Nerdbeak builds and owns agent-operated vertical marketplaces.
Investor memo · June 2026
Most of commerce never got a marketplace. Every niche with real demand and scattered supply stayed unorganized for one reason: running a market takes a team, and most markets cannot pay for one. Agents just made the operator nearly free, and that flips the entire long tail of consumer commerce and services into ownable markets. Nerdbeak builds and owns those markets. Each gets its own brand, seller workflows, buyer experience, and commerce rails. Nerdbeak takes a cut of every sale and uses agents to run the operator work: find a vertical, launch the market, seed supply, activate demand, read the metrics, and decide what to test next. Collectibles is the wedge.
The bet is simple: can agents operate those markets for less than those markets earn?
Each market earns take rate and costs compute. Agents already do real marketplace work: they research verticals, configure launches, create content, draft outreach, find supply, price inventory, and review the work with me. The missing layer is the heartbeat: a continuous loop that runs those jobs, keeps context between them, and decides the next move without me kicking it off. If take rate beats compute, one company can own many markets because the marginal cost of the next market is tokens, not headcount.
ISec.
The problem is operations, not creation.
Launching a marketplace is getting cheap. Keeping one alive is still hard. Marketplaces die when they never reach liquidity, stretch one marketplace into categories that do not belong together, rely on GMV too thin to carry a standalone team, or spend too much on operations before the flywheel starts.
District's $14.7M a16z/Kindred-led round proves the market believes software makes marketplace creation cheap. We agree. But District is the people-in-the-loop version: tools for humans to launch marketplaces. Nerdbeak is the agent-operated version: agents that find, launch, seed, measure, and operate markets. District scales with human operators. Nerdbeak scales with compute.
IISec.
Collectibles is the wedge, not the company.
We started there because the markets are real, fragmented, and operationally broken. The first vertical has done $15K+ in paid volume with almost no ad spend, while content has brought 3,000+ visitors. But the bigger question is not whether people buy collectibles. They do. The question is whether agents can generate enough liquidity that take rate beats compute. That is what the next phase tests.
IIISec.
The operator becomes software.
A human operator has to pick a market, build the site, source supply, make content, run outreach, watch the numbers, and decide what is next. We turn that into an agent loop: find a vertical, launch the market, seed supply, activate demand, read the metrics, decide what to test next. The unit is not a storefront. It is an autonomous market: agents handle the repeatable work, and founder judgment teaches tone, timing, and collector taste.
IVSec.
The pieces are working.
One config stands up a branded marketplace with shops, profiles, payments, and admin. Nine verticals exist in the system today; new ones launch from a config, not a roadmap. Agents research markets, find communities and sellers, draft collector-specific outreach, write search-led content, and test supply discovery against real comps. The operator console is live: every market, its bottleneck, and the next agent action on one screen. We demo it. Today the loop still starts when I press run, and I review for taste before work ships. The next phase closes that gap: a heartbeat that finds, launches, seeds, measures, decides, and repeats on its own.
VSec.
The first vertical is live.
In five months, collectible art has done $15K+ in paid volume, 140+ orders, ~40% monthly repeat buyers, and a 3.6-day median time to sale, on about $535 of total ad spend. Content has brought 3,000+ visitors. We did not chase maximum GMV in one market. We spent the time building the primitives agents need to run many. The honest open risk is liquidity. The next proof is that it repeats.
- $15K+
- paid volume
- 140+
- orders
- ~40%
- monthly repeat
- 3.6-day
- median to sale
- ~$535
- total ad spend
- 3,000+
- visitors
VISec.
We launch ecosystems, not storefronts.
Every launch is an Etsy-grade platform tuned to its niche from day one: shops, branded seller sites with their own domains, drops, auctions, listings, newsletters, orders, and admin. The press prints the whole platform, and it scales with the market it serves. We believe in bringing a nuke to a gunfight: markets that run on Discord threads and spreadsheets get hit with an entire platform.
That matters because a serious seller today juggles Shopify, eBay, Discord, a newsletter, Instagram, a Facebook group, and a spreadsheet. We collapse that stack into three workflow shapes: Studio for handmade makers, Booth for inventory-event sellers, and Catalog for high-volume known-product sellers.
VIISec.
The prize is bottom-up.
The market math is not "collectibles is huge." It is the number of markets we run, times the GMV of each, times our take rate. At a 10% take rate, a $1B outcome means roughly $700M-$1B of GMV flowing through the system. A $10B outcome means roughly $7B-$10B of GMV, around Whatnot scale. And the ceiling is higher than collectibles: if agent-operated markets work for consumer goods, they work for services. The long tail of commerce has never had an owner, because owning it took headcount. That is the prize.
The anchors are real. Small markets at $100K-$500K GMV are proving grounds that only work if compute replaces a team. Markets at $1M-$10M already exist: District's own portfolio includes NikNax at $5M+ in sales and Stacked Golf at roughly $150K per week. Category markets are much larger: eBay did over $1B of U.S. trading-card GMV in one quarter, and Whatnot reported $8B of live sales in 2025. Sports cards were a ~$13B global market in 2023, trading card games were ~$13.3B in 2025, and North American comics and graphic novels were $1.94B in 2024.
Sources: Whatnot 2026 State of Live Selling, eBay Q1 2021, District seed coverage, Spherical Insights, Mordor Intelligence, ICv2.
VIIISec.
Why us.
Ricky Eckhardt has spent 15 years building marketplaces: founding engineer at Fitted Retail ($100M+ GMV), built BookOutdoors (sold to Hipcamp), scaled GoCamp past 100K users to an exit. This is his eighth marketplace. The human in the loop is not ops labor; it is the judgment and collector-specific taste the agents learn from.
IXSec.
The next phase.
The next phase is the rollout: more markets, the always-on heartbeat, live arbitrage, and the push to $10K–$20K monthly GMV across verticals. The next major proof point is $100K monthly GMV.
The opportunity is to back the operating system before the heartbeat is closed and the next milestone is obvious.