Your read on each client, made explicit.
Tacit keeps one evolving judgment model per client — built from the fragments you already collect, checked against the hires that actually happen. It’s the part of your practice that has always lived in your head, finally on the record.
Early access · A small design-partner cohort · Built for boutique and retained search
// illustrative panel — every real model is built from a partner’s own engagements
Built around how a search actually runs.
Not a form to fill before you start. A loop that runs alongside the engagement you’re already working.
Collect
Forward the JD, paste the WeChat message, note what the founder said on the call. Fragments, as they come.
Read
The client’s model takes shape as a document — dimensions, evidence, and honest blanks where evidence is thin.
Evaluate
Candidates are assessed against this client’s model, not a generic rubric. Every judgment cites its evidence.
Hear back
An offer, a rejection, a remark in passing. Each outcome is reconciled against what the model expected.
Recalibrate
Where reality diverged, a revision is proposed — and shown to you before it takes effect. Then the loop runs again.
Each pass through the loop, the model gets a little closer to how this client actually decides.
Quiet, precise, and honest about its limits.
Tacit is built for people who already know their craft. It behaves accordingly.
One model per client
Not an average of the market — a specific, evolving record of how this client decides. Readable end to end, transferable on your terms, and yours.
Fragments in, structure out
No forms, no tags, no cleanup ritual. Messy input is the normal case, not the exception — organizing it is the product’s job, not yours.
Nothing changes silently
When an outcome shifts the model, the revision appears for review first: apply it, or keep the current model. State never moves behind your back.
Comfortable saying “not enough evidence”
Dimensions without sufficient evidence stay blank, and low-confidence judgments are labeled as such. A grounded “no judgment yet” beats a fluent guess.
The ones we actually get.
Is this another AI resume screener?
No. Screeners push volume through generic criteria. Tacit works on the opposite problem — a small number of high-stakes placements, judged against one specific client’s revealed preferences. The measure of value is how precisely the model reflects this client, not candidates per hour.
Where does the judgment come from — me, or the AI?
From your inputs and your clients’ outcomes. Tacit organizes the evidence and surfaces the patterns in it; it doesn’t substitute its own taste. When it proposes a model revision, the change is visible before it takes effect, and applying it is your call.
What does the model actually look like?
A readable document per client: the dimensions that drive decisions, the evidence behind each, a calibration history, and honest blanks where evidence is thin. If you can’t read it, you can’t trust it — so everything is readable.
Couldn’t I just do this in ChatGPT?
For a single evaluation, roughly — and that’s a test we hold ourselves to. What a stateless chat can’t do is persist: hold a model that survives across sessions, accumulate one client’s outcomes over months, and reconcile past predictions against what actually happened. The persistence is the product.
Does Tacit contact my clients or candidates?
Never. It works only on the material you bring back from your relationships. The relationships stay yours.
What does it cost?
There’s no public pricing yet. The current cohort is a paid design partnership with terms agreed directly with the founder. If that sounds premature for where you are, it probably is — write to us anyway and say so.
Start with the client you know best.
Bring one client and the fragments you already have. If the model can’t earn its place on that engagement, you’ll know within weeks — and so will we.