Join us

ContentUpdates and recent posts about Unsloth..
Link
@varbear shared a link, 1 month ago
FAUN.dev()

Why I'm leaving GitHub for Forgejo

The Dutch Ministry of the Interior launched code.overheid.nl, a self-hosted Forgejo instance for government source code. This move was driven by the need to own and control the platform where code is published. Forgejo was chosen over GitLab for its open-source nature and alignment with the ministry.. read more  

Link
@kaptain shared a link, 1 month ago
FAUN.dev()

v1.36: Deprecation and removal of Service ExternalIPs

Kubernetes v1.36 deprecatesService.spec.externalIPsand starts the removal path, finally closing CVE-2020-8554, the trust-everyone hole the field has carried since the early days. The project has recommended disabling it via theDenyServiceExternalIPsadmission controller since v1.21, but SIG Network h.. read more  

Link
@kaptain shared a link, 1 month ago
FAUN.dev()

What kubectl debug doesn’t tell you: The silent evidence gap

kubectl debugsessions leave almost no forensic trace: by design,EphemeralContainerStatushas nolastStateorrestartCount, so the exit code, session duration, target container, and debugger logs disappear from the Kubernetes API the moment anything else updates the pod. That breaks incident handoffs (th.. read more  

What kubectl debug doesn’t tell you: The silent evidence gap
Link
@kaptain shared a link, 1 month ago
FAUN.dev()

Mirantis has entered into an agreement to be acquired by IREN

Mirantis has agreed to an acquisition by IREN. The companies have announced no customer-facing product changes... read more  

Mirantis has entered into an agreement to be acquired by IREN
Link
@kaptain shared a link, 1 month ago
FAUN.dev()

When AI agents become contributors: How KubeStellar reached 81% PR acceptance

The KubeStellar Console team learned that AI coding agents improve after engineers build deterministic feedback loops into the codebase. Engineers who grant more autonomy give agents more room to guess, with no new correction signal... read more  

When AI agents become contributors: How KubeStellar reached 81% PR acceptance
Link
@kaptain shared a link, 1 month ago
FAUN.dev()

Extending AI gateways with Rust

Every gateway ships with a set of built-in policies. Authentication. Rate limiting. Request routing. Prompt guards. These cover most use cases. But what about the ones they don’t cover? What if you need to add a custom header based on a database lookup? What if you need to transform a request body i.. read more  

Extending AI gateways with Rust
Link
@kala shared a link, 1 month ago
FAUN.dev()

Create Custom MCP Catalogs and Profiles

Docker made Custom Catalogs and Profiles available for MCP servers. Admins can distribute server catalogs they approve, and teams can package per-developer configurations as OCI artifacts... read more  

Create Custom MCP Catalogs and Profiles
Link
@kala shared a link, 1 month ago
FAUN.dev()

How Code works in large codebases: Best practices and where to start

Anthropic breaks down the patterns behind successful Claude Code rollouts in monorepos, legacy systems, and codebases spanning thousands of developers, arguing that Claude Code performs agentic search over a live filesystem instead of relying on a RAG index that drifts out of sync with active engine.. read more  

How Code works in large codebases: Best practices and where to start
Link
@kala shared a link, 1 month ago
FAUN.dev()

Claude’s next enterprise battle is not models: it’s the agent control plane

New data shows Microsoft and OpenAI leading agent orchestration, but Anthropic's rising stake signals a shift in control of AI infrastructure. Anthropic's move from model to orchestration layer hints at a strategic battle over agent runtime platforms where operational AI work happens... read more  

Link
@kala shared a link, 1 month ago
FAUN.dev()

Tokenomics: the 62.5-minute rule for Claude's cache

Ryan Skidmore works out the tokenomics of Anthropic's prompt cache and lands on a single rule: if you expect to need a cached prefix again within 62.5 minutes, keep refreshing it with cheap reads; past that, let it expire and rewrite, because a 5-minute cache write costs 1.25x base input and a read .. read more  

Tokenomics: the 62.5-minute rule for Claude's cache
Unsloth is an open-source toolkit for training and fine-tuning large language models faster and with less memory than a standard Hugging Face stack. Its core library replaces PyTorch's default autograd with custom backpropagation kernels written in OpenAI's Triton language, which is where most of its speed and memory savings come from. It supports LoRA, QLoRA, full fine-tuning, reinforcement learning, pretraining, and 4-bit, 16-bit, and FP8 training, across more than 500 text, vision, audio, and embedding models.

The practical draw is hardware reach. QLoRA workflows in Unsloth let you fine-tune an 8B model on a single 12 GB consumer GPU, and the project headlines roughly 2x faster training with about 70 percent less VRAM versus baseline implementations, though the exact figures vary by model, GPU, and config. A 2026 update added faster mixture-of-experts training, with models like Qwen3-30B-A3B fine-tunable on about 17.5 GB of VRAM. It runs on NVIDIA (including Blackwell and DGX Spark), AMD, and Intel GPUs, with free Colab and Kaggle notebooks for trying it without local hardware.

It fits cleanly into the local-AI workflow. Unsloth integrates with Hugging Face transformers and TRL, and uses llama.cpp to save and run models, exporting to GGUF for Ollama or LM Studio as well as safetensors. As of 2026 it also ships Unsloth Studio, a local no-code GUI that covers the full lifecycle from dataset creation to training to running and comparing GGUF and safetensors models, with tool-calling, web search, and an OpenAI-compatible API, all running offline on Mac and Windows, with the core library under the Apache 2.0 license.