Stars and mentions
Last 30 days, Oct 4 – Nov 2
Package downloads
Top repositories
cloudkj’s repositories ranked by new stars, as on trending.
Contributes to
Repositories outside cloudkj’s account where cloudkj opened or reviewed pull requests in the past 365 days.
Who noticed cloudkj
Everyone who starred or forked cloudkj’s repositories since the archive began in 2011.
Notable accounts
Ordered by the stars their own repositories have received.
Biggest fans
Accounts that starred the most of these repositories.
Breakout weeks
Each repository’s best week of new stars, with the release published around it.
- Gold
- 01st place finishes
- Silver
- 02nd place finishes
- Bronze
- 03rd place finishes
- Awarded repositories
- 0Owned by cloudkj now
No trending awards yet
None of the repositories cloudkj owns has finished in the top three of a completed trending day or week, by stars, forks or PR authors. See who is trending now.
A repository earns an award for finishing first, second or third in a completed trending day or week, ranked by new stars, new forks or pull request authors. Awards follow a repository through renames and transfers and count toward its current owner. Weeks run Monday to Sunday, UTC.
Mentions
Cocktail-party audio as a reverse Turing test: humans beat speech models, but not by enough [P] Hello - I wanted to share a small project I worked on test whether picking one voice out of background chatter could separate humans from machines, in the spirit of an audio CAPTCHA. Setup - 40 synthetic clips: one target sentence read by a synthetic voice, mixed over multiple background voices reading random short fragments (staggered starts, 4 dB attenuation) - 146 human transcriptions from 18 sessions, unlimited replays allowed - 8 model configurations, one pass per clip: Whisper base/small, Vosk, Wav2Vec2, Qwen3-ASR 0.6B, Parakeet CTC 0.6B, and Gemini 3.8 Flash with a naive prompt and with an instruction to focus on the main speaker - Primary metric: word accuracy (1 − WER, floored at 0); default passing threshold = word accuracy ≥ 0.5 TL;DR: humans are better at the task but the gap is too narrow to be a reliable CAPTCHA Results - Humans 0.71 mean word accuracy vs. 0.54 for instructed Gemini (and 0.55 for Qwen3-ASR) - Pass rate: 77% humans vs. 55% for Gemini - Note: no threshold from ≥0.5 to ≥0.9 gets near my initial rough mental target: humans > 80% and machines < 20% - Paired same-clip gap vs. Gemini: +0.13 (95% CI +0.03 to +0.23, session-level bootstrap); AUC 0.62 - The instruction prompt alone lifted Gemini from 0.34 to 0.54 - Replays didn't measurably help humans; people replayed the clips that were hardest for the models too - Misses rise toward the end of the sentence for humans and every model alike Limitations: small sample (18 sessions, mostly native English speakers), a single synthetic setup, no source separation, and no large or commercial ASR beyond Gemini. I dusted off my dormant blog and did a write-up (https://www.kelvinjiang.com/2026/10/can-machines-follow-voice-through-crowded-room.html) and also shared the code, data, and audio corpus: https://github.com/cloudkj/crosstalk Feedback on the method is very welcome, especially ideas for audio that stays easy for people but is consistently hard for models!