Product Manager and UX β’ building @OnyonSystems
Earth
Joined September 2009
- Tweets1.7K
- Following650
- Followers664
- Likes16K
ANDR retweeted
Introducing π»π΅πΌπΉ: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> π»π΅πΌπΉ-π·.π½π±: text, 256k context window.
> π»π΅πΌπΉ-π
π»-πΈπ±: vision-language model with video-input support for spatial understanding
> π»π΅πΌπΉ-π°ππππ-πΈπ±: tool calls, voice chat, reasoning in 18 languages.
> π»π΅πΌπΉ-πΎπππ-πΈπ±: an omni model combines vision, audio, and language.
> ππ·-πΈπ± and ππ·-πΎπππ-πΈπ±: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> π»π΅πΌπΉ-π
π»-πΈπ± is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> π»π΅πΌπΉ-π·.π½π± is #1 on the Pareto frontier of device intelligence πππππππ.ππππππ.ππ
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ π§΅
lately building Perla is export, recovery, and what happens when you lock your phone halfway through a recording. no demo video for any of it
I hate apps that throw confetti at you for keeping a streak. Perla has no streaks: skip a week, and it just waits for you.
Day 5 of building in public, and its been a blast. Thanks to everyone whoβs followed me
My private local ai journal app Perla, ended up needing 3 models on the phone. @OpenAI whisper for transcription, @Alibaba_Qwen Qwen for tagging, @Microsoft's miniLM-L12 to find related entries.
all that so i can talk into my phone privately and find the thought again later
i've been working on Perla for 3 months and barely sharing any of it. posting feels weird when you're used to just building.
the discussion "is it art or design" has been around for centuries. art has soul, design is made for an audience. "is ai alive or just imitating us" is the same split
In this era of artificial intelligence, it is becoming urgent to distinguish human art from what machines produce. There is an ontological difference, even before an aesthetic one, between art and what a machine can generate through statistical calculation based on millions of images created by others. Algorithms lack the spark of humanity. For this reason, the Church wishes to renew an alliance with artists and cultural institutions to safeguard our humanity.
Why brit though?
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
Itβs the first Human Interaction Model (HIM).
Readers added context they thought people might want to know
The 48% figure and "video Turing test" claim are from Tavus's own study of 54 one-minute calls, not independently verified or using a standard protocol. Griffin-Lite leads NVIDIA's VideoFDB benchmark on their public leaderboard.
cellcog.ai/blog/tavus-griβ¦
research.nvidia.com/labs/amri/projβ¦
tech-ish.com/2026/10/02/tavβ¦
thankss as a solo founder running 7 agents on Buzz my company lives in the sidebar π
Buzz Desktop v0.5.26 π
More reliable sidebar sync across devices: sections, sorting, stars, and mutes. Plus clearer report errors, better handling of query timeouts, and a new admin console for relay operators.
github.com/block/buzz/releasβ¦
