This document specifies the Agentic-AI Platform: a first-party probe console that lets a single operator — the Platform Prober — send autonomous capability and limitation probes to the Agentic-AI Platform interface at https://8080.ai/ using the current model selection «Dopamine Ultra 1 ・ 10M context», and read back what the model can and cannot do, together with a truthful verdict on whether a «Add more credits (or subscribe) in order to let the AI continue processing» pay-wall was encountered.
The product intent is derived from the requester's stated goal: find out the capabilities and limitations of the 8080.ai platform under Dopamine Ultra 1, while hoping not to run into a credit or subscription pay-wall, and with the probing decisions delegated to the builder as autonomous decisions. The console therefore does not ask the operator to author individual test cases; it runs a curated, builder-chosen probe suite on the operator's behalf and reports results, limits, and pay-wall evidence.
The audience is the single Platform Prober operating the console. The 8080.ai platform itself is an external, provider-owned system that the console inspects; it is not built, hosted, or modified by this product.
The Agentic-AI Platform is a first-party web console with three destinations: Landing, Probes, and Probe Results. All three are anonymously reachable; no account, sign-in, or identity establishment is required, because the accepted journey is a single operator inspecting an external platform and no durable actor-specific state, commitment, entitlement, or value transfer must be bound to a named participant.
The console owns the probe orchestration and the presentation of results. The actual model work happens on the provider side: the backend sends requests to https://8080.ai/ with the Dopamine Ultra 1 model through the provisioned LiteLLM gateway, and the provider's responses — including any credit or subscription interruption — are recorded and surfaced. The console never purchases credits, never subscribes, and never attempts to bypass a pay-wall; it detects and reports one.
Current scope is the probe lifecycle: start a probe run, watch it execute, and read the results and pay-wall verdict. Future ideas such as scheduled recurring probes or multi-model comparison are explicitly out of current scope and are listed only in the future section.
Removed from scope
The reference directive for https://8080.ai/ declares domain_context and feature_reference uses only; it does not declare content_source, so no source content inventory is included. The verified observations below are preserved as domain and feature context, with unverified items explicitly marked.
Platform context (domain_context)
Feature reference
Unverified aspect
FR-1 — Learn the platform's capabilities and limitations (explicit) As a Platform Prober, I should be able to find out the capabilities and limitations of the Agentic-AI Platform at https://8080.ai/, so that I know what the platform can and cannot do.
FR-2 — Probe the current model selection (explicit) As a Platform Prober, I should be able to probe the current model selection «Dopamine Ultra 1 ・ 10M context», which appears to be recently released, so that my findings reflect the model actually in use.
FR-3 — Detect a credit or subscription pay-wall (explicit) As a Platform Prober, I should be able to see whether a «Add more credits (or subscribe) in order to let the AI continue processing» pay-wall occurs, so that I know whether continued processing is gated.
FR-4 — Autonomous probing decisions (explicit) As a Platform Prober, I should be able to leave the probing decisions to the builder, so that I do not have to author test cases myself.
FR-5 — Send probe requests to the provider endpoint (required_inference) As a Platform Prober, I should have my probe requests sent to https://8080.ai/ with the Dopamine Ultra 1 model, so that the accepted journey is executable.
FR-6 — Start, stop, and re-run a probe run (required_inference) As a Platform Prober, I should be able to start a probe run, stop it while it is running, and re-run a completed or failed run, so that I control the probe lifecycle.
FR-7 — Read per-probe results (required_inference) As a Platform Prober, I should be able to read each probe's outcome, capability finding, and limitation finding, so that the results are usable rather than a single aggregate.
The creative direction is authoritative for this section: Data made physical for an agent probe console, after the muse Refik Anadol — a live instrument that shows data, not a marketing page.
Mode: dark only.
Colour tokens by role
#08090B — black ground holding roughly 80% of the page.#12141A — panels and rows.#F2F4F8 — primary readable text.#39E6C0 — teal, marks live agent data only.#FF6A3D — orange, marks a credit wall or a limit hit only.#7C8494 — secondary labels and metadata.#2A2E38 — 1px panel and row borders.Typography
clamp(44px, 9vw, 128px).Shape language: full-bleed dark canvas with thin luminous strokes and soft particle fields; panels are flat with 1px #2A2E38 borders; no rounded card grid.
Layout: full-bleed generative canvas hero on Landing; below it and on Probes and Probe Results, a single wide column of probe rows with tabular data and hairline rules.
Imagery: generative particle and fluid renders built from the probe data itself; no stock photos and no clip art.
Accessibility: readable text and controls stay whole at 375px, 768px, and 1280px, wrapping or scaling to fit; decoration and motion may bleed and crop but never cover readable text or controls; with prefers-reduced-motion, a usable static arrangement is provided.
The public entry is a full-bleed black canvas of slow teal particle flow. A left-aligned headline in Space Grotesk sits over it at clamp(44px, 9vw, 128px), and the model label «Dopamine Ultra 1 ・ 10M context» shows as a small monospace tag above the headline. The particles are generated from the probe data itself, so the hero reads as a live instrument rather than a marketing banner. The call to action "Start probes" sits in teal beneath the subline. When a run later hits a credit wall, the same visual language flips to flat orange #FF6A3D with a hazard rule, making the limit feel physical rather than decorative.
Interaction Model: Animated Motion Tempo: cinematic Hero Dimensionality: webgl
Landing Hero Motion Brief
#08090B ground.Landing Hero 3D Scene Brief — DIRECTION-DERIVED A single crafted real-time particle system — a compact field of luminous teal points whose flow is driven by the probe data — occupies the full-bleed hero. It shows the product's defining state: a live probe instrument reading the model. The scene is one object, not a scene graph of props; it degrades to a static frame under reduced motion and to a static dark field when WebGL is unavailable.
prefers-reduced-motion, a usable static arrangement is provided.
Dopamine Ultra 1 ・ 10M contextprobe target · 8080.ai
Send agent probes to 8080.ai and see the results and any credit wall.
Every run goes to the provider with the current model selection. The suite is chosen for you: seven categories, 47 probes, each written to surface a capability or a limit rather than a feature tour.
Counts are the probes planned per category; nothing here reports a run. Results and any credit-wall evidence appear on Probe Results once a run has executed.

Dopamine Ultra 1 ・ 10M contextprobe target · 8080.ai
Send agent probes to 8080.ai and see the results and any credit wall.
Every run goes to the provider with the current model selection. The suite is chosen for you: seven categories, 47 probes, each written to surface a capability or a limit rather than a feature tour.
Counts are the probes planned per category; nothing here reports a run. Results and any credit-wall evidence appear on Probe Results once a run has executed.
No comments yet. Be the first!