How The Chonkerton works

And what it won't do. The creature is goofy; this page is the part that isn't.

The short version

The Chonkerton turns recent news into a short audio briefing you control. You pick the topics you want, ban the ones you're sick of, and say how many minutes you've got. It builds a fresh bulletin from stories it just pulled, reads it aloud, and — this is the load-bearing part — it ends.

The creature is goofy. The plumbing underneath is deliberately boring and inspectable. Here's all of it.

What it actually does, step by step

1. Pulls recent stories from a hand-picked list of sources, on a short refresh cycle, all day.
2. Summarizes each one into a brief script — new writing that states the facts, not a copy of the article.
3. Reads it aloud in one of our narrator voices.
4. Assembles your briefing when you press play: it takes the recent stories that match your topics, packs them into the length you asked for, and skips anything you've already heard or banned.
5. Stops. There's no autoplay rabbit hole, no "up next," no feed to fall into. When it's done, it's done.

Where the news comes from

This is the part people are right to scrutinize, so here's the whole rule.

Every source has to fit one of three buckets, or it doesn't run:

• Open RSS feeds. Publishing an open feed is a publisher saying, in machine-readable terms, "pull this and reuse it, with attribution." That's the default for most sources — newsrooms, blogs, podcasts.
• Author consent. If a creator doesn't publish an open feed, we email them and get explicit permission before featuring their work. We record when and how.
• Public domain, or reporting on a public statement in our own words — the same basis a journalist uses to tell you what someone said, with attribution and a link, without reproducing their work.

Every story is attributed out loud ("from TechCrunch…", "on Simon Willison's blog…") and linked to the original on its page.

What we never touch

• Paywalled outlets — NYT, WSJ, Bloomberg, paid Substacks, paywalled Medium — even when we personally subscribe. A subscription buys reading rights, not the right to redistribute. The only way paywalled work gets in is the author handing us a public copy themselves.
• Sites with no feed, scraped from their HTML.
• AI content farms of murky origin.
• Bulk social-media firehoses. (One creator who cross-posts to their own feed is fine. Hoovering a platform's API is not.)

When we're unsure about a source, the answer is no. It is always fine to not have a source; it is never fine to keep one we can't stand behind. A background audit re-checks active sources and automatically pulls any that drift — added a paywall, changed their terms, took their feed private — out of rotation until a human re-clears them.

What's AI here — plainly

Two things are AI, and we'd rather just tell you than have you find out:

• The summaries. A language model writes each briefing script from the facts of the story.
• The voices. The narrators are text-to-speech — our "cast." Some run on a graphics card in a spare room; that's a cost decision, not a secret.

Some things are not AI: the source list is picked and audited by a human, the sourcing rules above are a human policy, and which stories are flagged sensitive (so the jokes stop) is a human guardrail.

We label the AI because hidden AI is the thing that deserves distrust. Disclosed AI you can check — which is the entire point of the next section.

So should you trust the summaries?

Treat a briefing like a sharp friend catching you up over coffee, not a citation you'd put in a footnote. The summary is faithful to its source, but summaries compress, and compression loses things.

So for anything that actually matters — a number you'll act on, a quote you'll repeat, a claim that sounds too clean — the original is one tap away on every story's page. Go read it. We built the links to make checking easy, not to look diligent. A news product should make verification less theatrical, not more.

Mistakes, corrections, and takedowns

We will get things wrong sometimes. Here's the recourse:

• Spot an error? Email [email protected] and we'll fix or pull it.
• Are you a creator or outlet who wants out? We remove you within 24 hours, no negotiation, and we don't re-add you without fresh permission. Same address.

Your data

• No third-party trackers, no ad pixels, no cross-site tracking — so no annoying consent banner. The one cookie we set is a first-party sign-in cookie that keeps you logged in; it does nothing else and goes nowhere else.
• We store your account and your topic choices — that's what makes the briefing yours.
• Email is used for sign-in links and, unless you opt out, a short morning note when your briefing is ready. One click in any of those emails turns the morning note off for good; sign-in links keep working.
• You can export everything or delete your account yourself from account settings — deletion cancels any subscription and takes your data with it.

Who makes this

A solo builder — Damlab, under Conch Labs LLC — with a lot of help from AI coding tools and one bare-metal server. Not a newsroom, not a hedge fund, not a forty-person growth team. If something here matters to you and this page didn't answer it, write to [email protected]. A human (the one who built it) reads those.

See also: FAQ · Terms of Service · Privacy Policy.

The radio is a creature. The sourcing is boring on purpose. That's the deal.