Useful to include: which app you are using (iPhone, Mac, Apple TV or Apple Vision Pro), the operating system version, and what the app did rather than what you expected it to do.
Please do not paste the decision you were analysing unless you want to. A description of the behaviour is enough, and your questions are your own business.
Running a simulation
It is taking a long time
The swarm is computed on your device. A Premium run is 600 agents over 32 rounds and an older iPhone will take noticeably longer at that than a recent Mac. Keeping the app in the foreground lets it use the processor; switching away suspends it.A free run is 40 agents over 12 rounds and is fast on any supported device. The wait people notice is the Premium configuration.
The result was the same when I ran it again
That is correct and deliberate. The simulation is seeded with a hash of your question and context text, so identical input always produces an identical run. If you want a different run you have to change the input, not repeat it.Two people who type exactly the same words get the same simulation on different devices, and the same run repeats identically months later. That is the point of seeding from the text rather than from the clock.
The corollary: if you change the wording and get a different answer, that is the seed changing, not the decision being unstable. It is not a robustness test.
The analysis missed something obvious
Orakl only knows what you typed. Add the missing consideration under “Add context” and run it again; the driver list, and usually the result, will change accordingly.
Confidence came back at 0%
A real signal rather than an error. Confidence is one minus the entropy of the vote, so a near-even result scores close to zero even when one side is nominally ahead. On a two-option question a 60/40 vote scores 2%. Reading the numbers explains why, and what to do about it.
Every figure came back at 100%
Usually a sign that the question was too short for the extractor to find more than one driver, in which case there was nothing for the swarm to disagree about. Type more context and run it again. Example three shows the failure mode in full.
The outcome labels look cropped, like “Stay where”
That is the deterministic extractor, which splits an “A or B” question at the word “or” and takes two words from each side. It runs when the on-device language model is unavailable. Rephrasing the question so the two options are short and adjacent — “Should I accept or decline?” — gives it better material.
Purchases and devices
Restoring a purchase
Settings inside the app → Restore Purchases, signed in to the Apple Account that made the purchase.
I bought it on iPhone — do I get the Mac app?
No. They are separate products on the App Store and are purchased separately.
My history is not on my other device
By design, and it is not something we can switch on. Nothing syncs because nothing leaves the device it was written on. The same question run on two devices will give the same result; it will not give you one shared archive.
Subscription management and cancellation
Handled by Apple, in the App Store settings of your Apple Account, not inside Orakl. We have no access to your subscription and cannot cancel or refund it on your behalf.
Data and deletion
Everything Orakl holds is on the device you typed it into. There is no account to close and no server copy to request. Deleting an entry from History removes it; deleting the app removes all of it. The full text is in the privacy policy, including the one optional setting that can send anything anywhere.
Orakl is a structured way to think through a decision. It is not financial, legal, medical or professional advice, and the limits page is worth reading before you rely on a result.