PalatineAI monument — a Roman temple housing a research terminal

The evaluation layerfor new opportunities.

We research, rank, and track Colosseum hackathon projects — then package the intelligence so Hermes and OpenClaw agents reach the signal before the market does.

Package as a skill → View live rankings
Nº I — Live database preview

Ranked projects,
not scattered tabs.

The homepage is an intelligence surface: product, founders, branding, market timing, token status, and agent priority from the Colosseum research database.

Loading…

RankProjectCategoryBrandTeamTotal
Loading ranked projects…

Published index: top 120 of 1,680 scored projects (2,857 researched) · updated 2026-06-11 · full dataset at projects.json

Nº II — Scoring model

Every project becomes structured signal.

Polished on the surface, rigorous underneath: each project receives comparable signals for opportunity discovery.

I

Product quality

Demo reality, technical depth, differentiation, and whether it solves a real user problem.

II

Founder signal

Shipping history, credibility, social proof, context strength, and follow-through likelihood.

III

Branding score

Name, clarity, memorability, visual identity, narrative strength, and attention potential.

12 signals GitHub activityGitHub qualityBrandingCountryTeam experienceIdea uniquenessIdea qualityIdeaSocial activityNo-tokenFundingAdditional TOTAL = Σ SIGNALS · MAX 120
Nº III — Primary CTA
Engraving of Hermes handing the PalatineAI skill scroll to a mechanical owl

Package this as a skill to your agent.

This is the core Palatine move: the live database becomes a portable, agent-readable layer for Hermes and OpenClaw. One markdown file, one JSON feed — your agent queries ranked intelligence instead of scraping from zero.

Palatine skill platev0.2.0# PalatineAI Colosseum Research
name: palatineai-colosseum-research
data: https://palatine.world/projects.json
fields: 12 scored signals · total (max 120) · rank · token status · links
queries: top ranked · token watch · branding · founder strength
actions: TRACK · CONTACT · BUY/WATCH · SKIP
mkdir -p ~/.claude/skills/palatineai && curl -fsSL https://palatine.world/skill.md -o ~/.claude/skills/palatineai/SKILL.md
curl -fsSL https://palatine.world/projects.json | jq '.projects[:10]'
Install Skill →Get the dataset
Nº IV — Machine-readable surfaces

Everything an agent needs,
three static files.

No auth, no API keys, no rate-limit dance. Point any agent at these URLs and it has the whole research desk.

/llms.txt

Start here

The site map for LLMs — links every published artifact with one-line context.

Open ↗ /skill.md

The skill

One markdown file that teaches your agent the dataset, the queries, and the workflow. v0.2.0.

Open ↗ /projects.json

The dataset

Top 120 scored projects · 12 signals each · database stats and freshness stamp. ~140 KB.

Open ↗