Find Medium Publications That Fit Your Story

Too Many Plausible Homes

A technology story may fit a software publication, a workplace publication, a founder publication, or a general-interest one. Their names rarely settle the question. The writer still has to inspect recent stories, topics, submission status, guidelines, and whether the publication accepts work that has already been published. Medium itself advises writers to check each publication's guidelines because requirements vary.

That work becomes awkward across a large catalogue. Medium's Boost Nomination Program list alone includes more than 100 publications and links to their submission guidance. Our tool starts from that kind of current source, visits each candidate, and records what it can establish. If a requirement is unclear, it stays unknown rather than being filled in with a convenient guess.

A Small, Visible Algorithm

Matching is deterministic and runs locally. Existing Medium tag overlap supplies 40 points, title and subtitle topic context supplies 25, acceptance of previously published stories supplies 20, recent activity supplies 10, and preferred publication size supplies 5. The weights are intentionally easy to inspect. Medium tags remain the strongest signal, and the vocabulary expansion around titles and subtitles is kept small.

A score out of 100 makes a long search easier to sort, but it must be read with its parts. Seventy-five points may mean excellent topic overlap with uncertain submission rules. A lower result may identify an active niche publication whose tags do not describe the story particularly well. The tool stores the component points, matched terms, and rejection reasons so the number never has to stand alone.

Built for Interpretation

This is where we think the project becomes more useful. The utility can produce a local report containing imported stories, active candidates, the strongest eligible matches, rejected pairs, and the reasons behind them. An agent can interpret that material, compare trade-offs, and explain why one publication looks more suitable than another. It gets structured evidence instead of being asked to browse Medium and improvise an answer from memory.

The division of work is deliberate. The algorithm applies the same scoring rules every time. The agent helps the writer understand a messy set of results in ordinary language. The writer makes the editorial decision. That is a healthier use of AI than asking a model to choose a publication, draft an application, and click through the whole process without showing its reasoning or the publication's current rules.

What It Cannot Know

Publication suitability is not story quality. A score cannot tell whether an argument is original, whether the opening works, or whether an editor has already received five similar pieces that week. It also cannot turn an inactive publication or closed submission route into an opportunity. The tool treats those obvious conflicts as hard rejections rather than awarding a hopeful score.

Medium's interface and guidelines can change. Some pages do not expose every detail cleanly, and publication prose cannot always be classified with confidence. The assistant therefore starts in dry-run mode, stops on ambiguous controls or changed guidelines, and requires explicit approval before external actions. We would rather leave a writer with one manual check than let brittle automation send the wrong story somewhere public.

An Open-Source Contribution

We made the project local and open source because publication matching should not require handing a third party every draft, browser credential, and editorial decision. The tool keeps its own state in a local SQLite database and uses Medium's normal browser interface for sign-in. It does not ask for or copy passwords, cookies, or session tokens, although the persistent browser profile still needs to be protected like any signed-in browser.

The public repository also lets people challenge the weights, maintain selectors when Medium changes, or adapt the reporting to their own workflow. That is the point of a small algorithm: its assumptions are visible. Use it to reduce a huge search to a defensible shortlist, let an agent help read the evidence, and keep the final choice with the person whose name is on the story.