About the show — production notes from the first season.

One finished episode of this show, fully edited, fully transcribed, fully mailed back to the people who participated, costs about $42 in hard expenses to make. That number is the floor; it does not include any of the labor. The labor is donated, the labor is mine, the labor is two volunteers who help me edit and transcribe. The $42 is just the tape stock, the call charges, the envelopes, the postage, and the ten dollars I put on a phone account for somebody on the inside who otherwise could not get to a phone that month.

I am writing this as a production note for new listeners and for other producers who have asked, over the last year, how the show actually gets made. The honest answer is "slowly, by mail." Below is the longer version.

where the audio comes from

Most of the audio on the show comes from one of three places.

Outbound prison-phone calls.

The person on the inside calls a recorded line we set up through the facility's contracted phone provider — usually Securus or ViaPath, depending on the system. Their account has been pre-loaded with money I send to them. They call a number that rings into a small interview booth in San Diego. The call is monitored by the facility, on the facility's terms, and we make that clear to every participant in advance. We tell them, in writing, what monitoring looks like and what it means, and we ask explicitly whether they want to proceed knowing every word will be on the facility's record.

The hard cost of one of these calls runs between $3.15 and $11.40 per fifteen-minute block, depending on the state, the facility, and which contractor has the current contract. Federal calls are the cheapest. Michigan, after the 2023 rate cap, sits at about $0.21 per minute. Texas remains the most expensive. We budget $14 per planned interview and go over almost half the time, because a person inside who has been told that someone outside actually wants to listen is rarely a person who finishes the thought in fifteen minutes.

Cassette and digital recorder dropoffs.

Some of our participants are housed in facilities where the phone-line interview model does not work — solitary housing, long-haul transfers, facilities whose phone provider does not allow third-party numbers. For those participants we work in cassette. A volunteer mails the participant a small handheld recorder (when allowed) or a stack of cassettes (when the recorder is not allowed but a borrowable player is). They record onto the tape on their own time. They mail the tape back. We digitise it. The pipeline takes about a month per cycle. The recorder route is the better audio. The cassette route is the more reliable route.

I have written elsewhere about how durable cassette is as a medium for this work. The summary: a tape is unkillable, cheap, easy to share, easy to record over for the next cycle, and a player can be checked out from a facility library in most states. The audio quality is fine. We are not making music records.

Letters read aloud by their authors.

Some of the most affecting audio in the first season came from participants who wrote a letter first, with the understanding that they would record it aloud later. About two-thirds of the time they followed through. The letter-and-read approach gives the writer time to compose. It is the rare format that survives the loud, busy environment of a shared cellblock — because the writing happens in pencil on paper, on their schedule.

the transcription side

Every episode gets a full transcript. The transcript is mailed back, on paper, to the participants who appear in the episode, before the episode airs. They have ten days to read it and write back. They can request a quote be removed. They can request a quote be reworded if our edit changed the sense of it. They can request a pseudonym, even after the fact, if they think the episode might draw attention they did not anticipate. The default is that we honor the request. The rare cases where we have not honored a request involved factual claims about third parties that we needed to leave in for the integrity of the reporting; in those cases we discussed it with the participant and with counsel.

The transcript pipeline is run on a local Whisper deployment on a small box in my office, not on a third-party cloud service, because the audio includes voices of people whose participation we are obligated to protect. Cloud transcription services log audio in ways we cannot audit. Local Whisper does not. The model runs on the same Rock 5B edge-AI cluster I use for OSINT retrieval work, which makes the marginal cost of one episode's transcription about nine cents of electricity. I have written about that cluster elsewhere; it is the boring infrastructure that makes the careful work possible.

After Whisper produces a draft transcript, two humans pass over it. One is me. One is a volunteer named Mara who has been transcribing trial proceedings for twelve years and knows the rhythm of carceral speech better than any model does. She catches the slang the model softens and the acronyms the model spells wrong. The final transcript is hers more than the model's, and we publish it on the episode page when the episode airs.

the source-protection floor

Three rules govern what we publish about a participant. They are non-negotiable and applied to every episode.

First, we do not name a participant who is currently incarcerated without their explicit, written, on-file consent — re-confirmed within ninety days of air. People change their minds. People get transferred. People get out. The consent that was given in February does not necessarily still hold in August. We re-ask. If we cannot re-ask, the episode runs with a pseudonym, even if we previously had clearance.

Second, we do not publish geographic details that would allow someone with mediocre research skills to identify a participant beyond the level the participant chose to disclose. We will say "Marquette" if the participant said "Marquette." We will not say "the C unit on the upper tier" because that is the kind of detail that gets people retaliated against. The fact-checker on every episode has a checklist. It is not a soft pass.

Third, we do not publish anything from a participant that they have not heard or read first. Period. No exceptions. The cost of this rule is that some episodes take six months instead of two. The cost of breaking it would be that participants stop trusting us, and at that point the show no longer exists.

what the show does not take

We do not take sponsorship from corrections vendors. This includes prison-phone companies, commissary providers, electronic-monitoring companies, parole-supervision software companies, jail-management software companies, "behavioral analytics" vendors, and any private operator of any corrections facility. We do not take sponsorship from anyone who is a defendant in pending litigation that we are covering or might cover. We do not take sponsorship from politicians, candidates, or PACs. We do not take advertising or promoted-episode placements from anyone, including non-profits in adjacent spaces. The reason for the last one is simple — we cannot evaluate every nonprofit's funding chain in a way that would let us be confident the money is clean, so we just take none of it.

The show is funded by listener contributions, by a small grant from a foundation that has agreed in writing not to ask for editorial input (the grant is administered through the press's 501(c)(3) paperwork in flight at Closed Press), and by direct sales of season transcripts in printed booklet form to readers who like the show and like physical objects. The transcript booklets are the most reliable income line. They cost $14. They are printed in San Diego on the same paper stock as our other print work. Each one ships with a postcard the buyer can return, write a name on, and we mail a free copy of the season's transcript into a facility on their behalf.

The full economics of the show are publishable. If a listener asks, I send them the spreadsheet. There is nothing to hide and a great deal to copy.

what comes next

Season five is in scoping. The theme is the new private parole companies — the for-profit operators who have moved into the post-release supervision space as states have outsourced their probation and parole functions over the last decade. The reporting is already underway alongside The Counted Press and an investigator from a partner newsroom. Episodes will air through the second half of 2026.

If you are inside and you want to participate, write to the PO Box. The door is open by letter. We answer letters from inside facilities first.