"""The About tab — a compact, self-contained tour of Multi-Agent Land. What it is, how it works (an HTML/CSS recreation of the agent-ensemble architecture so it scales crisply and lives in the CRT theme rather than a flat PNG), and the links a judge or visitor needs: the submission video, the GitHub repo, LinkedIn, and the three Hugging Face field-notes articles. Pure HTML — no state, no callbacks. Everything is scoped under ``.fishbowl`` so the theater stylesheet applies; the About-only rules live in the inline ``") gr.HTML( """

◉ Multi-Agent Land

small minds · one ledger · ≤ 32B

A small-model, multi-agent interactive story engine built on an append-only event ledger. Agents never call each other — they post typed events to a shared log, and every view (stage, memory, UI) is a projection of that log. One tiny event-sourced engine powers many delightful worlds: a whimsical forest theater, a mystery-solving blackboard swarm, a tool-using oracle grove — not three apps, but three YAML configs of the same engine.

Genuinely alive

Watch small specialist agents write, judge, remember, and render strange interactive scenes in real time.

AI is load-bearing

The multi-agent drama is the product — the magic comes from the models, not the scaffolding.

One engine, three costumes

Agents bind to models by declarative YAML profile; each runs on the small model that fits its job.

""" ) gr.HTML( f'
' f'
How it works · the agent ensemble
' f"{_architecture_html()}
" ) gr.HTML(f'
{_links_html()}
') return {}