Independent reporting on artificial intelligence.


The Frontier Wire

Topic

Open source

19 articles


Editorials

Open-weight AI just lost its most powerful corporate patron

Jensen Huang made Nvidia open-weight AI's biggest corporate backer and its loudest voice in Washington. The commercial logic behind Nemotron and the Hugging Face deal survives him. The political defence does not, just as export policy turns, and a campaign to reinstate him will not fill that gap.

17 min read

News

LiteLLM's Rust gateway benchmark, checked against LiteLLM's own data

LiteLLM says its Rust gateway has about 7x lower overhead, 9x less memory and 4.5x more requests per dollar than the next-closest gateway. LiteLLM's own benchmark files show a latency and throughput near-tie, a mislabeled widest-pair memory chart, a modeled cost and a config gap, plus Rust leads on CPU and memory.

17 min read

Comparisons

Best Claude Code skills for teams shipping LLM apps and agents

A Claude Code skill is a folder of instructions the agent loads only when a task calls for it. Six MIT-licensed skills for security, observability, cost, MCP, coding-agent governance and skill management, ranked on stated criteria, installed and linted in a clean repository.

14 min read

Tools

npx skills and skills.sh: a complete guide to the skills CLI

The skills CLI installs SKILL.md folders from GitHub into 79 coding agents, and skills.sh is the directory built on its telemetry. Every command in version 1.7.4, project and global scope, the symlink layout, the lock file and its hashes, and the warning printed on every install, checked against the source.

13 min read

Comparisons

The 10 best agent observability tools in 2026, compared

An AI agent makes dozens of model and tool calls for one task, and a failure can sit in any of them. Ten observability tools, from gateways that see every call to SDK-based tracing platforms, compared on published criteria and the OpenTelemetry conventions they follow.

20 min read

Comparisons

OpenRouter alternatives in 2026: seven gateways compared

OpenRouter made one API key for hundreds of models the default way to prototype. Credit fees, a third party in the data path and governance that lives in someone else's account send production teams looking for alternatives. Seven options, ranked against published criteria.

18 min read

Research

Self-hosted LLMs: how to run a local LLM, from laptop to GPU cluster

Running a model on your own hardware is now a configuration problem more than a research one. The hard parts are sizing GPU memory, choosing between Ollama, llama.cpp, LM Studio, vLLM and SGLang, reading the licence, and putting a production API in front of a server that was built for one user.

19 min read

Comparisons

Best agent orchestration frameworks in 2026: 10 open-source options

Agent frameworks have converged on the same feature list, but they still differ in how they model control flow, where state lives and how much they assume about your model provider. Ten code-first, open-source frameworks, compared against published criteria with GitHub activity checked on 5 October 2026.

21 min read

Comparisons

Top 10 open-source AI gateways for self-hosting in 2026

An open-source AI gateway is only as open as its license file and its paid tier allow. Ten self-hosted LLM gateways, read at the level of the LICENSE file, the dependency list and the release history, with criteria published before the ranking.

19 min read

Surveys

The state of open-weight licensing, mapped

Most flagship open-weight releases of the past year ship under Apache 2.0 or MIT. The exceptions carry user caps, revenue gates, prohibited-use policies and remote kill switches. This survey reads the licence texts, not the launch posts, and maps what each one actually permits.

13 min read