Archived Work
Two of our early technical reports are withdrawn as standalone papers. Both documented mechanisms that are real and implemented; neither had results that survived the measurement discipline the rest of our research is held to. We record here what each claimed and why it was withdrawn, in the same register as the correction in our market paper. Please do not cite either report. Where the underlying work continues, this page points to it.
ICCD — CRDT intent propagation over contact-plan DTN
Withdrawn 2026. Originally published 2025 as “ICCD: CRDT-Based Intent Propagation over Contact-Plan DTN for Million-Agent Swarms at Hour-Scale Light Delay.”
What it claimed. That mission intent replicated as a conflict-free data type over a delay-tolerant contact plan stays fresh, with age of information scaling as hops × per-hop delay.
Why it is withdrawn. Two reasons.
- The headline was true by construction. With a hand-authored contact plan, both the hop count and the per-hop delay were inputs, so “AoI ≈ hops × delay” restated the assumptions rather than measuring anything. A plan derived from real orbital geometry makes both quantities measured — which is the only version of the question worth answering.
- The delay framing was wrong for constellations. Hour-scale light delay is a deep-space regime. Inside a satellite constellation, light crosses any link in milliseconds; what delays a message is waiting for a link to exist at all. We now measure that contact-wait directly, and it — not light travel — accounts for essentially all of a constellation’s delay.
Where the work lives now. The replication layer is part of Gossamer, and the relay question runs against a real, geometry-derived contact plan in the contact-plan-aoi task of the Constellation Benchmark. A measured study of intent routing over inter-satellite links is in preparation.
Density-modulated Boids and stigmergic coverage
Withdrawn 2026. Originally published 2025 as “Density-Modulated Boids and Stigmergic Coverage: Phase Transitions in 500k-Agent Exploration Swarms.”
What it claimed. That a Boids variant with density-dependent weights undergoes an order–disorder transition in the three-dimensional Vicsek universality class, with fitted critical exponents. An update on this site in February 2026 repeated the exponent claim.
Why it is withdrawn. Identifying a universality class from a custom simulator requires a finite-size-scaling study that resolves a genuine critical point — a crossing of the Binder cumulant across system sizes, from runs long enough to have equilibrated. Our own later, more careful measurement does not meet that bar: it shows a transition, but it does not establish a critical point, and until it does we do not claim exponents or a universality class. The earlier claim outran its evidence, and we withdraw it.
Where the work lives now. The criticality instrumentation the report introduced is in Gossamer, and the validation move it made — checking that the substrate reproduces Vicsek ordering at all before any other result is trusted — is now a permanent benchmark row (vicsek_transition) in the Swarm Benchmark. Our standard for any phase-transition claim is set out in Criticality and Phase Transitions.
Why we keep this page
A lab that only publishes what went its way is advertising, not measuring. Both reports were public; links to them exist; anyone who read them deserves to know what changed. See also our principles and reproducibility policy.