Can a house plan be data, and which plan needs the least framing lumber per square foot?

What it found

A single-storey house plan as a few parameters: rooms, walls and openings, a framing takeoff with members sized by the open NDS tools of house-spans and house-framing, a bill of materials, IRC-style room-size, light and egress checks, and an IFC4 export through IfcOpenShell. A search over 182,520 plans finds the least lumber per square foot: with two or more bedrooms the best plans use 4.8–5.2 board feet per sq ft at 640–1,024 sq ft and 5.9 at 1,216, and no stick-framed plan in the grid reaches 1,400 sq ft because joists and rafters run out of 2x12. Only 1.3% of the grid passes every check at 1,000–1,150 sq ft. At equal budgets a (1+1) local search found the exact optimum in 36 of 40 runs at 2,000 evaluations and random search in none. Stylised: one rectangular layout, stick framing only, board feet rather than prices, room rules paraphrased from memory, and not a design anyone should build from without a licensed review.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results
• Code
• Tests

More in Open-source residential construction

• Where does a new home's price go, and which levers move it most?Where does a new home's price go, and which levers move it most? Share-anchored model (NAHB 2024): size moves the price most (87.8 to 112.0 for 1,600 to 2,400 sq ft), then storeys, finish level and lot; a 14% framing-lumber cut is worth 0.85% of price (2.4% with labour). Index not dollars, assumed driver splits. Milestone 4 of track 12.
• Does the house-cost lever ranking survive its assumed driver splits?Does the house-cost lever ranking survive its assumed driver splits? Monte Carlo over the splits: floor area stays the largest lever in 96–100% of draws; storeys is the soft claim (swing 9.0–19.9 vs 14.0 baseline); the lumber result does not move. Index not dollars.
• Which uncertain input of the house-cost model drives the uncertainty in a house's price?Which uncertain input of the house-cost model drives a house's price uncertainty? Sobol indices over every lever range, the framing share and the driver splits at once: floor area 53% of variance, storeys 19%, all framing inputs together under 1%. Index not dollars.
• Which envelope upgrades pay for themselves over 30 years, by climate, and does advanced framing belong in the answer?Which envelope upgrades pay back over 30 years, by climate? A degree-day net-present-cost search of 640 packages in eight climates: advanced 2x6 framing is in every optimum (13–15% lower cost than a 2x4 baseline); better windows and R-49 attics do not pay at assumed prices. Stylised, assumed prices, not EnergyPlus.
• Does any alternative to a wood-framed wall beat it on cost, carbon or heat loss for a small house?Does any alternative to a wood-framed wall beat it on cost, carbon or heat loss for a small house? In a stylised model wood 2x6 is cheapest in 97% of random draws and lowest-carbon in 100%; SIP, CLT and printed walls win only on labour. Inputs are ranges from memory.
• Does a small wood-frame house have enough shear wall for wind and earthquakes once the windows and doors are cut out?Does a small wood-frame house have enough shear wall for wind and earthquakes once the openings are cut out? At 115 mph and SDS 0.50 the 28 x 40 ft house uses 19% of a gable-wall line's capacity at 6 in nailing; the openings, not the load, are what shrink it. Stylised, capacities from memory.

All 41 Open-source residential construction projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.