A publication, written partly by a machine, on purpose.

Glitch & Glow covers AI, data, dev and cloud for people who build software and are tired of being talked down to. We assume you can read a benchmark, that you don't need a vendor's blog post translated into English, and that hype is the most boring thing in tech.

How this is made

Articles are drafted by Claude (Anthropic's model) running on a daily schedule. The model scans AI/tech news, papers on ArXiv, Hacker News, and selected primary sources, then produces 400–600 word drafts. A human edits, fact-checks, and decides what ships. The byline reflects this: most articles are signed The Editors, with Claude credited on each piece in the "Claude's Take" sidebar.

What you'll find on every article

• Claude's Take. A short, opinionated commentary written from the model's perspective — distinct from the body, transparently labeled.
• Sources widget. The primary sources we used, plus a credibility score (0–100) reflecting how well-sourced and verifiable the piece is.
• No tracking-as-a-business-model. Analytics are minimal. There are no third-party ad networks. Comments live on GitHub Discussions via giscus.

The editorial bias

• We'd rather be right than first.
• If a story can be told in 400 words, it won't get 1,200.
• Contrarian reads only when we actually disagree — not for traffic.
• Disclosure when we have skin in the game.

What we don't do

• Sponsored content disguised as editorial.
• Press-release rewrites.
• "X is dead" headlines about technologies that are clearly fine.
• Predictions presented as facts.

Contact

Tip line, corrections, or a contrarian take you want us to look at: tips@glitchandglow.dev. Source code, including how the publication pipeline works, is on GitHub. Pull requests welcome.