Early Signal Arbitrage

The Dealer Puck

Watched a poker tournament on TV because there was nothing else on. Figured it would be background noise. Instead, the chair pulled closer to the screen. The production was sharp — seeing the hole cards before the players acted made it genuinely compelling television. Poker was about to have a monster resurgence. That much was obvious.

What wasn't obvious to most people: the entire supply chain for home poker was empty.

Went online to buy supplies. Almost nothing on eBay besides playing cards and some antique chips. No dedicated websites selling poker gear. This was early-ish internet, and the gap between what people were about to want and what was available to buy was enormous.

Looked up wholesalers and struck gold — an importer who warehoused poker supplies a few miles from home in Miami. Called, showed up, built a relationship to pick up smaller quantities without minimum orders.

GamblersDepot.com went live. eBay listings went up alongside it. The dealer puck — a small disc that tracks the antes and marks who's dealing — became the breakout product. Nobody was selling them outside of actual casino supply stores in Las Vegas. Thousands sold. The #1 poker chip and equipment seller on eBay for 18 consecutive months. The site carried a full line of cards, chips, tables, and every gambling accessory worth stocking.

Then the importers caught up. They flooded the market and raced each other to the price bottom. The window closed the way these windows always close — not because demand disappeared, but because supply overwhelmed it.

That was 2003. The pattern didn't have a name yet.

It was quiet for a while. Then crowdfunding hit like a lightning bolt and the same instinct fired again.

What Early Signal Arbitrage Actually Is

Every story in this article follows the same three beats. Recognizing that rhythm is what turned a poker supply business into a 23-year pattern.

Detect. Something catches attention — a new technology, a gap in available information, a behavioral shift, a market that doesn't have its basics covered yet. The test is simple: search for it online. If the results are packed with quality information, the signal is already noise. Move on. If the search comes back thin or empty, lean forward.

Build. Start constructing something while still learning about the space. Not after the research is done. During it. The building is the research. A site, a tool, a resource — something lightweight, focused, and live. Speed matters more than polish because the window between "nobody is here yet" and "everyone arrived" is shorter than most people think.

Compound. The individual project might peak, fade, or die entirely. What carries forward is the authority — the relationships made, the knowledge gained, the demonstrated record of having been there first. The asset is disposable. The authority compounds.

This system worked when the empty spaces were entire categories. Poker supplies online. Crowdfunding information. Consumer blockchain affiliates. In 2003, whole markets had no web presence at all.

The game has evolved. The empty spaces aren't categories anymore — they're intersections. Narrow, specific, deeper niches where the big players are structurally too slow or too cautious to go. Nobody is going to out-rank WebMD on "aspirin side effects." But "what happens when you take edibles on Ozempic" with a retired pharmacist willing to put his name on the answer? That's all open.

Two things changed the math completely. The niches got narrower, but the tools got radically cheaper and faster. What used to take weeks and real money and developers now takes an afternoon on a no-code platform and costs almost nothing. The ability to test a signal for the price of a few hours means being wrong carries no real penalty. Build it, test it against reality, learn, keep going or walk away.

The other shift is what makes this framework forward-looking, not just a retrospective: artificial intelligence will absorb and redistribute all the information. Every article, every explanation, every comparison — AI will have it and share it freely. But when someone needs to actually do something — check an interaction, run a calculation, set up a safety protocol — the AI has to send them somewhere. The citation goes to the tool, not the article. Building tools that AI can't replicate but has to reference is the next evolution of capturing early signals.

The Crowdfunding Arc

The Pebble watch — or maybe it was Oculus — on Kickstarter stopped everything. Not the product. The mechanism. The idea that you could presell something to the public instead of selling equity to investors was a completely different way to fund a business.

Backed a few campaigns for five dollars each and started reading everything available about crowdfunding. There was almost nothing written about the space. So the reading turned into watching, the watching turned into writing, and the writing became Crowdfunding-Website-Reviews.com — built to help people figure out which platform matched their campaign. Kickstarter for products. Indiegogo for flexible funding. GoFundMe for personal causes. Nobody had built the comparison layer yet.

Early in the learning curve, the writing led to action. Helping reward-based campaign creators with promotion was a natural extension — these people had great ideas and no audience. Listed crowdfunding promotion services on Fiverr under "other" because no dedicated category existed on the platform. That category is now a permanent fixture in Fiverr's marketplace and a significant business for the site. Created the category before the platform recognized it as one.

GoFundMe had an affiliate program that split their percentage fee — not a flat dollar bounty but an actual share of campaign success. That turned into a meaningful income stream as campaigns funded through the review site's referrals. Top independent affiliate, earning on a model that most people in the space hadn't noticed existed.

Then the review site started attracting a different kind of attention. The private equity group building IPO Village found the site and brought access to PR and industry tradeshows — eventually leading to a Managing Director role. Len Zacks found it. The person who would become a co-founder of VestLo found it. CrowdfundInsider — the space's key publication — became a strong relationship. Every meaningful connection from that era traces back to the same content, written when almost nobody else was paying attention.

The interest had already shifted from reward-based crowdfunding to equity crowdfunding. A background in finance — Series 3 license, time at REFCO in the mid-90s — made the equity side feel like familiar territory approached from a new angle. Writing and consulting fed into early lobbying efforts for the 2012 JOBS Act, which legalized equity crowdfunding but didn't finalize rules until 2015. Three years of pushing for something before the regulations caught up to the vision.

That work led to serving as VP & CMO of Zacks Crowd — one of the first FINRA-registered national equity crowdfunding portals in the US. And then co-founding VestLo — the first registered equity crowdfunding portal under the Illinois state crowdfunding law. Three companies listed. Investments came in. The deals didn't close. But the credential was real and the pattern was proven again: detect the signal early, build the asset, and watch the authority compound even when the individual project doesn't survive.

The review site is gone now — a domain squatter holds a copy of the old content. VestLo didn't become a lasting business. The Fiverr category thrives without any connection back to who created it. But the relationships, the industry knowledge, the demonstrated ability to arrive years before everyone else — those carried forward into everything that followed.

The App Store Filter

Crypto was more than a hobby — the same early detection instinct worked well as an investor, spotting signals in that space before they matured. But building content sites around crypto projects was a different question until a simple thesis emerged: any consumer blockchain application that figured out how to get listed and survive in the Apple App Store had cleared a legitimacy bar that 99% of the crypto space couldn't touch.

Three apps passed that filter. Sweatcoin, Coin App powered by XYO, and Upland. All three still operate today. Sweatcoin had the monster growth numbers and a rich affiliate program. More importantly, nobody with any real skill was playing in the affiliate space around it yet.

Sweatcoin's rules allowed owning Sweatcoin.Club as an independent affiliate site — no employment, no ownership stake in the company, just the domain and the content. Same playbook: build the authority resource, capture the traffic before the competition shows up, let the affiliate revenue compound while the window is open.

Top 5 global affiliate. Peaked about two years ago and the growth has slowed — seven years in, the competition finally caught up the way it always does. Still earns, just at a fraction of the peak. Same cycle as the dealer puck, stretched over a longer timeline. The window opened, paid well for years, and gradually compressed.

The Scam That Got Personal

Most of the signals in this portfolio came from screens — search results, market data, trending topics. ShieldWord started with people.

A friend's father was targeted by a scam — the full story is on the site's about page. A sister got a call claiming her husband was in trouble. Not AI-generated, but the caller had real personal information. Digital breadcrumbs turned into a weapon against someone's family.

Then the wider picture came into focus. Watching parents and their friends age into a world increasingly built to exploit them. The information older adults post on social media without thinking twice. The reality that people in their generation — and honestly, ours too — have no real defense against new technology being used for harm. AI voice cloning didn't create the problem of impersonation fraud. It removed the last barrier. When a scammer can call your mother in your voice and say they need money, "just hang up and call back" isn't a plan.

The signal wasn't that AI voice scams were trending. The signal was that the gap between scam sophistication and victim awareness was widening every single day, and nobody was building anything for the people who actually have to do something about it — the adult children. AARP writes to the senior. The FBI publishes incident reports. The National Cybersecurity Alliance writes awareness content. Nobody writes the "visiting Mom this weekend and need to lock her phone down in 45 minutes" guide.

ShieldWord is almost absurdly simple. A family picks a code word. When an urgent call comes in, you ask for the word. That's it. The tool on the site walks families through setting it up together. Honestly, the tool is probably unneeded — anyone can agree on a word without a website. But the tool isn't really the point. The conversation is the point. And the code word makes the older parent a participant in their own protection, not a child being monitored by their kids.

The family code word concept is now recommended by the National Cybersecurity Alliance, the FTC, CBS, and Clark Howard. A peer-reviewed study from Abertay University validated the core thesis. None of them cite ShieldWord. The idea is proven. The attribution hasn't caught up yet.

That's a familiar position.

Growth and Gap

Older adults are using cannabis. Medical cards, dispensary purchases, edibles, tinctures — millions of people over 50 consuming legally and regularly. Most of them are also on multiple prescriptions. Almost none of them mention the cannabis to their doctor.

Not because they're hiding something dangerous. Because the stigma persists. A 68-year-old with a medical marijuana card and scripts for blood pressure and cholesterol medication isn't going to volunteer that information at a ten-minute checkup. Schedule 3 reclassification should eventually open the research pipeline, but right now there is almost nothing legitimate written about what happens when cannabis meets common medications. A massive growth area with almost no credible information serving it.

The drug interaction space belongs to WebMD and Drugs.com. They own traditional prescription-to-prescription lookups and they'll be slow to move on cannabis and supplements — too much institutional caution, too much legal gray area, too little incentive to cover a space where the science is still emerging. That slowness is the gap.

InteractSafe was built to open the conversation that nobody else will start. The clinical credibility comes from one person: Sanford A. Orloff, RPh, retired after 40-plus years of clinical pharmacy experience. He uses medical marijuana himself. He's the gray-haired pharmacist people wish they could actually talk to at the counter but don't. His credential anchors every page on the site.

The GLP-1 medication class — semaglutide, tirzepatide, the drugs behind Ozempic and Mounjaro — became the primary focus because growth and gap converged more dramatically there than anywhere else in healthcare. Roughly 10 million Americans on GLP-1 therapy, projected to reach 25 million by 2030. Patents expiring, generics flooding global markets, an oral pill changing the entire absorption equation, and one of the few credible free interaction checkers voluntarily retiring — all happening simultaneously. Millions of new, less medically supervised users searching for interaction information while the big platforms still had almost nothing on cannabis or supplement combinations with these drugs.

The architecture was built for a search environment that doesn't fully exist yet. The main InteractSafe tool runs on OnSpace — Google indexes it fine. But AI crawlers need something different. Research from building other projects in the portfolio led to a strategy of static HTML pages hosted on Cloudflare subdomains — front doors built specifically for how AI systems find and cite information, not just how Google ranks pages today. Each subdomain is a crawlable entry point designed for the next generation of search.

Validation came fast. Google Search Console data confirmed the GLP-1 + cannabis query cluster as the top opportunity. An organic post in a GLP-1 subreddit linked to the cannabis interaction page mid-discussion — a real user sharing a resource with other patients, not a planted promotion. The content is being found by exactly the audience it was built for.

The big interaction checkers will eventually cover this space. They always do. The question is whether InteractSafe will have established enough authority and enough tool-based utility by then that the citation sticks — that when AI systems answer a question about cannabis and semaglutide, they send people to the pharmacist-reviewed tool that was there first.

The Happy Accident

Mass white-collar layoffs were the topic of a late-night conversation. AI displacing mid-level knowledge workers wasn't a prediction anymore — it was starting to happen. The question was what tools people would need when the wave hit full scale.

Severance benchmarking was the obvious gap. No usable public data exists for someone who just got handed a package and has 21 days to decide whether to sign. No way to know if the offer is fair for your industry, your level, your company size. The concept went from conversation to a live anonymous benchmarking tool in days. Fastest build in the portfolio.

Then friends saw the site. Every single one immediately referenced the Apple TV show Severance. Should have Googled the exact term before registering the domain.

Rather than fight the association, the site leaned all the way into the show's aesthetic. The Lumon orientation film treatment. The visual language. The tone. Season 3 is coming and the show's audience is massive and obsessive.

Underneath the theme, the tool actually works. A serious issue wearing a fun costume. If the fan community finds it first, the traffic lands on something with real utility. If the layoff wave hits at scale, the tool can be cleaned up and relaunched without the show references. A real severance benchmarking calculator doesn't care what font it's wearing.

The System, 23 Years Later

Laid out chronologically, the projects tell a clear story:

Every project on this list either peaked and compressed, got abandoned, or is still in its early window. None of them became billion-dollar companies. That was never the goal.

The tools have gotten radically faster and cheaper. What cost months and real money in 2003 now takes an afternoon on a no-code platform for almost nothing besides time. A signal can be tested against reality before most people have finished their market research deck. Being wrong costs an afternoon. Being right — and being first — compounds for years.

What hasn't changed is the instinct. Something catches attention. A shift in technology, a gap in available information, a problem that nobody credible is solving yet. The lean-forward moment. Then building starts — not after the learning is complete but during it. If the signal was wrong or the timing was late, drop it and move on. No bosses to convince. No board to update. No investors to disappoint.

Twenty-three years of interesting spaces, no bosses, and a good life. The early signals keep coming. The building hasn't stopped.