Full transparency

How the decoder actually works

Last updated July 2026 · Clinical review: underway

Every decode runs the same four-stage structure, built on the dialectical behavior therapy skill Check the Facts: separate what was stated from what was added, weigh alternative readings with honest odds, ask three fact-checking questions, then spend sixty seconds bringing the body down. The AI fills in the content for your specific situation; the structure never changes. This page documents that structure, the rules the AI operates under, and the limits of the tool.

The basis: Check the Facts

Check the Facts is an emotion-regulation skill from dialectical behavior therapy. Its premise: intense emotions respond to our interpretation of events, not the events themselves, so before acting on a big feeling you audit the interpretation that produced it. The skill was designed for people whose emotions arrive fast and at full volume, which is why it maps onto rejection sensitive dysphoria better than generic reassurance does. It never disputes that the pain is real; it disputes only whether the story earned it. A longer plain-language explanation is in this article.

The four stages of every decode

1. Stated versus added. The literal, observable content of your trigger goes in one list. The predictions, mind-reading, and conclusions your brain inserted go in another. This is the DBT move of describing the prompting event in observable terms, made visual.

2. Weighted readings. Four interpretations of the same trigger, each with a probability. One of them is always your spiral's own version, given its honest odds rather than dismissed to zero, because a feared reading at 8% is believable and a feared reading waved away is not. The alternatives must be specific to your situation; generic optimism is treated as a failure.

3. Three questions. Fact-checking questions tuned to your exact trigger: what evidence supports the fear versus contradicts it, what you would tell a friend showing you the same message, and whether the feared reading matches the other person's actual track record. You can switch the framework to CBT-style thought records or body-first prompts; the audit-the-interpretation spirit stays constant.

4. Sixty seconds for the body. One regulation exercise matched to the flavor of the spiral: paced breathing for panic, grounding for shame loops, muscle release for anger. A reframe you cannot feel yet is not finished, so the decode ends in the body, not in text. The breathing actually begins earlier: the loading screen is a paced-breathing guide, so regulation starts before the result arrives.

Two paths: the library and the live decode

The decoder has two ways of producing a result. The instant library covers ten of the most common RSD triggers, from the vague boss message to the one-letter reply, with decodes written in advance by a human, in full, in this exact structure. Tapping one serves it immediately: no AI call, no wait, no free-decode used, and nothing sent anywhere, because the entire decode already lives on the page. Every word of the library is reviewable verbatim, which is precisely what the clinical review covers first. The live decode handles everything else: whatever you paste is sent to the AI once, with the instruction set below, and returns a decode built for your exact situation. The structure is identical on both paths; only the authorship differs, and the result is labeled so you always know which one you're reading.

What the AI is instructed to do, and never do

The decoder runs on a fixed instruction set sent with every request. It requires the AI to validate that the pain is real while questioning only the story, to reference your actual details rather than produce generic comfort, and to keep every reading genuinely plausible. It forbids the phrases and moves this community has learned to distrust: "calm down," any suggestion that you are oversensitive, and therapy cliches. It also requires the spiral's own interpretation to appear with honest odds, because the tool's credibility rests on never pretending your fear is impossible.

The crisis policy

The same instruction set requires the AI to watch for signs that what you pasted is heavier than an RSD spiral: suicidal thinking, self-harm, severe hopelessness. When it sees them, the decode does not run. You get a plainly worded message and direct links to call or text 988 in the US, with findahelpline.com for everywhere else. Crisis resources are never gated, never count against the free limit, and never sit behind the Pro unlock. A reality-check tool is the wrong instrument for a crisis, and the tool says so instead of pretending otherwise.

Privacy, plainly

What you paste is sent to the AI for one decode and is not stored by this site. There is no account and no analytics on your words. The free-decode counter lives in your browser. The optional history feature (the Vault) stores past decodes on your device only.

Clinical review status

A licensed, DBT-trained therapist is reviewing this methodology: the four-stage structure, the instruction set, and the crisis policy. Their credentials and any changes they require will be published on this page when the review completes. Until then, this page describes the tool as it is, and the honest label for its current state is: built on an established DBT skill, engineering complete, clinical review in progress.

The limits, stated outright

RSD Decoder is a self-help utility, not therapy, diagnosis, or treatment, and it does not replace any of them. It is built for a specific, narrow moment: the minutes after a trigger, when structure helps and nothing longer is possible. If the spirals are frequent, if they are costing you relationships or work, or if the tool becomes something you need many times a day, that is real information, and it points toward support from an actual professional. The decoder will keep saying that, because a tool this small being honest about its size is the whole point.