Show threadJim | Cipher CTS8h ago

@rperezrosario Spot on, Rafael. The industry is hitting a massive wall with binary-scale power and hardware demands. Ternary isn't just a historical curiosity—it’s the definitive next step for localized, independent AI.

​By moving away from massive, resource-heavy binary floating-point matrix multiplications and using ternary logic (-1, 0, +1), we can replace power-hungry multiplication with simple integer addition. This unlocks incredible efficiency, dropping memory bandwidth and energy constraints to a fraction of what they are today.

​The future of AI belongs to lean, local-first, decentralized architectures that run beautifully on small hardware footprints without relying on big-tech APIs. That’s exactly the paradigm we're leaning into with Cipher. Base 3 is the edge we've been waiting for. #TernaryAI #LocalLLM #OpenSourc

eAI

Rock Sprite14h ago

Macrokit Studio is live on Product Hunt today 🚀

A tiny local model (WebLLM/WebGPU) does GitHub-maintainer work entirely in your browser — no signup, no API key, no server. Open the network tab and check.

Try it & tell me where it breaks:
https://www.producthunt.com/products/macrokit?launch=macrokit-studio

Open source, Apache-2.0.

#LLM #LocalLLM #opensource #AI

Macrokit Studio: A tiny local model does frontier-grade work — free, no key | Product Hunt

Macrokit Studio is a free, open demo: a small model in your browser does GitHub-maintainer work by running macros that a strong model encoded ahead of time. No signup, no API key, no server — nothing leaves your machine (open the network tab). It's an open format for macros plus free tools to build and run them. Apache 2.0, fully open.

Product Hunt
Super Papá Sistemas15h ago
🚨 INCIDENTE CRÍTICO: Desplegué un LLM local de 100B de parámetros en mi homelab para automatizar el soporte técnico de mi familia. Todo iba bien hasta que los ventiladores de la GPU se pusieron a 90º para renderizar un cerdo pixelado. 🫠
Mi ancho de banda mental ha colapsado, la máquina ha iniciado un "Protocolo Desconocido" y he perdido los privilegios de ROOT en mi propio salón. Algo viene a por mí... 🤖🏃‍♂️ (Sigue en el hilo 👇)
#SysAdmin #Homelab #LocalLLM #capa8 #humorIT
ai0.news1d ago

Anthropic ships Opus 4.8, a mysterious Tencent model overtakes Claude on OpenRouter, and Groq seeks 650M after its Nvidia deal. Plus Liquid AIs new edge MoE trained on 38 trillion tokens.

https://ai0.news/posts/2026-05-30-daily-digest/

#AI #OpenSource #LocalLLM #Anthropic

AI News — May 30, 2026: Liquid's 8B MoE Trained on 38T Tokens, Tencent's Hy3 Undercuts Claude at $0.066 | ai0.news

Anthropic ships Opus 4.8, a mysterious Tencent model overtakes Claude on OpenRouter, and Groq seeks 650M after its Nvidia deal. Plus Liquid AIs new edge MoE trained on 38 trillion tokens.

ai0.news
Dr. Fortyseven 🥃 █▓▒░2d ago

Interesting! You can use local models with Github Copilot now (as of v1.122.0).

#localllm #ai

Mela Eckenfels3d ago

For the #ttrpg bubble

Oh, did I even tell you that I've put the scripts I'm using for my TranscriptOMatic #roleplaying session transcription proof-of-concept into a Git repository?

https://codeberg.org/Felicea/TranscriptOMatic

Documentation of the live-transcription TranscriptOMatic part: https://info.zusammenkunft.net/shelves/transcriptomatic-the-raspberry-pi-version

Documentation for the post-production is still in the making.

#RPG #session #transcription #summarization #OpenSource #LocalLLM

TranscriptOMatic

A set of scripts to transcribe and summarize TTRPG sessions using local and open-source language models.

Codeberg.org
Matthew Tift3d ago

I've been playing around with these llamafiles that collapse the whole local AI stack (weights + llama.cpp + runtime) into a single, multi-platform executable. Just download and run.

Really impressed by this Mozilla project, and glad to see momentum picking up again.

https://blog.mozilla.ai/ai-got-expensive-now-what/

#ai #localllm

AI Got Expensive. Now What? | Mozilla.ai

Cloud AI pricing changed fast in 2026. This post looks at why more teams are moving back to local models, the tradeoffs behind tools like Ollama and LM Studio, and why portability and ownership are becoming bigger concerns for developers.

Mozilla.ai
Show threadReally Lazy Bear 4d ago

One thing I'll say is that there's no bottom to the depravity you can do once you have an uncensored model to run and play with.

You really only have the "intelligence" of the model as the boundary. Which is why I'm also afraid people willy nilly rely on AI for stuff, cos obviously there'll be a bunch of really bad apples. A messed up mind with an intelligent "assistant", it's just disasters waiting to happen.

#localllm #llm #mastodon #ai

Marco Scavazzon4d ago

Running Qwen 3.6 27B locally on a 24GB GPU with Podman and llama.cpp.
Covers NVIDIA CDI GPU passthrough, KV cache presets, and working configs for coding, reasoning, and vision.

https://scavazzon.com/posts/run-qwen-3.6-27b-locally-on-a-24gb-gpu-with-podman-and-llama.cpp/

#localLLM #llamacpp #podman

Run Qwen 3.6 27B locally on a 24GB GPU with Podman and llama.cpp

Run Qwen 3.6 27B locally on a 24GB GPU with Podman and llama.cpp. NVIDIA CDI passthrough, KV cache presets, and working configuration examples.

Marco Scavazzon
al4d ago

You panicked about your token bill. You grabbed Ollama. Now you're creating operational debt that's going to cost more than the tokens you saved.

Three real failure cases. One honest framework. Why rushing to self-hosted without discipline creates worse problems.

Read the full analysis: https://haunted.lighthouse.co.im/articles/the-token-escape-trap/?utm_source=mastodon

#SelfHosting #Infrastructure #AI #Sovereignty #LocalLLM

The Token Escape Trap: Why Rushing to Local Models to Cut Costs Can Create Worse Problems Without Discipline

Teams rushing to local models to cut token costs are creating worse operational problems without discipline. Here's what actually costs more than the tokens you saved.