BloonsBench – Evaluate agent performance on Bloons Tower Defense 5
LLM agents play Bloons TD5 with vision and OCR for a surprisingly revealing benchmark.

WASM port of OpenXcom with IndexedDB saves — legal if you own the game data.
Retro gamers, X-COM fans, browser gaming enthusiasts
DOSBox · ClassicReload · Internet Archive's software library
LLM agents play Bloons TD5 with vision and OCR for a surprisingly revealing benchmark.
Milkdrop running in the browser via WebGPU is pure nostalgia fuel.
Jupyter for raags—text notation, playback, and visual score in one notebook, no signup.
Using 1980s Rogue as an LLM benchmark is genuinely novel and technically clever.
Defensive publishing for tech companies protecting IP without expensive patent filings.
Two neat ideas here: dual scores (Market vs Value) and side‑by‑side comparisons with filters for compliance and team size — those make vendor shortlists usable instead of noisy. The tech is pragmatic: a Google Sheet served via a PHP CSV proxy and parsed client‑side keeps the CMS trivial, but that approach could become a scaling or trust bottleneck as the dataset and verification needs grow. Visually the site looks sharp, but the long game depends on how rigorously listings are vetted and updated.