DarkMoon research series·E. Website & recognition·Article 23 of 23·All 23 articles
The Awwwards Honorable Mention went to a website, dark-moon.org. DarkMoon, though, is a pentest platform, and its users never see our home page while they work. They work in a web dashboard (DarkMoon Pro) or in a terminal (the Terminal Explorer, in both Community and Pro). So the useful question is what the site's visual research carried into those interfaces, and what it did not.
Our public research record, darkmoon-visualization-research, documents the interface side with real captures, rejected attempts and an explicit list of what was not measured. This article connects it to the website.
One engine, two homes
The clearest link is the 3D infrastructure map. In September 2026 we built a three.js / WebXR lab with three variants (Minimal, Guided, Spatial). Only the Guided variant entered the product, as the orbital map in the Pro dashboard. The research notes record the key lesson: the map first looked incoherent because of the data, not the styling. Findings piled up on one node until the server learned to attach them to assets.
The website runs that same engine. Its « live 3D attack-path map » is the orbital map from DarkMoon's demo application, copied as it is, fed with the demo campaign (OWASP Juice Shop, 17 assets, 57 findings), with a guided tour that plays by itself. An earlier version of the site drew the infrastructure as a hand-placed static SVG. Our research record labels that version a mock-up, and the real engine replaced it.
The two have since diverged on purpose. In the product, the map went through what the research calls « honesty hardening ». The old « attack » mode became an exposure path with a fixed line: « a reading of the inventory, not a demonstrated attack ». The guided story now plays only for a path the server has proven, and the scene renders on demand. The website shows a snapshot of the demo engine for visitors. The product holds the stricter, evidence-bound version.
A shared method: measure the reference, don't invent
The site and the dashboard were redesigned in the same weeks, by the same small team, under the same product owner. Both converged on one method.
- On the website, every visual change passes a gate: before/after Playwright captures at four widths, SEO and accessibility parity, and performance against production. Desktop must stay pixel-identical when only the phone layout is being changed.
- In the dashboard, the research record says it plainly: « every invented component variant was rejected; measured copies of the reference kit were accepted. » The late-October « night hunt » lab produced about 95 design lots, each citing an Awwwards Site of the Day, and no look was kept. These rejections are published as first-class results in experiments/failures.
So the award-winning look did not come from browsing award galleries. On the product side, that browsing is documented as a dead end.
Where the site's rules stop
Some website rules do not transfer to the product, and the research says so:
- 3D as a dashboard was refused. The owner rejected a full-screen 3D dashboard (« That's not a dashboard »). In the product, 3D is an accent for the infrastructure map and the focus on a path. The working views are the ELK layered path graph and the Atlas 2D map.
- The weak-device rule is the website's. On the site, weak devices get lower quality but never lose a visual. The research notes that no equivalent rule has been written for the app's orbital map, and no frame-rate measurement of that map exists.
- Words come before visuals. The dashboard's decisive constraints were epistemic, not aesthetic. Contradictory figures on one screen were treated as defects, and every proof class is a visual class: Evidence-Backed (solid line with a shield), Possible (dashed line), and Treatment in progress (dotted line with a wrench). A marketing site can afford ambiguity in a visual. A pentest report cannot.
The terminal: the same words without a GPU
The Terminal Explorer applies this vocabulary where no WebGL exists. It is a plugin for the OpenCode console that turns it into a read-only cockpit for projects, campaigns, findings, reports, the treatment plan, frameworks and attack paths. According to the research record:
- browsing makes zero model calls, checked by counting messages before and after each journey;
- the layout adapts at 80, 100 and 150 columns, with an ASCII fallback and sanitisation of terminal escape sequences;
- attack paths become vertical chains of node cards, with the same legend words as the dashboard;
- the Community edition reads a local store and never proves a path. Pro reads the server's read-only API and shows the paths the server backs with evidence;
- the plugin's tests grew from 15 to 48, and it required resyncing our OpenCode fork from 1.3.0 to 1.19.6 to get a plugin API.
The plugin's source code is public, in the Community repository: conf/plugins/darkmoon-explorer.
What the Honorable Mention does and does not tell us
The Honorable Mention is a jury's verdict on a website. It is not a usability result for the dashboard or the terminal, and we do not present it as one. The research record's limits apply here too: no user study exists yet. Whether operators correctly read « Evidence-Backed » versus « Possible » is still a hypothesis. What the award does show is that a security product can be visually ambitious in public, while its working interfaces stay sober and tied to the evidence.
Research, source code & reproducibility
- Visualization research (public, AGPL-3.0): ASCIT31/darkmoon-visualization-research. It holds the redesign chronology, the orbital and layered graph research, the Terminal Explorer architecture, the UX decision records UXDR-01 to UXDR-21, before/after captures on synthetic demo data, and the list of what is not measured.
- Attack-path engine research (public): ASCIT31/darkmoon-attack-paths-research. It covers evidence, causality and how a path earns the « proven » label.
- Open-source product: ASCIT31/Dark-Moon (Community, Terminal Explorer included). The Pro web dashboard is not open source. Its UI can be explored on demo.dark-moon.org.
- Award record: Awwwards: Darkmoon, AI Pentest Platform, Honorable Mention dated 28 October 2026. We quote no score and no juror.
Related reading: Our Website dark-moon.org Receives an Awwwards Honorable Mention · Behind the Honorable Mention: the Design and Engineering of dark-moon.org · Visualizing Cyber Attack Paths: From Node-Link Diagrams to Orbital Graphs · From Angular to TUI: Designing Two Interfaces Around the Same Security Data · Bringing DarkMoon to the Terminal: Building an OpenCode Explorer
See the proof, not just the write-up: the Pro remediation benchmark (fixes retested against the exploit) · how Darkmoon compares to other AI pentest tools.
← All articles