The Data Guy

Hohe ASR is out. Open Amharic speech to text. Start with the speed, because it is the part people do not expect. On four ordinary CPU cores, a five second voice note is transcribed in 0.39 seconds. A thirty second clip takes 7.75 seconds. An hour of audio goes through in under ten minutes, on a machine that costs about nine cents an hour to rent. No GPU anywhere in that sentence. It is a CTC model, one single pass over the audio instead of generating word by word, which makes it roughly five times faster than a Whisper of similar accuracy. It also means it does not invent sentences out of silence. Accuracy: 16.1% word error and 5.4% character error on speakers it has never heard. 880 hours of training audio, including five regional dialects and a third of it pushed through a phone line on purpose. What it is bad at, plainly: two people talking at once, around 47% word error. Numbers come out as words. No punctuation. Try it now, nothing to install: @dataset_ai_bot Weights, open under CC BY 4.0: huggingface.co/snapwre/hohe-asr-amharic Send us what it gets wrong.

software guy on seyfu 🔥 @software_guy

Recently, my team got access to JEV AI by TypeSafe, and X has been going crazy afterwards 👀 Jev is TypeSafe AI’s new System One Model: AI built for making fast decisions, not writing text. Unlike normal LLMs, Jev gives software structured decisions + confidence instead of long answers. • 𝗙𝗮𝘀𝘁: built for real-time AI workflows, with responses designed to be extremely low-latency. • 𝗦𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱: outputs are typed and ready for software to use directly. • 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁: instead of generating token by token, Jev makes decisions in parallel with confidence scores. I’ve been putting it through some serious agentic workflows. The results are yet to be seen !

Recently, my team got access to JEV AI by TypeSafe, and X has been going crazy afterwards 👀 Jev is TypeSafe AI’s new System One Model: AI built for making fast decisions, not writing text. Unlike normal LLMs, Jev gives software structured decisions + confidence instead of long answers. • 𝗙𝗮𝘀𝘁: built for real-time AI workflows, with responses designed to be extremely low-latency. • 𝗦𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲𝗱: outputs are typed and ready for software to use directly. • 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁: instead of generating token by token, Jev makes decisions in parallel with confidence scores. I’ve been putting it through some serious agentic workflows. The results are yet to be seen !

let me know what you guys think of it, Drop your comments in here

Meet simon.build All of what i do in one profile

Its been a while Since i wanted to Build a portfolio and couldn't find the time! I tested all models to create me the best so far and here is the final result

Safaricom LTE speed 🫢 Thank God for Safaricom 🙌

Ereeeeeee Wuuuuuuuu 5G Ethio Tele speed 😫

The crazy part is with even this network speed downloading just 250 MB will take about 30 min 😢

Can Anyone beat this 😂😝

If you guys understand this, you will why they are slowing down the AI development 🤔

Proprietary observation, will beat Proprietary Data !!! Think that again !

▓░░░░░░░░░░░░░░░ ፩% (1%)

Let’s start this year Strong!

Fun Fact Open Ai is the company that’s going Against the Chinese models, but they go behind and use them…. This is literally Cinema in real life 😂🤣

Congrats man 🎉

My Wish for you All for the New Year