yazin

@yazins • 8,243 subscribers

messing with AI / founder @amalinvest (YC, W22)

Shorts

Introducing: OpenGranola 🔥 I built an open source meeting copilot for macOS. It transcribes both sides of your call on-device, searches your own notes in real time, and hands you talking points right when the conversation needs them. No audio leaves your Mac. Point it at a folder of markdown files, pick any LLM through OpenRouter (Claude, GPT-4o, Gemini, Llama), and it just works. It's invisible to screen share too — nobody knows you have it. The whole thing is open source. Link below

Introducing: OpenGranola 🔥 I built an open source meeting copilot for macOS. It transcribes both sides of your call on-device, searches your own notes in real time, and hands you talking points right when the conversation needs them. No audio leaves your Mac. Point it at a folder of markdown files, pick any LLM through OpenRouter (Claude, GPT-4o, Gemini, Llama), and it just works. It's invisible to screen share too — nobody knows you have it. The whole thing is open source. Link below

293,633 次观看

every mirath calculator I could find was either buried in ads or wanted me to click through 10 pages of forms just to divide shares. so I built the easiest one you'll find. one page. pick who's alive, get the breakdown. 133 test cases, all passing. link:

every mirath calculator I could find was either buried in ads or wanted me to click through 10 pages of forms just to divide shares. so I built the easiest one you'll find. one page. pick who's alive, get the breakdown. 133 test cases, all passing. link:

27,660 次观看

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

FINALLY. solved verse tracking and having a really strong shot at doing verse correction AT THE SAME TIME sll this is made possible because of the amazing work on ZipFormer by the folks at Quran Lab, especially Mustafa Mehdi and Tarek Mansour. i took that same model and fine-tuned it on a larger dataset, and I was able to improve on the results of the vanilla model they'd made available. (i did hear they're in the process of developing an even better model; look forward to seeing what that's about!) so, what does solved verse tracking mean? It means that you can do tracking from voice for any Quranic verse, and it is *extremely* accurate. it worked 100% of the times that I've actually tried it.. there's a public demo on that you can try out yourself what's more, the model is only ~65MB, which is absolutely insane! how do I know it's working well? we have a corpus of information on the open source repo, and the thing KILLED. The only samples it failed on were those that were ambiguous, where you wouldn't even be able to tell which verse it is until you read the entirety of the verse... So that's how good it is! the remaining weaknesses are around correction. It's not always accurate, and sometimes it forgets to correct certain errors. that's something I'm taking a crack at, and will keep the SDK up to date. Check out the demo on and if you're working on a Quran app or web app, we have SDKs that make it a piece of cake to integrate this directly into your apps!

FINALLY. solved verse tracking and having a really strong shot at doing verse correction AT THE SAME TIME sll this is made possible because of the amazing work on ZipFormer by the folks at Quran Lab, especially Mustafa Mehdi and Tarek Mansour. i took that same model and fine-tuned it on a larger dataset, and I was able to improve on the results of the vanilla model they'd made available. (i did hear they're in the process of developing an even better model; look forward to seeing what that's about!) so, what does solved verse tracking mean? It means that you can do tracking from voice for any Quranic verse, and it is extremely accurate. it worked 100% of the times that I've actually tried it.. there's a public demo on that you can try out yourself what's more, the model is only ~65MB, which is absolutely insane! how do I know it's working well? we have a corpus of information on the open source repo, and the thing KILLED. The only samples it failed on were those that were ambiguous, where you wouldn't even be able to tell which verse it is until you read the entirety of the verse... So that's how good it is! the remaining weaknesses are around correction. It's not always accurate, and sometimes it forgets to correct certain errors. that's something I'm taking a crack at, and will keep the SDK up to date. Check out the demo on and if you're working on a Quran app or web app, we have SDKs that make it a piece of cake to integrate this directly into your apps!

27,477 次观看 • 26 天前

BREAKTHROUGH 85% accuracy at just 115MB (!) i almost can't believe it 📹 demo video below for context: i spent days on offline Quran recognition. Whisper, Moonshine, wav2vec2 fine-tunes, layer pruning, CTC rescoring, beam search tricks. nothing cracked 81% accuracy. i then plugged in NVIDIA's pretrained FastConformer and immediately got 90.7%. also, it's 10x faster 🐇 (0.33s vs 3.24s) and 10x smaller 👌 (115MB vs 1.2GB). i tried fine-tuning it on Quranic audio, figuring I could push it even higher. Ran four different configs. every single one came back worse than the stock pretrained model. no idea what's going on there yet. code/repo in next tweet

BREAKTHROUGH 85% accuracy at just 115MB (!) i almost can't believe it 📹 demo video below for context: i spent days on offline Quran recognition. Whisper, Moonshine, wav2vec2 fine-tunes, layer pruning, CTC rescoring, beam search tricks. nothing cracked 81% accuracy. i then plugged in NVIDIA's pretrained FastConformer and immediately got 90.7%. also, it's 10x faster 🐇 (0.33s vs 3.24s) and 10x smaller 👌 (115MB vs 1.2GB). i tried fine-tuning it on Quranic audio, figuring I could push it even higher. Ran four different configs. every single one came back worse than the stock pretrained model. no idea what's going on there yet. code/repo in next tweet

62,404 次观看 • 7 个月前

BREAKTHROUGH on offline-tarteel today Cyberistic hit 100% recall, 100% precision, and 100% sequence accuracy, with 0.84s average CPU latency. I can't believe i'm writing this sentence! His idea was simple: instead of trying to "understand everything" in one shot, it listens to the recitation, finds the Qur’an verses that sound most likely, then checks the small audio clues against those verses until the right one wins. That's what made it fast enough to run offline, and accurate enough to feel real. This is the part that makes me happiest: open source worked exactly the way I hoped! I put my earlier work out there. Someone else picked it up, pushed it way further, and now we have something I genuinely didn’t think we’d reach this soon. link below!

BREAKTHROUGH on offline-tarteel today Cyberistic hit 100% recall, 100% precision, and 100% sequence accuracy, with 0.84s average CPU latency. I can't believe i'm writing this sentence! His idea was simple: instead of trying to "understand everything" in one shot, it listens to the recitation, finds the Qur’an verses that sound most likely, then checks the small audio clues against those verses until the right one wins. That's what made it fast enough to run offline, and accurate enough to feel real. This is the part that makes me happiest: open source worked exactly the way I hoped! I put my earlier work out there. Someone else picked it up, pushed it way further, and now we have something I genuinely didn’t think we’d reach this soon. link below!

31,011 次观看 • 3 个月前

Introducing: auto-maintainer 🔥 Your repo hits the front page. 147 issues, 38 PRs pile up overnight. You triage, review, fix, merge -- and wake up to more. auto-maintainer handles it. Write your rules in a Markdown file, and it triages issues, reviews code, ships fixes, merges PRs, and cuts releases on its own. 95% resolve without you touching anything. The 5% that are actually ambiguous or high-risk? It asks you. ⚡️ npx auto-maintainer init Works with your Claude subscription. No extra API costs. Open source.

Introducing: auto-maintainer 🔥 Your repo hits the front page. 147 issues, 38 PRs pile up overnight. You triage, review, fix, merge -- and wake up to more. auto-maintainer handles it. Write your rules in a Markdown file, and it triages issues, reviews code, ships fixes, merges PRs, and cuts releases on its own. 95% resolve without you touching anything. The 5% that are actually ambiguous or high-risk? It asks you. ⚡️ npx auto-maintainer init Works with your Claude subscription. No extra API costs. Open source.

18,402 次观看 • 7 个月前

没有更多内容可加载