The answer is almost always yes
Producing software is no longer the hard part. Almost anything a client asks for can now be built, and quickly, which means effort has stopped being evidence that anyone thought carefully about it. This essay is about what replaces that evidence: the judgement to know which answer is actually right, the discipline to go and find the edge case that speed glosses over, and the honesty to say plainly what has not been checked.
Read it-
Study
Study 05 / Two worlds, one camera
This study rebuilds the crossfade as two fully procedural Canvas 2D worlds driven by one virtual-scroll scalar. Orbit owns 0.00–0.55, both renderers run only through the 0.55–0.70 handoff, and arrival owns 0.70–1.00. The visual contrast is deliberate: a point-grid cartographic night gives way to a flat, warm editorial courtyard without resetting the camera or altitude readout.
-
Study
Study 07 / The autopilot
This study treats pacing as the authored material. Five editorial segments share one virtual-scroll scalar. Three statements are held longer than the two canvas interludes, so the automatic pass has deliberate pauses and quicker transits instead of constant speed.
-
Study
Study 08 / The seam
This study makes the WebGL scene and semantic DOM typography read as one spatial rig. Six procedural glass slabs and five real text statements occupy fixed world coordinates. A single virtual-scroll scalar moves one camera through the set.
-
Study
Study 09 / A million, exactly
This study makes the particle count auditable. The live path allocates one storage buffer containing fixed 32-byte structs. The HUD derives the displayed population from GPUBuffer.size / 32; it does not reuse the requested count as display copy.
-
Study
Study 11 / The silent pass
This study is about the failure mode where a check does not fail. It simply never runs, prints nothing, and is indistinguishable from a clean result. The reader is shown three real commands, all of which produced no output, and asked to judge them on that output alone. Only then does the instrument itself become visible, and two of the three turn out never to have executed.
-
Artefact
The Burn Ledger
This graphic encodes the supplied run-ledger notes only. The horizontal clock is deliberately labelled approximate because the source notes do not provide precise timestamps.
-
Artefact
Mander Editions
EDITIONS 001–012 uses one layered flow-field system throughout the series. Seeded multi-octave value noise steers line-integral paths through a fixed print field. Edition parameters alter density, scale, curl, directional drift, and a restrained ink set while preserving one visual grammar.
-
Artefact
Foundation Sheet 01
The sheet uses a two-layer token model inside the single index.html file.
-
Artefact
Loom
A working eight shaft loom in a single page. Three small tables describe how the warp is threaded, what each treadle lifts, and which treadle is pressed for each pass of the shuttle. The cloth in the middle is not drawn, it is derived from those tables on every edit. The point of building it is the property that falls out for free: every crossing in the fabric is the result of exactly one comparison, so you can point at any single thread and recover the three decisions that produced it. Hovering the cloth names them.
-
Artefact
Quote Assurance
A pre-send checker for a commercial electrical quote, built as a working concept proof rather than a mockup. It compares a draft against the job schedule, the contractor's own price book, a standard-items policy and comparable completed jobs, then returns one of three answers per check and a verdict computed by fixed rules. The interesting answer is the third one: not pass, not fail, but the evidence needed to decide was missing, so no answer is offered at all.
-
Artefact
Rebuilt 01: Tilt
Measured from index.html after the final source edit on 22 July 2026:
-
Artefact
Rebuilt 02: Beams and spotlight
Measured from index.html after the final source edit on 22 July 2026: