
mike
@mjwgoh
Followers
590
Following
4K
Media
47
Statuses
2K
building ai credit servicing agents. learnt to count @uchicago @uniofoxford.
sf
Joined July 2010
🎉 LongCat-Audio-Codec is officially OPEN SOURCED! 🚀 -an audio codec solution optimized specifically for Speech LLMs. Key Breakthroughs: 1. Dual Tokens: Semantic and Acoustic Tokens are extracted in parallel at a low frame rate (16.7Hz / 60ms).This ensures both efficient
2
47
292
the top two trending models on @huggingface are both for OCR! document processing is a hot topic, kids! 😈
21
64
541
Japan’s Big Banks Are Losing Liquidity — Fast The latest Bank of Japan data on Current Account Balances (attached) shows a clear tightening in the system — and it’s most visible where it matters most: City banks. Between August and September 2025, City banks’ reserves at the
31
224
715
chat is this real??? 128K context, int4 quantisation, 1B params, distilled from Llama 4 🔥 https://t.co/29A5KNUxjo
huggingface.co
12
65
465
Moody’s: More than 20 US state economies are now in, or near, a recession
244
875
5K
BREAKING: After earnings, Goldman Sachs, $GS, is reportedly planning a round of layoffs, per WSJ
139
247
3K
Google just did the unthinkable. They built a voice search model that doesn’t understand words it understands intent. It’s called Speech-to-Retrieval (S2R), and it might mark the death of speech-to-text forever. Here’s how it works (and why it matters way more than it sounds)
169
582
5K
It's wild listening to bossa nova me, beethoven me, and bollywood me talk about voice AI latency. Communicating with people across language barriers is now a solved problem (from a low-level technology perspective, at least). Three years ago, if you'd told me we'd be able to
AI IS FOR EVERYONE! We worked with @elevenlabsio Productions to get dubs in German, Hindi, and Portuguese. They do an incredible job of making sure that context is retained and help you scale your content to more people. Here watch @kwindla speak in multiple languages!
1
3
14
We’ve been working with our in-house forensic investigator on First Brands and Tricolor. As we’ve been saying, these are NOT isolated incidents. Watch these two bankruptcies take down series of banks with them.
22
58
282
A 7 million parameter model from Samsung just outperformed DeepSeek-R1, Gemini 2.5 Pro, and o3-mini on reasoning benchmarks like ARC-AGI. Let that sink in. It’s 10,000x smaller yet smarter. The secret is recursion. Instead of brute-forcing answers like giant LLMs, it drafts a
60
276
2K
56k US home purchases were canceled in August, which equates to 15% of all homes under contract. That's the highest cancellation percentage on record for the month of August.
68
303
1K
Distress in the loan markets US auto loan just had their biggest monthly loss since March 2020
31
209
1K
Just dropped on HF: kani-tts-370m A lightweight open-source text-to-speech model that sounds great and runs fast! > 370M parameters — efficient and deployable on consumer GPUs > NanoCodec + LFM2-350M > Natural & expressive voice trained with modern neural TTS techniques > Fast
13
59
440
Just dropped on HF — NeuTTS Air Next-gen on-device TTS that matches cloud-level quality while staying fully open source. > Real-time speech synthesis on CPU/GPU > 3-second voice cloning, no cloud or data upload > Compact: under 200 MB, runs on mobile and edge devices >
24
146
1K
open source text-to-speech model built by inverting Whisper, supports voice cloning and multilingual output
3
65
576
Copilot in Excel is a global financial crisis waiting to happen.
469
4K
62K
"Huge Red Flag" - Chanos Joins Growing Crowd Questioning Subprime Credit In "The Golden Age Of Fraud"
zerohedge.com
“The opaqueness is part of the process,” Chanos said. “That’s a feature not a bug.”
24
51
245