Almost every coach, practitioner and serious seeker I know is now using AI somewhere in their
inner work, and almost none of them has a test for when to stop believing it. The instruction
they have instead is the one this field repeats more than any other, which is to trust
themselves.
Trust yourself is half an instruction. Delivered on its own it is indistinguishable from
permission to guess, which is why the people most likely to take it literally are the people
least well served by it.
The missing half is a check. Choose, then look at what happened. Trust yourself, then verify.
Somebody who does both is exercising discernment. Somebody who only does the first is guessing
with more confidence than a guess deserves. And that distinction is about to matter a great deal
more than it used to.
WHAT AN ORACLE IS FOR
An oracle answers. That is its function, and its authority rests on the fact that you cannot
check the answer. You go to it precisely because you lack the information yourself, and the
value proposition depends on you accepting what comes back.
An instrument reports. A thermometer does not tell you what to do about the temperature. It
tells you what the temperature is, and everything after that remains yours. Crucially, an
instrument can be tested. You can put it in ice and see whether it reads zero. Its authority
comes from being checkable, which is the reverse of how an oracle’s authority works.
Both of these things now exist in software, they can sound almost identical, and the difference
between them is doing far more work than most people using them realise.
THE VERSION OF THIS I BUILD
In the programmes I run, there is a tool that knows the full sequence of initiations, asks a
person questions about where they are, and suggests what they might be ready for next.
Described in one sentence that sounds exactly like an oracle. It is not, and the difference sits
entirely in what happens after it answers.
The suggestion is a reading, not an instruction. The person is told, before they ever use it,
that their own sense of readiness outranks anything the tool produces. They are told to choose
something, do it, and then stop and ask themselves how that landed before going any further. If
they are uncertain, the correct action is to say so rather than to push on. And there is a
human, me, available to be asked, whose stated aim is to become progressively less necessary.
Take those away and the same tool becomes something else entirely, without a single word of its
output changing. That is worth sitting with, because it means you cannot tell an instrument from
an oracle by reading what it says. You can only tell by looking at what surrounds it.
THE TEST THAT ACTUALLY SORTS THIS
The question that does the most work is not whether a tool is accurate. It is what happens when
you disagree with it.
An instrument has an answer to this. Your reading is the senior one, here is how to check, here
is what would tell you the tool has got it wrong. An oracle has no answer, or worse, has an
answer that reframes your disagreement as the thing that needs fixing. You are resisting. You
are not ready to hear it. Your scepticism is the block.
Watch for that second move particularly, because it is the identifying mark of everything
harmful in this field, and it long predates the technology. The faculty you would use to detect
a problem is defined as the problem. Once that has been established, no evidence can reach you,
because the mechanism for evaluating evidence has been disabled by agreement.
WHAT THIS ASKS OF A PRACTITIONER
The uncomfortable version of this question is not about software at all.
Any practitioner can occupy either position. If a client comes back and says the thing you named
did not fit, what do you do with that. If the honest answer is that you explain why they are not
yet able to see it, you are operating as an oracle, and the tool you build will inherit that
whether you intend it to or not.
The direction I want to move people in is from needing an oracle to not needing one. That is a
strange commercial position, and I hold it because dependence is the failure mode this entire
industry is built on and mostly declines to discuss. A practice that produces people who can
read their own situation is a smaller business than one that produces people who cannot. It is
also the only version I am willing to run.
THE SENTENCE UNDERNEATH ALL OF THIS
Trust yourself, then check.
It sounds mild. Applied properly it disqualifies a great deal, because it insists that your own
discernment stays in your hands throughout, and that anything asking you to put it down is
asking for the wrong thing. That test does not care whether the thing asking is a person, a
tradition, or an extremely fluent piece of software.
The tools are going to keep improving. The question of who is holding the judgement will not get
easier as they do.
