kay (@kodeefr) header
kay (@kodeefr) profile picture

kay

@kodeefr

ml researcher and founder @EmperoAI closing the loop one optimizer step at a time

kodee.cc
Joined October 2024
  • Tweets1K
  • Following175
  • Followers197
91Media
Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)Photo by kay (@kodeefr)
Pinned Tweet
kay (@kodeefr) profile picture
kay
@kodeefrMay 7
what if the universe is a markdown file
1
7
1,590
it doesnt refuse PP
Image from kay (@kodeefr)
10
kay retweeted
kay (@kodeefr) profile picture
kay
@kodeefr16h
This is actually so fun to play around with! Im so excited for the next releases wish i could reveal more
Empero (@EmperoAI) profile pictureEmpero
@EmperoAINov 3
Meet Pi and Vo, the two voices of PiVo Studio. Local text-to-speech in a single Rust binary: ~4× faster than real time on one CPU thread, first audio in ~30 ms, OpenAI-compatible API. Runs on desktop, iOS, Android and in the browser. github.com/empero-org/PiVo-s…
1
10
329
DM me 👀
Empero (@EmperoAI) profile pictureEmpero
@EmperoAI3h
We have something very special to present soon. Currently we are looking for testers with credentials/experience. If you are interested send us a DM or an e-mail to [email protected]
3
111
kodee.cc/posts/etok?v1
a tokenizer that works everywhere — Leon Lehmann

a tokenizer that works everywhere — Leon Lehmann

etok: a deterministic universal tokenizer I train (very) small models. That means the tokenizer ships inside the product, on hardware somebody already…

kodee.cc
1
4
142
code: github.com/empero-org/etok
GitHub - empero-org/etok: Our internal tokenizer

GitHub - empero-org/etok: Our internal tokenizer

Our internal tokenizer . Contribute to empero-org/etok development by creating an account on GitHub.

github.com
6
manta cc53ff7f889a8c36e4bc1b7192a402be9db650d8e382f1d030c3ea23d08f02d8
2
76
my agent in christ you are on a 8xGB300 server WHY would you offload to RAM?
2
11
351
We are about to release one of the most interesting demos in a long while
3
70
Image from kay (@kodeefr)
2
7
272
gemini argon CoT is a work of art insanely bullish on google
Image from kay (@kodeefr)
1
5
125
i really wanted to use argon through opencode but i didnt want to use google-sdk so i made proxima. kodee.cc/posts/proxima?v1
proxima - when gemini speaks gpt — Leon Lehmann

proxima - when gemini speaks gpt — Leon Lehmann

I wanted to use gemini-4-argon in my everyday client. The client spoke OpenAI; the endpoint spoke Google's Generative AI format. So I built a small proxy…

kodee.cc
1
6
196
code: github.com/kodee2k/proxima/t…
1
17
This is actually so fun to play around with! Im so excited for the next releases wish i could reveal more
Empero (@EmperoAI) profile pictureEmpero
@EmperoAINov 3
Meet Pi and Vo, the two voices of PiVo Studio. Local text-to-speech in a single Rust binary: ~4× faster than real time on one CPU thread, first audio in ~30 ms, OpenAI-compatible API. Runs on desktop, iOS, Android and in the browser. github.com/empero-org/PiVo-s…
1
10
329
kay retweeted
Empero (@EmperoAI) profile picture
Empero
@EmperoAINov 3
Meet Pi and Vo, the two voices of PiVo Studio. Local text-to-speech in a single Rust binary: ~4× faster than real time on one CPU thread, first audio in ~30 ms, OpenAI-compatible API. Runs on desktop, iOS, Android and in the browser. github.com/empero-org/PiVo-s…
Empero (@EmperoAI) profile pictureEmpero
@EmperoAINov 1
How many parameters does a voice need? Pi(co) Vo(ice): two English voices in 1.4M parameters, vocoder included. The whole model is 5.8 MB and runs on a CPU, no PyTorch or GPU. huggingface.co/empero-ai/PiV…
7
16
2
126
5,259
LLMs are slowly letting us whiff what the smartest heads at three letter organisations have been playing with for 60 years
0x45 (@0x45o) profile picture0x45
@0x45oNov 2
this guy figured out how to leak your exact location from any video with sound using the Doppler curves of nearby airplanes
Image from 0x45 (@0x45o)
55
932
14
21,628
411,191
kay retweeted
Empero (@EmperoAI) profile picture
Empero
@EmperoAINov 1
Its free on our API because why not? api.empero.org/models/empero…
PiVo v1

PiVo v1

PiVo turns text into clear, natural English speech in two voices, Pi and Vo. It speaks OpenAI's audio API and is free with any Empero API key; the text you send is neither logged nor stored.

api.empero.org
Empero (@EmperoAI) profile pictureEmpero
@EmperoAINov 1
How many parameters does a voice need? Pi(co) Vo(ice): two English voices in 1.4M parameters, vocoder included. The whole model is 5.8 MB and runs on a CPU, no PyTorch or GPU. huggingface.co/empero-ai/PiV…
4
3
38
1,898
kay retweeted
Empero (@EmperoAI) profile picture
Empero
@EmperoAINov 1
How many parameters does a voice need? Pi(co) Vo(ice): two English voices in 1.4M parameters, vocoder included. The whole model is 5.8 MB and runs on a CPU, no PyTorch or GPU. huggingface.co/empero-ai/PiV…
30
42
13
471
129,835
kay retweeted
Empero (@EmperoAI) profile picture
Empero
@EmperoAINov 1
Rate limits that reset every 3 hours are a tax on deep work. Kojo just dropped ours. Binge a whole project in one sitting, then take the weekend off. Your plan, your rhythm. Plans start at $1/month kojo.sh
Kojo: free AI chat that searches, sees and builds

Kojo: free AI chat that searches, sees and builds

Free AI chat that searches the web, reads images and builds pages, games and art for you. Made by Empero AI.

kojo.sh
5
4
27
1,992
iceman-contributor-free on @opencode is a MAI-Base-1 based model! (1T-A35B)
Image from kay (@kodeefr)
Image from kay (@kodeefr)
6
392
had V1-alpha rework my personal website! kodee.cc
Leon Lehmann

Leon Lehmann

leon lehmann (kodee) — founder & researcher at empero. writing up experiments in efficient language models.

kodee.cc
1
6
224
Load more