Dealoneby Belkins
Proposed method · intake not open

The proposed method, published.

A design you can inspect beats a claim you have to trust. This page specifies the work Dealone would perform, in order, if the release gates clear and Cohort 001 opens.

The proposed Cohort 001 is four seats, run by hand by one person. Nothing on this page describes software finding buyers. It describes a Belkins operator reading pages you could open yourself, with a model helping him read faster and draft first versions. Intake is not open.

01

Four moves. Fourteen days. Eight hours.

The proposed Deal Sprint is $1,000 with a draft cap of eight operator hours and a 336-hour delivery clock from kickoff. No contract, checkout, or active delivery workflow exists today.

MoveWhat it coversOperator hours
01Kickoff, readiness verdict, signal review3.5
02Shortlist and enrichment1.5
03Voice kit, sequence, deliverability pre-check2.0
04Delivery call1.0
Total8.0

The proposed eight-hour scope includes two Dealone calls, kickoff and delivery. It excludes buyer calls. Those would sit in the separate proposed Live Desk at $600 a month, capped at three months. Neither offer is currently available.

02

The four moves, in full.

01
Move one

The operator would read the public record for people already saying the problem out loud.

The proposed move would begin with a 60-minute recorded kickoff. The founder's product description and buyer guess would be logged before research so the guess could later be scored against evidence.

The intended readiness verdict has three states: GO, FIX FIRST, or REFUND. The exact response times and refund action remain draft operating terms until the control plane is verified.

A person would then open pages available to a logged-out visitor. A model could structure results, but the operator would accept or reject each row and record the reason. Rejected rows would remain as evidence of what did not qualify.

Proposed artifact
A written readiness verdict and buyer-signal map whose kept and rejected rows carry their evidence and reasons.
Proposed timing
Kickoff on day 1 or 2 after a future validated intake, with the signal map targeted within 72 hours.
Founder inputs
Existing writing for the voice kit and a fixed excluded-account list. No intake for either exists today.
02
Move two

Three best-evidenced buyers, each with the dated thing that put them on the list.

The intended process would cluster kept signals by hand into buyer patterns, then select up to three accounts carried by a specific individual and dated artifact. The final selection would remain human.

Every proposed candidate would have to clear all six points on the evidence bar below.

Proposed artifact
One evidenced page per buyer plus a short walkthrough explaining the selection and its disqualifiers.
Proposed timing
Within 48 hours after the signal map, targeting day 7.
Founder role
A future founder would approve or veto the selection and record the reason.
03
Move three

Four proposed messages in the founder's words, sent by the founder.

The proposed voice kit would use founder-supplied writing to identify phrasing and banned words.

A model could draft first, with a human rewrite measured and shown as a provenance byline. The intended sequence would use four touches on days 0, 3, 7, and 14, all approved and sent by the founder.

A future deliverability pre-check would block outreach when SPF, DKIM, or DMARC failed. Contact, bounce, and complaint limits would require verified monitoring before any message could be queued.

Proposed artifact
A voice kit, founder-approved four-step sequence, provenance bylines, and deliverability pre-check result.
Proposed timing
In parallel with move two and before the day-14 delivery call.
Founder role
Read, edit, approve, and send every external message.
04
Move four

The part everyone else leaves you to do alone.

In the proposed sprint, move four would be the delivery call where artifacts and the operator's current assessment are handed to the founder.

In the proposed model, buyer calls would sit in the Live Desk, a separate $600 monthly opt-in, capped at three months. We are saying this here rather than burying it, because "we are on your sales calls" is the most attractive sentence on our whole site and it is not what the $1,000 buys.

A future desk would target a prep sheet before each buyer call and would reschedule when prep is missing. The intended artifact would carry evidence, objections, price timing, one required question, and the next step.

The operator would be a silent observer unless introduced. The operator would never pose as the founder. Any recording would require the buyer's explicit consent; otherwise the operator would take notes only.

Proposed deliverable
A delivery call inside the sprint. On a future desk: prep, objection and reply artifacts, with every external message sent by the founder.
Proposed timing
Delivery call on day 14. The intended desk cap is six operator hours each month.
Founder responsibility
The founder would stay on every buyer call, state the price, and sign the customer. Dealone would not replace the founder or send as them.
03

What has to be true before we call someone a buyer.

Three buyers is the thing your $1,000 buys, so it needs a definition tight enough that you can audit it yourself and tell us we got it wrong. First, the word: a buyer is a named person. Nothing else earns it.

TierWhat it isWhat we are allowed to call it
1A named individual, at a named company, with a dated public signal and a verified reachable channelBuyer. Counts toward your three.
2A named company and a known role, no individual identified yetTarget. Never presented to you as a buyer.
3A role at a type of company, such as ops managers at 20 to 50 person shopsSegment. An input to the message, not a deliverable.

You can already describe your segment. That is not what is missing. What is missing is a name and an inbox next to it, so tier 2 and tier 3 do not get to wear the word.

All six of these must hold. Five is a fail.

1. A dated artifact
Authored by that person or directly attributable to their organization, inside that signal type's freshness window. The named-individual requirement is carried by the tier table above: nothing below tier 1 gets to the bar at all.
2. Their words, not ours
The problem stated verbatim, with a live URL and the publication date on it. Our inference about their problem is not evidence. Their sentence is.
3. Function ownership
Their title, or hiring-manager attribution on the job post, or their stated role on the review, shows this person owns the problem or approves the spend.
4. A verified channel
The address passes a deliverability check before it ever enters a sequence.
5. No conflict
Not a competitor, not an existing customer, not someone you already approached. Checked against the list of contacts you named at kickoff. We keep it inside the six rather than beside them, because a name you already own is not a buyer we found.
6. Serviceable size
Inside the company-size band you can actually deliver to. A buyer you would drown servicing is not a buyer.

On freshness windows. Each signal type has its own, because artifacts go stale at different speeds. Every freshness window we currently use is an assumption, set from how quickly the underlying artifact stops describing a live situation. None of them is measured yet, and all of them get re-fitted against observed reply rate by artifact age once there is data to fit against. We would rather publish that they are guesses than let you assume they are findings.

Proposed missing-buyer term. If only two buyers cleared the bar, the draft terms would offer two evidenced buyers plus $333 back, or a founder-elected full sprint refund. This refund control is not yet operational.

04

Three kinds of public evidence, and what each one is actually worth.

Everything below is a page a logged-out visitor can open. A person opens them and reads them. There is no crawler, no monitor, no scraper and no harvest anywhere in this product, and that is not a compliance posture, it is a description of how four seats run by one person actually works.

The complaint signal

Public forum and community threads

In niche professional places rather than startup places. Reddit is the obvious one and we say so rather than hiding it behind a euphemism, because reading a public thread is not a restricted act and pretending otherwise would be the tell.

Tells you: the problem in the buyer's own unprompted words, which is the vocabulary your opener should be written in. At its strongest, someone walked into a room of strangers and asked for a vendor.

Does not tell you: that they have budget, or that they own the decision. Plenty of loudly frustrated people cannot sign anything.

The spend signal

Public job ads and contractor postings

Tells you: money is already leaving the building for this outcome. A job description listing the manual task by name means someone approved a salary to solve it, and your price is a rounding error against that salary. A contractor posting is better still, because it puts an hourly rate on the exact output your product produces. That is the purest evidence of willingness to pay that exists in public.

Does not tell you: dissatisfaction. Nobody in a job ad says the current approach is failing. And the post stops being true the day the role is filled.

The displacement signal

Public review pages for what they use today

Tells you: this person already pays for this category, which removes the hardest objection in a first sale, and they have written down in public the thing their vendor does not do. Renewal or migration language on a recent review means there is a date on their calendar you did not have to create.

Does not tell you: who they are. Reviewers are usually partly anonymized, which is exactly why these start life as tier-2 targets and only become buyers if a named individual can be established through licensed data.

Where contact data comes from

Licensed providers, under a DPA

Contact details are bought from licensed B2B data providers, under their terms and under a data-processing agreement. Nothing is harvested off a platform. If a licensed provider does not have a working address for a person, that person does not become a buyer, no matter how good the signal was.

05

What is a person, what is software, and what would have to be true before that changes.

The fourth column is the one worth your time, because "we will automate this later" is free to say and means nothing without a number attached. Every row has a measured trigger, and a step cannot be automated before its trigger fires. The trigger has to be a saved view returning a threshold with a stated n. Not a feeling.

Two rules sit above the whole table. No pattern is automated until it has shown up in at least three different founders, because a pattern from one founder is a fact about that founder's product, not about the market. And automation is revocable: if an automated step's quality metric drops below its trigger for two consecutive weekly reviews, it goes back to manual within 48 hours and stays there until the threshold is cleared again.

StepTodayWhat changes laterWhat would have to be true first
Signal searchTooling plus a personScheduled searches file raw results automatically. A person still accepts or rejects every row.At least 3 query templates proven, each having produced a contact for at least 3 different founders, plus at least 200 signals across at least 6 founders, plus a measured keep-rate of 25% or better. Reverts if keep-rate falls under 15% for two weeks.
Buyer shortlistA person, onlyA rules-based score, not a model, replaces the operator's manual ranking as the default sort.At least 10 closes with unbroken lineage from query to deal, and the operator's top-quartile picks replying at twice the rate of his bottom quartile across 200+ touches. If human judgment does not beat the bottom quartile, the score does not get automated, it gets deleted.
Outreach draftingModel drafts, person editsThe draft goes straight to you without an operator rewrite.Median operator edit distance under 20% across the last 10 founders' sequences, and no founder rejecting a message at the read-aloud step. Reverts the moment median edit distance goes back above 30%.
Follow-up cadenceCode schedules, you sendTiming, suppression and channel logic run without a hand on them.Cadence unchanged for 10 consecutive founders, bounce rate under 3%, zero spam complaints in 30 days, and the deliverability gate enforced in code rather than as a habit. A single spam complaint reverts it.
Call prepA person assembles itAuto-assembled from the signal, contact and objection tables. The operator approves rather than builds.New objections in under 20% of the last 20 calls, meaning the written library already covers four out of five things that come up, plus at least 10 active objection cards each used 3 times or more.
The live callA person. Always, for nowAsync coaching after the call first. Real-time prompting during the call is last, deliberately.For async: model objection-tagging agreeing with manual tagging at 85% or better on a 30-call holdout. For real-time: that, plus call prep green for 30 straight days, every top-ten objection type carrying 3 recorded responses with results, offline replay against 50+ calls surfacing the right card inside 3 seconds at 80% precision, and 20+ recorded founder-freeze events with the topic logged. It stops permanently if it ever prompts a founder to say something false.
Closing the dealYouNothing. There is no trigger in this row.This one is not on the ladder. You sign the customer, we do not sell as you, and we take nothing on a deal we did not source.

One condition applies to every row and it is not a yield trigger. A step may only be automated against an official platform API inside its own tier terms, licensed job-posting data, or licensed contact data under a data-processing agreement. A step that cannot be automated compliantly is unbuildable, not merely unwise, and it stays in the left-hand column forever.

06

Three intended integrity rules.

The proposed service would never pad the list. If only two buyers cleared the six-point bar, the founder would get two and the drafted $333 credit. The temptation here is obvious: a tier-2 target dressed as a named person would fill the gap and you would probably never notice. Padding once would make every other sentence on this site worthless, including the fee, which is the whole argument for taking us seriously.
Dealone would never send as you or log in as you. Every message would leave your mailbox, in your words, after you have read it out loud. No shared credentials, on any account, ever, including the ones you would happily hand over on day one to save yourself an afternoon. Sending as you would burn your domain on our schedule rather than yours, and an outreach thread where the founder never actually wrote anything falls apart the moment a buyer replies with a real question.
The intended screen would gate payment. A future intake would reject ineligible products before any payment path and record the reason. Products with revenue would route to a different sales motion.

The screen, payment controls, and $199 workshop are not live. If they open under cleared terms, workshop spend would not influence screening or count toward a Deal Sprint.

07

The rule that takes the word AI off this website automatically.

Under the proposed provenance rule, every artifact would carry a data-generated byline such as Produced by: Deal Lead (human) or Drafted by: model, edited by operator, edit distance 34%. The percentage would be measured from the actual rewrite.

The reasoning is simple enough to check. If a person is rewriting more than two words in five, the model is not doing the work. A person is doing the work with a rough first draft in front of him, which is a real and useful thing, but it is not what a reader hears when a page says AI. At that point continuing to say it would be a claim we could not substantiate from our own instrumentation, so the instrumentation removes it instead of waiting for someone to decide to be honest. The tripwire only ever fires in the direction that costs us the more attractive word.

It runs on the same measurement as a build decision, in the opposite direction. Automatic drafting only gets promoted past the operator's rewrite when median edit distance falls under 20%, and it gets demoted again above 30%. One number governs both what we are allowed to claim and what we are allowed to build, and the claim threshold is the loosest of the three. We cannot quietly ship the automation while the copy stays honest, or keep the copy while the automation quietly rots.

If the median operator edit distance across the last ten sequences goes above 40%, the word AI comes off this site within 48 hours.

It cannot fire yet, and pretending otherwise would be its own dishonesty. The rule is defined on a median across the last ten sequences, and Cohort 001 is four seats, so at four sequences there is no window to take a median over. Until then, every artifact still carries its own individual edit-distance byline from the very first sequence, so you can see your own numbers while the fleet-wide rule is still filling up. A tripwire nobody can read is decoration. This one you can read on your own delivery.

See the bar now. Wait for the gates before intake.

The self-check is public, but applications and payments are disabled. The launch page lists the operational, privacy, legal, and payment evidence required before a founder can submit anything.

Review launch status See what each rung costs