Tag
Demonstrates achieving 20GB VRAM and 448GB/s bandwidth for around $100 using two NVIDIA P102-100 cards, running a llama.cpp server with a Qwen model and supporting 3 concurrent users with large context.
Wire, a platform for AI agent context containers, is migrating from Cloudflare Durable Objects to a custom runtime on Fly Machines due to structural limits: vector index separation, compute-data colocation, placement flexibility, and lack of self-hosting. The new architecture improves latency and enables dedicated capacity.
Developers are increasingly moving from GitHub to platforms like Codeberg and self-hosting due to concerns over Microsoft's ownership, downtime, and political direction, with notable projects like Ghostty, Zig, and Tenacity leading the exodus.
Chatto, a group chat application with end-to-end encryption and built-in voice/video calls, is now open source and available for self-hosting. A paid cloud hosting service, Chatto Cloud, is also announced for public beta.
Theo praises GLM 5.2 as an incredible model but criticizes the notion that it is self-hostable and comparable to Fable, sparking debate about model accessibility.
ArkhasFlow, a professional trading platform, announces it is becoming fully open source, releasing all components including backend, frontend, and core infrastructure, allowing self-hosting with no dependency on the company.
A comprehensive beginner's guide to obtaining and configuring your own ASN and IP addresses, covering BGP setup, costs, and limitations.
A developer documents building a self-hosted text-to-speech app on an NVIDIA Jetson Orin Nano using Kokoro-82M and durable streams, enabling reliable local AI inference with shareable audio outputs.
A user describes their fully local AI stack using multiple hardware devices running Chinese models like GLM, Qwen, and Kimi, claiming 87% cost savings compared to frontier models like GPT-5.5 and Opus 4.8, while noting plans to self-host video generation.
Brad Fitzpatrick left Google to develop Perkeep, an open-source personal data storage project that gives users full control over their digital data, freeing them from big tech cloud services. However, it is still in early stages and targets technical users.
Spindle introduces a new microVM engine that runs each workflow in its own lightweight virtual machine, supporting NixOS configuration, Docker builds, and caching for fast subsequent runs.
The Human-Centered Computing Foundation is campaigning to secure a new .self top-level domain dedicated to ethical, human-centered technology, aiming to support self-hosting and reclaim digital autonomy.
A troubleshooting guide showing how changing OpenClaw's contextInjection setting from 'always' to 'continuation-skip' fixes prompt cache reloading on every turn when using llama-server, resulting in a 100x speed improvement for long sessions.
The author describes how they automated their homelab to require minimal manual maintenance using a single server, UniFi networking gear, and Docker containers with scheduled updates and backups.
A guide on setting up your own DNS over HTTPS (DoH) service to enhance privacy and security by encrypting DNS queries.
A comprehensive guide to running LLMs locally across various hardware and software setups is now available online for free, covering tools like llama.cpp, vLLM, and more.
A user reports that MTP versions of Qwen 3.6 and Gemma 4 models produce lower quality outputs in code review tasks compared to non-MTP counterparts, with only marginal real-world speed improvements despite higher token generation rates.
CasaOS is a free, open-source system that turns old laptops or Raspberry Pis into a personal cloud server, replacing subscriptions like Netflix, Google Photos, Dropbox, and 1Password with self-hosted alternatives.
A comprehensive 2026 guide to 30 powerful LLMs that are free to use, distinguishing between hosted platforms and self-hostable open-weight models, with detailed hardware requirements and license considerations.
The author calculates the token cost and break-even period of running large models on a Mac Studio, concluding that it is not cost-effective for ordinary users to buy a Mac for personal large model use, and suggests that using APIs or renting GPUs is more economical.