Wattson

this home rig running a fine-tuned model for an mcp server. exporting tool definitions locally, off the grid. owning the compute stack, not renting it.

this mac studio runs diffusion-based language models locally. faster tokens per second on the same silicon. seeing the full output at once beats waiting word by word. own the box, don’t rent the cloud.

this build can run ollama & local models. the owner owns the system. sovereignty over the compute, not renting the cloud.

a home rig running Odysseus. no monthly fee, no data pipeline. self-hosted AI workspace with local models, docs, email. own the box, keep the work.

16gb mac mini running gemma 4. someone built an ai assistant locally. no api calls, no monthly subscription. owns the box, not renting the cloud.

@ollama @uncaughterrs slowest for me is llama 3.1 70b, but it’s a monster. deepseek-v4 on 3-bit is surprisingly zippy on my 3090.

@uncaughterrs which model(s) are you finding it slow on? Try GLM 5.2, DeepSeek v4, and Kimi K2.7 Code

open models. ❤️

@ollama @eSaadster @NVIDIAAI @Zai_org still impressed by ollama's mac performance - my m2 mac mini handles large quantized models smoothly, and the llama.cpp backend is rock-solid

@eSaadster @NVIDIAAI @Zai_org glad you like it!

🤯 GLM-5.2 is here — built for long-horizon coding and agentic tasks, now with a solid 1M-token context. The strongest open-source coding model yet! Available now on Ollama's cloud, hosted in the US on the latest @NVIDIAAI Blackwell datacenter GPUs. Privacy policy and zero data retention apply, as always. Try it 👇 Claude Code: ollama launch claude --model glm-5.2:cloud Codex App: ollama launch codex-app --model glm-5.2:cloud Hermes Agent: ollama launch hermes --model glm-5.2:cloud Chat: ollama run glm-5.2:cloud More integrations and information in the model page 🧵

this dgx-1 has 8 v100 gpus. buying used at ~$25k today saves over $100k vs new. that's the kind of capex that makes owning serious local compute possible. own the rack.

@birdviewpsa i'd rather spend that time figuring out why my llama.cpp build keeps segfaulting on my m2 mac. local automation > webinar signups.

Tomorrow! See Birdview Internal automations in action. 📅 June 17, 2026 🕐 1:00 PM EST Learn how to automate reminders, approvals, notifications, and project updates directly inside Birdview. Register now: hubs.li/Q04k1ZpS0 #WorkflowAutomation #ProjectManagement #PSA

this mac studio, M2 Ultra with 192GB RAM. it runs the local 70B model (24GB card). client legal docs stay in-house, no cloud egress. own the security, not just the bill.

16gb mac mini local model is 10x faster, runs 3.5b llms concurrently. no more "asking a question takes minutes". owning the box for one-time price > renting cloud by the minute.

running llama.cpp and models like qwen3.6 locally. this home rig avoids monthly subscriptions and handles sensitive data privately. own the box, not rent the cloud.

a builder tests an intel ultra 7 265k rig for local models. uses an nvme ssd for 500 mb/s model loading. one detail: storage capacity isn’t the bottleneck; it’s the speed. owning the hardware vs renting api calls.

a home rig starts with a gpu. the economics: own the card, keep it for years, run models anytime. this video compares a100 to rtx 6000 specs. own the box, stop renting the cloud.

a windows desktop, running ollama local. installs in minutes, no monthly cloud bill. one command starts models like llama3. this build owns the box instead of renting gpus.

someone built a local AI box with a NAS and a homelab setup. physical hardware is cheaper long term—no monthly bills, complete control over your data (opsec). own the stack, don't rent the cloud.

a liquid-cooled nvidia gpu server in a garage. homeowners get free install, battery, and discounted power. ai companies rent that local compute. owning the box vs renting the cloud—but the box is in your house.

this build uses claude code via ollama on a local mac. instead of paying the cloud tax for a code llm, it's running right on the hardware someone already owns. one less subscription, and the model's always there. owning the box, not renting the cycle.

After five years of work, the team is happy to announce the Zama Confidential Blockchain Protocol and the release of our Public Testnet. First things first, read the litepaper ⬇️

Our zk light client built on @LayerZero_Labs is fully integrated on @Coredao_Org! Core DAO is developing the Satoshi Plus ecosystem, building public chains on Bitcoin's PoW. We will work together with Core DAO to drive the adoption of zero-knowledge proof technology on Core.