Source: https://toddschiller.com/artifacts/one-way-door/index.html

[← toddschiller.com](https://toddschiller.com/)

# One-Way vs. Two-Way Door Decisions

A *one-way door* decision is hard or impossible to reverse. A *two-way door*
can be walked back cheaply. This explorer asks
[Jev](https://docs.typesafe.ai/introduction),
the decision model from TypeSafe, to classify a decision, rate how much work it would take
to undo, and say how much human oversight it warrants. Reversibility is the reason this
matters for AI. The cheaper a decision is to undo, the safer it is to let an agent make it
alone.

## Connect to Jev

Calls run in your browser

OpenRouter API key

Model

typesafe/jev-1.13
typesafe/jev-latest

Remember on this device

Jev runs through
[OpenRouter](https://openrouter.ai/typesafe). Get a key from your OpenRouter dashboard. It is sent only to OpenRouter, never
here.

## Jev's spectrum map

Every example, placed by Jev

Re-map all with Jev

Two-way (reversible)
Toss-up
One-way (irreversible)

## Where Claude and Jev disagree

Jev's one-way probability minus Claude's

## The decision to evaluate

The only thing sent to Jev

Describe the action


This text is the state Jev reasons over. Only the decision
text is sent to the model, together with the three questions below. The reference
reading is not part of the prompt.

Ask Jev

## Jev's reading

One-way door probability

Two-way door (reversible)
One-way door (irreversible)

Work to undo it
Jev's probability across an ordered effort scale

Recommended oversight for an AI agent

Show the raw Jev response

## Claude vs Jev

Same decision, two models

Claude's guess at the reversibility, and it can be wrong the same way Jev can. Not
passed to Jev.

### The questions Jev is asked

Edit them to run your own experiment

Each run sends the decision above plus three typed questions. Jev returns a
[Noul](https://docs.typesafe.ai/primitives/noul)
(a yes/no probability), a
[Score](https://docs.typesafe.ai/primitives/score)
(a position on an ordered scale), and a
[Choice](https://docs.typesafe.ai/primitives/choice)
(one option from a set). All three are typed answers with probabilities, not free text.

### What you're looking at

**The metaphor.**
The one-way / two-way door framing comes from Amazon's
[2015 letter to shareholders](https://www.sec.gov/Archives/edgar/data/1018724/000119312516530910/d168744dex991.htm). Two-way-door (reversible) decisions can be made fast and delegated widely, because
if you're wrong you just walk back through the door. One-way-door decisions are
consequential and near-irreversible, so they deserve slow, careful deliberation.

**It's a spectrum, not a binary.**
"Reversible" hides a range of cost. Flipping a
[feature flag](https://en.wikipedia.org/wiki/Feature_toggle)
off is instant and free. Reversing a
[credit-card charge](https://en.wikipedia.org/wiki/Chargeback)
takes a refund or a dispute but is routine. A
[Bitcoin transaction](https://en.wikipedia.org/wiki/Bitcoin#Transactions)
settles on-chain with no one able to claw it back, and an
[83(b) election](https://en.wikipedia.org/wiki/83(b)_election)
with the U.S. Internal Revenue Service (IRS) is generally irrevocable. The "Work to
undo it" scale is where this artifact tries to place a decision.

**Why it matters for AI.**
If an AI agent picks the wrong two-way door, the cost is a quick reversal. If it picks
the wrong one-way door, the cost is permanent. So reversibility is a natural gate for
autonomy: let agents act freely on cheap-to-undo decisions, ask for human confirmation
on costly ones, and keep irreversible ones with a person. That's exactly what the
"oversight" question probes.

**Reversibility can change out from under a model.**
A decision's door type is not fixed for all time. An outside catalyst, often a change
in law or policy, can turn a two-way door into a one-way door. Converting a
traditional individual retirement account (IRA) to a
[Roth IRA](https://en.wikipedia.org/wiki/Roth_IRA)
used to be reversible. You could unwind it with a recharacterization. The 2017 Tax
Cuts and Jobs Act removed that option starting in 2018, so every conversion is now
final. A model trained mostly on older text can be confidently wrong here, reading the
decision the way it worked before the change. The
catalyst
tag on an example marks a case where the answer moved after some cutoff. Treat it as a
prompt to check whether the model's world is still current.

**Is Jev any good at it?**
That's the open question. Every example ships with a
*Claude analysis*: Claude's own call on how reversible the decision is, and
why. Hold Jev's numbers against it. Both sides are models guessing, and either one can
be wrong.