Skip to content

Rebuilding access, one block at a time

Roughly 1.1 billion people live in informal settlements where missing roads cut homes off from emergency services, water, and power. reblock screens a whole city for its most access-starved blocks, proposes the least-disruptive new roads to reconnect them, and grades every proposal on the same footing.

Displacing a matched 10.0% of homes, Loop Network (desire-routed) reaches 92.6% permeability. The footpath network residents have already worn into the same settlement reaches 24.1% with all 3,094 m of it, displacing only 1.4% — it never reaches that budget.

Stony Brook University · AI Innovation & Diffusion REU

Looped Tree's proposed roads through the Cape Town screen's top-ranked settlement
The Looped Tree's proposed roads — 7,997 m of them, displacing 10.1% of homes — through the Cape Town screen's top-ranked settlement, one street-bounded block of 6,619 parcels. See the full run on Results.
6,619
parcels in the single-block benchmark region
92.6%
permeability at a matched 10.0% displacement — Loop Network (desire-routed)
0.7%
of homes displaced to reach the shared permeability target — Loop Network (desire-routed)

What this measures

Every proposal is graded on one tradeoff: permeability — how easily every parcel can reach a street — bought against displacement, the homes a road set grazes. Both are computed the same way for every method, including the footpath network residents built themselves, so the comparison is like-for-like rather than a scoreboard of incompatible scores.

Explore it yourself →

Five stages on one Cape Town block, each one live: drag the screening floor across a whole metro, grow a region block by block, draw your own road and get its real permeability — solved in your browser by the same code that produced every number on this site, not a recording of it. Every control writes to the address bar, so a view you find is a link you can send.

Start here

  • Explore

    The pipeline, interactive — screen, grow, solve, and draw a road of your own.

  • Background

    The problem, why roads matter, and the prior work this builds on.

  • Methodology

    The data → screen → reblock pipeline, and what the two metrics actually mean.

  • Methods

    Eleven road-generation methods, each shown on the ground with its own numbers.

  • Results

    The settlement-scale benchmark and the permeability–displacement frontier.

  • Reproduce

    How to set up the project, and the exact command behind every figure on this site.

  • Team & References

    Who built this, and the research it stands on.

Nearly every figure, table, and number on this site is machine-generated from run artifacts committed in the repository, so it cannot drift from the data — the rare exception is a qualitative approximation stated in prose and flagged as such where it appears.