The
Trenches
A POSI is a place where somebody must trade without a view. Finding one is slow, expensive work — and the finding itself has almost no way to pay for itself. That is not a side effect of the problem, it is the problem: coverage voids exist because research economics broke. Pump.fun, stripped of the memes, is the most successful machine ever built for paying a creator a permanent royalty on attention with no fund, no mandate and no minimum AUM. So the question is worth asking properly: how much of that machine actually transfers to unearthing, arguing about and monetising structural inefficiency in listed equities — and where does it flatly not?
Why the question isn’t stupid
The POSI list ends with a family called work barriers, and its governing test is Perold’s: you should be paid for labour, not epiphanies. The trouble is that no mechanism pays for the labour unless you already run a fund. The ladder available to someone who finds a pocket today has four rungs and three of them are broken: buy it yourself (capped by your own capital), raise a fund (years, fixed costs, a minimum viable AUM), publish it free (reputation only), or sell research to institutions — which is the business that unbundling destroyed, and the reason archetype 09, the coverage cliff, exists at all.
Pump.fun’s creator-fee tiers are the first widely-used mechanism that routes a share of an asset’s trading volume to the person who created it, permanently, on-chain, with no fund wrapper and no accreditation gate. On the first day of the new tiers it paid out roughly $2.4m. Whatever you think of what is being traded, that specific rung of the ladder is the one research has been missing for twenty years. It is worth understanding precisely before deciding what to do with it.
Read the mechanism, not the vibe
Asked for his answer to the casino critique, the founder’s reply was “have a higher IQ, I guess — look at what’s actually going on, read into it.” Fair enough. Here is the mechanism, in five primitives. Figures below are as stated by the founders in the three interviews (mid-to-late 2025) and in coverage of Project Ascend; treat them as of that date.
The curve
A coin is created in one click and prices itself along a virtual-reserve bonding curve. No order book, no counterparty, no listing committee, no minimum size. Ten seconds from idea to a market with a readable price.
The public program now generalises the quote asset beyond SOL — USDC-paired coins included — which matters more than it sounds: it is the difference between a casino chip and a denomination anyone can account in.
Graduation and locked LP
Cross a threshold and liquidity migrates to their AMM. Roughly $1.6bn of SOL sits in pools sent to a burn address — it cannot be pulled, by anyone, ever.
Newer “boost mode” redeploys the migration SOL as market buying in the first five minutes of a coin’s life — on their own napkin figure, $5–10m a day of injected bid.
Dynamic creator fees
The fee falls as market cap rises: about 0.95% below $300k, roughly 0.5–0.7% to $1m, then 0.2–0.05% above it. The creator’s share goes up to about 0.3% — call it a sixfold increase on what it was.
High friction where there is only hype; near-free where there is real liquidity. It is a deliberate tax on noise and a subsidy to depth.
The social layer
Feed, DMs, live streams, and positions and P&L visible to everyone. Their own phrase for what a post is: an atomic unit of social consensus — an opinion with a coin stapled to it.
The claimed payoff for a nobody: a stream with 20 viewers earning $300 in eight minutes, against roughly a dollar a month for the same audience on Twitch.
The buyback loop
Platform revenue buys the token back. The policy moved from 25% of revenue to 100%; something over $70m spent and more than 5% of supply retired, at roughly $50m of revenue a month.
This is the primitive everyone forgets, and the whole disanalogy lives here. PUMP is a claim on a real fee stream. An idea about a stock is not. Section six comes back to it.
The sign flip
The best idea in the three interviews is an argument about network effects, made to push back on the claim that a launchpad has none. It is worth restating carefully, because it is the one piece of this that maps cleanly onto research.
Negative-EV to share
- Your perpetual futures position. Telling people moves the price against your own fill.
- Your prediction-market position. Telling people shortens your odds. You are strictly worse off.
- Your investment thesis. Publish before you are positioned and you destroy your entry; publish after and it looks like talking your book, because it is.
Every incentive says stay quiet. This is why the good work never gets distributed, and a large part of why coverage voids persist at all.
Positive-EV to share
- A coin you hold and did not create. Telling your friends genuinely does move it.
- A coin you created. Every share pays you twice — the position and the fee stream.
- Anything where your compensation rises with the audience rather than falling with it.
The launchpad’s real contribution is not the coin. It is that it flips the sign on sharing, and it is the only consumer-scale mechanism that has.
Any mechanism that makes it profitable to broadcast the work rather than hoard it attacks the coverage-void family at its root. Whether you can flip that sign without also importing the casino is the rest of this page.
The three jobs
Unearth, discuss, monetise. For each: what the platform actually does, what part of it transfers, which POSI archetypes it touches, and what breaks.
What it does
13.5 million active wallets, something like 40% of all Solana activity, and a live feed plus streaming layer where a large slice of the marginal retail speculator population narrates what it is doing, in real time, in public.
What transfers
Use it as an instrument, not a venue. The original archetype list opens with the customer-investor mismatch: Ocado’s short sellers sat in New York and had never used the product. Its general form is archetype 47 — you have to observe the customer where the customer actually is.
For a whole cohort, that place is now a phone app with a feed. The founder’s own framing — if you want to know where the world goes in twenty years, look at what the kids are doing — is an argument for treating the platform as a sensor on the marginal buyer, the way a fund already pays for card-spend panels.
What breaks
A sensor pointed at twenty-year-old speculators tells you about twenty-year-old speculators. It is a read on flow and taste, never on cash flow.
It is also self-referential in a way a card panel is not: the platform manufactures the behaviour it is measuring. Treat it as one panel among several, and never as the thesis.
What it does
Makes disagreement legible and expensive at the same time. Positions and P&L are public in the feed, and the curve itself is a continuous, capital-weighted, real-time count of who believes the thing and how much they have behind it.
What transfers
Two of the original archetypes — high short interest and the bull-bear battle — are really just proxies for disagreement, and both are noisy and lagged; short interest reports on a fortnight’s delay. A curve is neither. The transferable object is a conviction meter: a live priced measure of crowding you read as a contrarian input rather than stand inside.
Second, and better: the discoverability argument. A few hundred concurrent streamers against a multi-million-person audience means an unknown’s odds of being seen are far better than on Instagram. That is coverage-void logic applied to analysts instead of companies — a thin venue with real demand and few suppliers is exactly where an unknown gets found. It is the strongest argument on this page for a small research venue over yet another newsletter.
What breaks
The mechanism rewards virality; Perold’s filter rewards verifiability. Those are close to orthogonal, and where they conflict the mechanism wins every time.
The most valuable POSIs are the least shareable things in existence — a semiannual Japanese reporter, an amortisation run-off schedule, a docket. A venue that pays for attention will systematically underfund precisely the work that matters most.
What it does
Routes a permanent share of an asset’s trading volume to whoever created it. No fund, no mandate, no minimum AUM, no accreditation gate, no quarterly letter. The missing rung.
What transfers
The fee mechanic — applied to the artifact rather than to a claim on a security. That distinction is doing all of the work, and section seven is entirely about it.
Perold’s third test says the durable edge is one where the return is payment for labour somebody else declined to spend. A creator royalty is, quite literally, a payment for labour that keeps paying. It is the closest thing anyone has built to that test as a business model.
What breaks
Everything that follows from the tradable object being a token whose thesis is a listed company. That is not a design problem to solve, it is a line, and it is drawn in section seven.
And the fee only exists if there is volume. Volume follows attention, attention follows the half-life — which is the next section, and the hardest one.
The half-life problem
Everything above is an argument about mechanism. This is an argument about time, and it is the one that does the most damage. Both axes below are the same axis — note that it is logarithmic, because it has to be.
How long the clock runs
Typical duration of the thing you are being paid to sit through. Log scale, 5 minutes to 7 years.
Show as a table
| Clock | Group | Typical duration |
|---|
The gap is three to four orders of magnitude. No fee mechanic closes it. And it is not a new problem — it is the POSI list’s own archetype 43, dead money duration: the market cannot fund a holding period longer than its attention span. A launchpad does not fix short-termism. It is the most concentrated expression of it ever built.
What does not transfer
Five things, stated plainly, in descending order of how much they matter.
The forcing is on the wrong side
A POSI pays you because somebody else must trade without a view — a mandate, a redemption, a rebalance. On a launchpad, everybody is trading without a view, including you.
You have not found a pocket of structural inefficiency. You have built one and then stood in it. This is the deepest objection on the page and there is no clever way around it.
Pump has revenue. A research coin has none
PUMP is a claim on roughly $50m a month of real platform fees, all of which currently buys the token back. That floor is why the thing is an asset rather than a game.
A coin representing an idea about a stock has no fee stream, no buyback and no floor — only reflexivity. Copy the launchpad without the revenue and you have copied the casino and left the business behind.
Verifiability inverts
Perold’s ideal is a market where everything is perfect except the price. A launchpad is a market where nothing is verifiable and the price is the only thing that exists.
That is the frontier-market failure mode from the POSI filter — inefficient, but not in a way you can act on — rebuilt deliberately, with much better UX.
The half-life
Three to four orders of magnitude, per the chart above. Worth restating because it is structural rather than fixable: fee income requires volume, volume requires attention, and attention on this venue is measured in hours.
Every archetype in family VI is defined by needing more time than that.
Survivorship — in their own data
Their bounty product got roughly 20 million views on launch and was killed within weeks, because attention never converted into usage. Three people, two weeks, shipped, measured, cut.
They are unusually honest about this. Anyone importing the mechanic should import the kill discipline with it, and expect most versions to fail the same way.
The line
Three tiers. This is the part where the interesting version and the legal version diverge, and it is worth being exact rather than vague.
Tokenise the work, never the security
The tradable object is the artifact: a model, a dataset, a translated-filings pipeline, a docket-tracking feed, a piece of written research. A creator royalty accrues on its usage. A public, timestamped, verified track record attaches to the author.
No ticker sits inside the asset, so nothing about it is a claim on anyone’s stock. This is a newsletter with a fee tier and a verified record — unglamorous, and the only version that is unambiguously fine.
Cash-settled contracts on the catalyst, not the equity
Will this name be deleted at the June reconstitution. Will the deal terminate before year-end. Will the readout hit its primary endpoint. Will the rating go through to high yield.
Whether an event contract is a regulated product — and which regulator — varies by venue and country, and the answer has been moving. Get advice for your jurisdiction rather than reasoning from someone else’s.
A coin whose thesis is “buy this listed microcap”
Launching a token that exists to promote a security you are positioned in is scalping — publishing a recommendation you stand to profit from without disclosing the position — and it has been settled US law since SEC v. Capital Gains Research Bureau in 1963.
Coordinating buying in an illiquid name you have tokenised commentary about is market manipulation in essentially every developed market. The token wrapper changes nothing about either. This is the version that would work best and it is the one not to build.
The defensible build
The honest landing zone, and it is a better one than it first sounds. Look again at the POSI page’s fifth exploitability test: you can name the clock — a dated, mechanical, verifiable event that ends the forcing. That is not merely compatible with an event contract; it is the specification of one. Both are selecting for the same property, which is why every archetype in family I resolves on a published date.
So the family of archetypes that most needs a market already has the right instrument, and it isn’t a coin. What the launchpad supplies is not the asset — it is the sign flip and the fee mechanic, and both survive the translation.
- Run the alert rack as the sourcing layer. The twenty-two standing alerts on the POSI page are already an event-detection pipeline. Every hit is a candidate with a date attached.
- Express each candidate as a dated binary. Deletion at the June review; termination by year-end; the readout; the downgrade. If it cannot be written as a binary with a settlement date, it failed test five anyway and should not be in the book.
- Publish the artifact with the position, timestamped. This is the whole sign flip: the thesis is public, the position is public, and both are stamped before the outcome. Nobody has to take your word for the sequence.
- Take the royalty on the artifact, not the underlying. The fee accrues to the work, which keeps the clean tier clean and removes every manipulation incentive at once.
- Make the verified record the reputation asset. Not AUM, not followers. The thing an unknown analyst can accumulate in a thin venue — which is the discoverability argument, working for you instead of against you.
- Import the kill discipline. Run each archetype as a cheap experiment with a stated hypothesis. Kill what does not produce inside two quarters. Pour capital into the two that do.
What this buys you: the sign flip — it becomes positive-EV to publish — without the coin, without the manipulation exposure, and without the three-order-of-magnitude horizon mismatch, because event contracts settle on the catalyst date and a POSI is defined by having one.
Scoreboard
Every POSI family against the three jobs. The verdict is narrower than the enthusiasm on either side would suggest.
| Family | Unearth | Discuss | Monetise | Why |
|---|---|---|---|---|
| I · Mandate mechanics8 archetypes | LOW | LOW | HIGH | The dates are already public — no crowd needed to find them. But a rebalance print or a rating action is a perfect dated binary. Instrument yes, venue no. |
| II · Coverage voids7 archetypes | MED | HIGH | HIGH | The best fit on the board. These exist because nobody is paid to look; a creator royalty and a thin discoverable venue attack the cause directly rather than the symptom. |
| III · Event plumbing9 archetypes | LOW | MED | HIGH | Corporate actions are announced, not discovered. But every one of them is a dated binary, and several are badly priced because nobody bothers. |
| IV · Ownership base9 archetypes | MED | MED | LOW | The feed is a genuine sensor on retail-dominated floats (31). But forced sellers finish on their own schedule, which resists expression as a clean binary. |
| V · Optics6 archetypes | LOW | MED | MED | Nothing social about an amortisation schedule. But the model that un-breaks the reported number is exactly the kind of artifact a royalty can sit on. |
| VI · The clock4 archetypes | LOW | LOW | NONE | The half-life problem in its purest form. These archetypes are defined by needing more time than the mechanism can hold attention for. Actively anti-correlated. |
| VII · Work barriers4 archetypes | MED | HIGH | HIGH | The other best fit, for the same reason as II. Perold’s “paid for labour, not epiphanies” and a creator fee are the same sentence in two vocabularies. |
The conclusion the table forces: the launchpad’s mechanic fits two of seven families — coverage voids and work barriers — and both for an identical reason, which is that their whole problem is that nobody gets paid to do the work. For three families it is merely irrelevant. For family VI it is actively harmful, and for the ownership-base family it flatters you with a sensor while giving you no way to express the trade.
Which is a real answer, and a usable one. Take the fee mechanic and the sign flip. Leave the coin, the curve and the half-life where you found them.