Contact InferenceWeekly

We read the papers so you can spend your week building. If something we published was wrong, unclear, or missing the artifact link you needed, tell us. Corrections get priority over everything else in the inbox.

Where to send things

• Corrections and clarifications: [email protected] — include the article URL and the specific claim. If a paper’s reported baseline is weaker than it looks, we want to know before our readers rebuild on it.
• Paper suggestions: [email protected] — send the arXiv link or lab blog post. Tell us in two sentences what changed in your own work because of it. That last part matters more than the paper’s citation count.
• Reproducibility notes: [email protected] — if you ran the code and it broke, or it ran and the numbers didn’t match, send the commit hash, environment, and what you saw. We publish these notes alongside the write-ups.
• Everything else: [email protected]

What we do with a tip

Every suggestion gets read. We triage on one question: does this change a decision an engineer would make this quarter? Papers that only move a benchmark number usually don’t make the cut. Papers that change a training recipe, a data pipeline, or an evaluation habit usually do.

If we write about your work, we’ll link the code, the weights, and the paper. We’ll also say plainly whether we got the repo running ourselves, and on what hardware. If we couldn’t reproduce it, we’ll say so in the piece rather than quietly dropping it.

Response times

Corrections: same week, usually faster. Tips: we read daily, but replies come in batches. If you’re a grad student with a paper that hasn’t found an audience, say so — those get a closer look than the press releases do.

We don’t take sponsored posts, and we don’t run guest articles that are really product announcements. If you’re pitching one, the answer is already no.