This is a screening record, not an industry report. Each of the ten below ran the same six questions: who signs the cheque, why now, what you sell first, where the ceiling is, what compounds, and what result would make you stop. The other 754 are not rejects — they are where these ten came from, they stay published, and you can check our judgement against them.
SourceTraced to the original thread, continuously updatedInventory764 needs, every one publishedThis issue10 businesses · 2 fundable · 5 cash flow · 3 not a company
Summary
Three conclusions. Each of them changes a build decision, which is why they come before the ten.
01
Discussion volume is not revenue. The three most-discussed needs in the inventory — directory submission, blocking AI content, a calendar that bends to flow — are the three hardest to monetise of the ten. Heat measures how much a complaint hurts; it does not say whether anyone will pay to stop it.
02
Six of them are already being paid for, step by step. 01, 02, 03, 04, 05 and 07 — the buyer is already spending today: contact data, model bills, compliance audits, quote gaps, character production, sales collateral. These are not new customers waiting to be convinced; they already pay for the adjacent step, so the product replaces a line item instead of creating one.
03
Only two have a clear window. One is a competitor shutting down — GummySearch stopped new signups in November 2025 and shuts down on 1 December 2026; the other is tokens moving from R&D spend to cost of goods, a turn happening in the next year or two. The other eight are not wrong, they are untimed: doable this year, still doable in five.
The portfolio on one page
Ceilings are bottom-up: reachable accounts × monthly price × 12, with the account count labelled wherever it is an assumption. Not TAM, and not a valuation.
#
Business
Buyer
Price
Ceiling
Call
01
A daily lead list from public comment threads
Outbound shops, GTM consultants
$99/mo
$4.8M
FUND
02
Flag the AI customers you lose money on
SaaS teams already shipping AI
$49/mo
$8.0M
FUND · CATCH
03
SOC 2 in six weeks, fixed price
1–3 person company closing a first enterprise deal
$4,900 flat
$0.3M
CASH
04
Defend a quote instead of discounting it
Independent developers, small studios
$29/mo
$12M
CASH
05
Keep multi-persona canon consistent
Streamers running more than one character
$12/mo
$2.9M
CASH · SMALL
06
Hand over a one-person business
Solo founders whose clients depend on them
$19/mo
$1.2M
CASH · HARD SALE
07
White-label audits studios resell
Web and WordPress studios
$49/mo
$6.0M
CASH
08
Submit to 80 live directories in 48 hours
Founders in launch week
$149 once
$0.3M
QUICK CASH
09
Move the calendar around flow
Remote workers, personally
$8/mo
$0.6M
NOT A COMPANY
10
Block AI content where you actually read
Ordinary readers
$3 once
$0.09M
NOT A COMPANY
764 needs, one square eachthe rust squares are the ten in this issue
754 still in the inventory, carried to the next issue10 taken into this memo
Six questions, before anyone spends
The same six, in the same order, on every memo below. If you read only two, read the first and the last: who pays, and what would make you quit.
#
Question
What a usable answer looks like
01
Who signs the cheque
Not "who has this problem" but who is allowed to spend money on it, and which line it comes out of. Replacing an existing line is a sale; creating a new line is a campaign.
02
Why now
The thing that changed in the last eighteen months: a price, an API, a shutdown, a rule, a model getting cheap. If nothing changed, the need has sat there for a decade and nobody bought.
03
The wedge, then the expansion
The smallest thing you can charge for this month, and what it opens. A wedge is not a small MVP — it is a complete answer to one narrow question.
04
The ceiling, bottom-up
Reachable accounts × price × 12. Not TAM. Every memo prints the arithmetic so you can argue with the input instead of the conclusion.
05
What compounds
In twelve months, what do you own that a competitor cannot copy in a weekend? Data, distribution, trust, switching cost. If the answer is "a nicer interface", you have a feature.
06
The falsification test
A specific spend, a specific window, a specific result that means stop — written down before you start, because after three months of work every founder can find a reason to continue.
Ten memos
Ordered by what each should become, not by how interesting the problem is: venture-backable first, businesses that pay one person in the middle, tools and services last. All ten are real; they differ only in which form fits — and each memo ends by naming that form, and the result that would make us stop.
People spend two hours a day reading comment sections by hand to decide who is worth messaging. That is a task a daily list can replace.
Evidence: a small SaaS operator describes exactly this loop — hundreds of relevant comments a day, each one judged and answered by hand.
Who signs the cheque
Independent go-to-market consultants and 2–10 person outbound shops. Contact data is already a monthly line item for them — seat-based databases, email verification, sending tools — so you replace an existing cost rather than opening a new one. The buyer already earns money from this work.
Why now
The category leader in social-comment prospecting stopped taking new signups in November 2025 and shuts down on 1 December 2026 after failing to license platform data — the gap is real, and so is the way it died. At the same time the cost of judging "does this comment contain buying intent" fell from minutes of human attention to well under a second.
Wedge, then expansion
Wedge: five communities, twenty human-checked leads a day, $99 a month. Expansion: from finding the person, to writing the first line to them, to charging per booked call. At that point you sell pipeline, and pipeline is priced ten times better than data.
Ceiling, bottom-up
Roughly 20,000 small outbound shops and independent consultants × $199 a month ≈ $48M addressable; the 2,000 a small team could serve and support ≈ $4.8M ARR.
What compounds
Intent labels — which sentence actually turned into a paid conversation. Only someone who ran the outreach owns that, and no model vendor has it.
Falsification test
200 comments, human-labelled against model-labelled: precision under 70% or recall under 50%, stop. The harder test — if you cannot get a commercial data licence and stand up a second platform, stop. That is the exact combination that killed the incumbent.
VERDICTThe cleanest of the two — the only one I would fund without a caveat. The risk is not demand, it is data access — the first line of the plan should be a licensable source and a second platform, not a landing page.
WHAT WE WOULD VALIDATE NEXTA licensable data source that survives a platform terms change, and a second platform already live. Without both, the incumbent's failure becomes our failure.
Incumbent: GummySearch — stopped new signups Nov 2025, shuts down 1 Dec 2026.
Once an AI feature ships, tokens become a variable cost. Most small teams know the total bill and cannot say which customer caused it.
Evidence: small SaaS operators discussing how to price AI features, converging on the same observation — heavy users quietly become the most expensive accounts.
Who signs the cheque
SaaS teams that already ship an AI feature and already pay a model provider by usage. The budget line is cost of goods sold: the money is being spent today, the only open question is on whom.
Why now
Through 2026 AI features moved from demo to production, so tokens stopped being R&D spend and became a margin problem. Usage-billing infrastructure matured for companies with a finance team; a five-to-fifty person product is still reconciling a provider invoice against payment records in a spreadsheet.
Wedge, then expansion
Wedge: not billing — an alarm. Connect payments and the model invoice, answer one question: which five customers were gross-margin negative last month. $49 a month. Expansion: from "who is losing money" to "what should you charge them". Pricing advice is the product; metering is the plumbing.
Ceiling, bottom-up
Assume 20,000 small AI products — an assumption, not a census. At $49–199 a month that is $12–48M addressable; the 5,000 accounts a small team could serve at $149 ≈ $8M ARR. Billing platforms will push down from above, because their marginal cost is lower than yours.
What compounds
A cost-to-revenue mapping per product type, and a cross-product gross-margin benchmark. The benchmark is the asset — nobody hands over real numbers until you already hold everyone else's.
Falsification test
Ask ten teams with an AI feature what each customer costs them last month. Eight will not know — and if most of those eight also say it does not matter yet, stop. They care about pricing, not cost, and you would be selling the wrong noun.
VERDICTFundable, but only if you describe it as a pricing decision rather than a metering tool. Metering is a feature; pricing is money. The catch: the honest version needs a year of invoice data before the advice is worth paying for.
WHAT WE WOULD VALIDATE NEXTFive teams willing to hand over invoice data, and three of them changing a published price because of what the report showed.
No direct incumbent: Orb, Metronome and Lago meter usage; none attributes gross margin per customer.
Enterprise procurement turned a security report from a funding milestone into a prerequisite for a five-person company. The tooling got cheap; the sequence did not.
Evidence: a founder days from launch asking what the requirement even means, and whether anyone has got through it under $10,000 — which means he has already accepted that he must spend.
Who signs the cheque
A one-to-three person product about to sign its first enterprise customer. The budget line is compliance, or the cost of closing one deal. Price anchors are public: platforms run $6,000–15,000 a year, an auditor $15,000–35,000.
Why now
Tooling became self-serve and its price fell below $10,000 a year. The remaining gap is not software — it is ordering: what to fix first, which auditor to book, what the observation window means. That gap is a person, and a person is what you sell.
Wedge, then expansion
Wedge: $4,900 fixed, six weeks, a Type I report plus an evidence pipeline the client can keep running. Four clients a month is $20k of revenue with no infrastructure. Expansion: the same customer returns for Type II, for a second framework, and for the questionnaire attached to their second enterprise deal.
Ceiling, bottom-up
48 engagements a year × $4,900 ≈ $235k of revenue for two people, against a service pool of $15–25M. This is a people business: hire three consultants and gross margin falls to about 40%.
What compounds
Auditor relationships and a control library that has survived real audits. In year one it is delivery; in year three it is a product you can sell without you.
Falsification test
Of the first ten enquiries, how many will prepay a $2,000 deposit? Fewer than three means they want free advice. Second signal: if every client needs custom work beyond the six-week scope, the fixed price is a lie.
VERDICTA cash-flow business. Do not raise for it. Its real value is that within a year you own the trust of a few dozen B2B founders at the exact moment they start spending on infrastructure.
WHAT WE WOULD VALIDATE NEXTThree of the first ten enquiries prepaying a deposit, and one audit firm holding a six-week slot.
Incumbent: Vanta, Drata, Secureframe — $6k–15k a year, software only, no sequencing or hand-holding.
AI made delivery cost invisible from the outside, so clients negotiate against the assumption that work got cheap. Quoting stopped being arithmetic and became an argument.
Evidence: developers describing unpaid proof-of-concept work that still gets declined, and quotes dismissed out of hand as too expensive.
Who signs the cheque
Independent developers and small studios quoting small-business projects. They are not buying a tool — they are buying the ability to hold a price, and the budget line is the gap between the quote and the counter-offer.
Why now
Delivery cost is no longer legible to the buyer, so credibility in a quote no longer comes from hours logged. It comes from scope boundaries and change terms — which, unlike judgement, can be templated and sold.
Wedge, then expansion
Wedge: not an estimator, since every estimator is guessing. A quote-defence pack: scope boundaries, a change-order template, three sentences for the "too expensive" reply, and one PDF you can send. $29 a month. Expansion: a deal room — proposals with open tracking, and a change order the client signs without a phone call.
Ceiling, bottom-up
Assume 3M independent developers who quote clients — an assumption, not a census. At a 1% paid rate that is $10M a year at $29; at 2% it is $21M. The realistic line for a content-led business is $12M. Acquisition would be editorial, so it ends up a content company that also sells a product.
What compounds
Real closing prices: every quote sent, every counter-offer, every number signed. That benchmark has no public version, and it is worth more than the templates.
Falsification test
Ask twenty freelancers to upload their last quote in exchange for a benchmark. Fewer than five will — the rest want to complain, not to change process. Also watch whether anyone opens the PDF a second time; a tool used once is a template pack, not a subscription.
VERDICTGood cash flow, ceiling set by your willingness to ask for data. Without the data it is another template shop with a $29 price tag and no reason to renew.
WHAT WE WOULD VALIDATE NEXTFive of twenty freelancers uploading a real quote, and the PDF opened a second time by a third of them.
Adjacent: Bonsai and Fiverr Workspace — proposals and contracts, no price defence.
Opening a second persona now costs tens of dollars instead of thousands, so the number of characters per streamer is rising. What did not scale is memory.
Evidence: a streamer asked how she keeps several original characters straight on air, answering with a joke because there is no system.
Who signs the cheque
Streamers performing as more than one original character. Avatars, voice packs and streaming software are all paid — tooling is their cost of production, not entertainment, which is why this audience converts better than its size suggests.
Why now
Voice cloning and image generation collapsed the cost of starting a persona, and the count went up. Continuity — what this character knows, how she talks, what happened last week — stayed manual.
Wedge, then expansion
Wedge: switching is already solved by the streaming tools themselves. Do the unsolved part: a one-card brief thirty seconds before going live, listing what this persona knows, what she must not know, and how she feels about the other one. $12 a month. Expansion: a canon timeline the audience can browse between streams.
Ceiling, bottom-up
Assume 20,000 active multi-persona streamers — an assumption, not a census. At $12 a month that is $2.9M addressable; realistically a few hundred thousand. This employs two people and sells at a low multiple.
What compounds
Persona assets and audience-built canon. There is a mild network effect when viewers contribute lore — not enough to support a venture story.
Falsification test
Ask twenty multi-persona streamers how they track continuity today. If fifteen say "I just remember", stop. If most show you a notes page or a spreadsheet, you have a wedge.
VERDICTCash flow, do not raise. Done well it is a $500k–1M business with a devoted audience; done badly it is a very good open-source project.
WHAT WE WOULD VALIDATE NEXTFifteen of twenty streamers already tracking continuity in notes or a sheet, and a paying cohort that returns weekly.
Incumbent: EchoDeck, Hyper Online, VTube Studio — persona switching solved, canon not.
Passwords can already be handed over on a timer. The operating business cannot: billing portal, client list, deployment notes, and the letter that tells clients what happens next.
Evidence: a founder who was hospitalised, lost the business, and could not get the operational detail to his second-in-command without upsetting clients.
Who signs the cheque
A one-person company whose clients depend on it being reachable. The problem is that there is no budget line — this is insurance, not software, and creating a new line is the expensive kind of sale.
Why now
Nothing changed, and that cuts both ways: it is the most durable need in the inventory and the least urgent. No event makes someone buy this week.
Wedge, then expansion
Wedge: not another password vault — three already sell that, one since 2008. Build the operational handover: billing portal, client list, deploy notes, client letter. $19 a month or $199 a year. Expansion: the same pack sold to two-to-five person studios as a continuity policy.
Ceiling, bottom-up
Assume 5,000 one-person companies — an assumption, not a census. At $5 a month that is $300k; at $19 it is $1.1M. Three incumbents selling tells you the market is real, and that three sellers fill it.
What compounds
Trust, and the accountant and lawyer channel. Nobody finds this through an ad; they find it through whoever already handles their affairs.
Falsification test
Fifty solo founders: how many will pay $199 now? Fewer than two means this is a content product — an article, a checklist, a paid template — and should be sold as one.
VERDICTDo not raise. If you build it, attach it to insurance or accounting distribution rather than building traffic, because the customer will never search for this on a Tuesday.
WHAT WE WOULD VALIDATE NEXTTwo of fifty solo founders paying $199 today, and one channel — accountant or insurer — willing to resell it.
Incumbent: deadmansswitch.net (since 2008), Killswitch, Vaulternal — all hand over credentials, none hands over the business.
Every individual check is already free. What a studio buys is the packaged document it can put its own logo on and send before the first call.
Evidence: a studio owner describing the same opening ritual with every new client — speed test, two SEO checkers, headers by hand, then translating all of it into plain words.
Who signs the cheque
Small web and WordPress studios, at the moment of onboarding a client. They are not buying an audit — they are buying an opening line that makes them look expensive and creates the next piece of work. Budget line: sales material.
Why now
The measurement is commoditised and the explanation is nearly free, so the value moved entirely to packaging and white-labelling. That shift is what lets a small product in.
Wedge, then expansion
Wedge: enter a domain, get a branded PDF with priorities and a suggested retainer, under the studio's logo. $49 a month, tiered by client count. Expansion: a monthly re-check the studio resells at ten times the price — their revenue, not yours, which is exactly why they keep paying.
Ceiling, bottom-up
Assume 350,000 small web and WordPress studios — an assumption, not a census. 1% paying $49–199 a month is $2–8M; the reachable slice is $6M. Low ceiling, but concentrated acquisition: they gather in a handful of communities.
What compounds
Templates, and a statistic nobody else has: which findings most often precede a lost client. Weak, but enough to make a second entrant start over.
Falsification test
Send fifteen studios a real report for one of their clients and ask whether they would put their logo on it and send it. Fewer than five means the report is not good enough to be a sales document, and the thesis fails.
VERDICTCash flow. Do not position it as an SEO tool — that market is free and crowded. You are selling a deliverable with someone else's brand on it.
WHAT WE WOULD VALIDATE NEXTFive of fifteen studios putting their own logo on a sample report and sending it to a client.
Incumbent: white-label reporting inside Ahrefs and Semrush — data without the branded document.
A launch comes with one week of budget and never comes back. This is a transaction, not a customer relationship — price it like one.
Evidence: a thread where every founder is choosing between doing it by hand, paying a service, or hiring someone; nobody is asking whether it works.
Who signs the cheque
A founder who launched in the last week and still has launch budget. The decision happens inside about 48 hours.
Why now — in reverse
AI-generated directory sites are multiplying, which lowers the value of each submission. The window is closing, not opening. That makes it a good way to earn this quarter and a bad thing to build a company on; most people reading demand data get this backwards.
Wedge, then expansion
Wedge: $149 flat, 80 directories that are still alive, delivered in 48 hours, plus a list of which ones actually sent traffic. Expansion: the list is the product — sell "which directories are worth it" as a subscription to every future launcher.
Ceiling, bottom-up
Assume 2,000 paid submissions a year — an assumption, not a census — × $149 with no repeat purchase: call it $300k a year for a well-run one-person operation, decaying as directories lose value.
What compounds
Only the quality score — which submissions converted into signups. The submissions themselves are labour.
Falsification test
Submit thirty directories for your own product and track signups for thirty days. If the middle of those directories sends fewer than five, you are selling a feeling — sell the honest list instead.
VERDICTQuick cash, not a company. $20k a month is achievable; a second year is not, unless you convert it into the directory-quality index.
WHAT WE WOULD VALIDATE NEXTThe middle of those directories still sending five signups after thirty days, and someone paying for the list of which directories are alive.
No named incumbent worth citing: submissions are labour, and the list of directories still alive is the missing product.
Tools that schedule already exist, and they are free. What is missing is the sensor: nobody knows whether you are actually mid-flow.
Evidence: a remote worker torn between a strong work session and a scheduled 45-minute break, who keeps choosing the work and skipping the break.
Who signs the cheque
A remote worker, personally, at maybe $8 a month. Meanwhile the main calendar assistants already defer habits and focus blocks when a conflict lands — documented in their own help pages, using almost this exact example, free on the major calendars. Your buyer has a free substitute for most of the behaviour.
Why now
Local activity signals — terminal, editor, browser — are finally good enough to infer focus at close to zero cost. The missing piece was never the scheduling tool; it was the sensor.
Wedge, then expansion
Wedge: a menu-bar utility. Twenty-five minutes of continuous input pushes the next block back thirty minutes and leaves a note in the calendar. Two weeks of work. Expansion: none that reaches a company — the honest expansion is a licensing deal with a calendar vendor.
Ceiling, bottom-up
100,000 paying users at $8 a month ≈ $9.6M, an optimistic pool; the realistic number is a few hundred thousand before the free incumbent is counted.
What compounds
Almost nothing. This is a function of the operating system and the calendar app, and both will ship it.
Falsification test
Get 100 installs. If fewer than fifteen still run it two weeks later, stop — sensor products do not retain on novelty, and this one has no reason to be opened on purpose.
VERDICTDo not build a company. Build an open-source utility, or take the prototype to a calendar vendor and sell the feature. It is ranked here by business value, not by how interesting the problem is.
WHAT WE WOULD VALIDATE NEXTNot a company, so the diligence is one question: will a calendar vendor license it? If none will, there is no deal.
Incumbent: Reclaim.ai (defers habits on conflict, free tier), Motion, Sunsama — they schedule, they have no flow sensor.
The loudest and least payable need in the library. Three extensions already sell it, one of them at thirty thousand users.
Evidence: users asking for a single one-click control that removes AI-written text from every feed and page they open.
Who signs the cheque
An ordinary internet user, once, for a few dollars. Highest emotional heat in the inventory, lowest willingness to pay — and those two facts are usually the same fact.
Why now — in reverse
Platforms are adding their own AI labelling. Every month that passes, third-party detection is worth less, because you are doing work the platform is about to do with better data. Being early does not help when your competitor is the platform.
Wedge, then expansion
Wedge: if you insist, shrink it — block AI summaries on the ten sites you actually read, $3 once, no subscription. Expansion: none. The moment it works well, the browser ships it.
Ceiling, bottom-up
Extension economics: $3 × 30,000 lifetime installs ≈ $90k, before churn and store fees. No version of this reaches $1M without a different product behind it.
What compounds
Nothing durable. Detection heuristics break weekly, and the users who care most are the least likely to pay.
Falsification test
Ship it with the $3 price on day one. If installs convert below 1%, you have your answer — people want the outcome and will not fund it. Run this before writing the second feature, not after.
VERDICTNot a company. Open-source it, take donations, or use it as the free top of a funnel for something people buy. The lesson is worth more than the product: heat is not demand, and a complaint about the whole internet has no buyer.
WHAT WE WOULD VALIDATE NEXTNot investable. The only test worth running: do installs convert above 1% at $3 once the platforms label AI content themselves.
Incumbent: AI Content Shield (30,000+ users), AI Blocker, BrowserBlock AI.
No cheque has been written against any of these ten. The calls — FUND, CASH, NOT A COMPANY — describe what each should become: a company, a cash-flow business, or a one-off service. Not what we would pay for it. That distinction is the body of the memo, not a footnote.
The other 754 are not rejects
A need that is real and cannot be sold is still a real need. The only difference between them and these ten is whether someone is already paying for it. They stay published and carry into the next issue — the ten are ten because the other 754 were read first.
Every number is an assumption, and it says so
Each ceiling is reachable accounts × monthly price × 12, with the account count stated or labelled as an assumption. Several are probably wrong by a factor of two. They are printed anyway, so the disagreement happens at the input instead of at the conclusion.
No receipts, and a source that is better at pain than at budgets
Nobody in the inventory posted an invoice: what the 764 contain is hours lost, quotes rejected, bills nobody understood, clients who left. That proves the pain; it does not prove anyone will pay, which is why every memo ends with the conditions that would produce that evidence. About seven in ten of what the library collects comes from developer sources — Hacker News, DEV.to and GitHub — so it is a strong instrument for what builders complain about and a weak one for budgets. No heat score appears on any memo: that number is internal to the pipeline and not comparable across days.
Snapshot, not a standing claim. The GummySearch shutdown lands on 1 December; a directory index decays monthly; platform labelling ships on the platform's own schedule. Re-read the dated claims before acting on them, and treat every ceiling as a hypothesis you can bet against.