y up
Massing v2 — terrace + lantern
- Footprint 9.60 × 7.70 m
- Walls 0.42 m. Internal 110.2 m² — ground 54.9, upper 55.3, after the entry recess (−5.2) and light well (−4.8).
- Flat terrace, parapet top 7.40 m
- Pavers level at 6.30 on pedestals; membrane below falls 2% north to two 110 mm outlets. Guard height 1.10 m.
- Lantern over the well — top 8.05 m
- All glazing vertical: 3.78 m² south, 1.14 m² east. North and west opaque. No horizontal glass anywhere.
- Shading split in two
- Brow at the floor line (proj. 0.35, soffit 3.00) for the ground floor; brow at the wall head (proj. 0.45, soffit 5.70) for the upper.
N north is −Z. Curves only where a body touches; the mass corners are square.
-
Stage 1c · metres · Y up
Massing v3 — curved figure
- Continuous ovoid, 12 vertices
- Radii by function: south face 4.40 · chest 3.85 · womb 2.20–2.50 · waist 1.75 · body-touch 0.35–0.40. Nothing square, nothing below 0.30.
- Symmetric envelope, asymmetric organs
- Plan mirrored about the N–S axis; spine at z −0.85, throat on the axis, gut and stair to either side of it.
- Everything solar kept from v2
- Brows 0.35 @ 3.00 and 0.45 @ 5.70; lantern glazing vertical only; lantern-to-spine run still 2.60 m, so all shadow lines hold.
N north is −Z. Entry limb on the north axis, unheated, thin skin.
-
Stage 1d · metres · Y up
Massing v4 — articulated
- Section steps twice as it rises
- Upper floor pulls in 0.35 over the womb (north), carried on the floor band; the parapet pulls in a further 0.55 behind a sloped shoulder. South face untouched.
- Brows cut to the solar arc
- Kept where the wall normal is within 50° of due south; projection tapers to nothing over the last 1.6 m so each brow dies into the wall.
- Parapet varies by orientation
- South 0.60 solid + 1.10 rail · flanks 1.10 solid · north 1.40 solid, thickened into a 0.48 seat facing the sun.
N north is −Z. Outline, wall, spine, throat and solar geometry unchanged from v3.
-
Stage 1e · metres · Y up · M = 520
Massing v5 — modulus + ground
- One body unit: M = 520 mm
- Plinth 1M · terrace seat 1M · ground sill 2M · upper sill 7.5M · step 3 risers of M/3 on M/2 treads · sill and plinth noses M/8, throated at M/16.
- Thresholds do work, not decoration
- Reveals M/4 deep taken in the insulation zone; sills throw water M/8 clear of the render; the stone family marks every edge a body touches.
- Ground: sun-trap, drip apron, cut bench
- South apron 7M deep on the chest curve · 2M gravel apron all round · bench cut 3M with a retaining edge · wind wall at 8M.
- Water moved down
- 2,500 L at the uphill bench gravity-feeds the ground floor; only a 400 L day tank stays on the terrace.
N north is −Z. Outline, wall, spine, throat and solar geometry unchanged from v3.
-
Stage 1f · metres · Y up · M = 520
Massing v6 — one piece
- Nothing added: lobes and recesses only
- The entry is a 3.5M swelling of the same outline at the east waist, dying back into the wall over 5M each side; the door is a mouth carved 0.75M into it.
- The approach is four stages
- Wide from the SE with no door in sight · a 1.35 m squeeze turning north · a 2.08 m threshold · release under the 7.28 m shaft of the throat.
- Stair rebuilt to suit the new arrival
- Half-turn at the chest's west end, 15 risers of M/3 on M/2 treads; the well keeps its full 3.20 × 2.40.
N north is −Z. Outline, wall, spine, throat and solar geometry unchanged from v3.