Workstream comparison + two new angles (Oct 13, 2026)
Verdict: data licensing is the higher-upside, more profitable workstream. Physical goods is the steadier, more controllable one.
| Dimension | Physical goods brokerage | AI data licensing (finder/advisor) |
|---|---|---|
| Capital required | ~$100k tied up in inventory | ~$0 (time + legal template) |
| Net margin | 20–30% on sales, after fees/DOA/freight | ~100% of fees (referral $25–50k; success fee 5–10% of $100k–$2M deals) |
| Ticket size | $500–$25k per item/lot | $25k–$100k+ per close |
| Path to $20k/mo ($240k/yr) | ~$100k/mo sales at 20% net = 1 full inventory turn/mo, flawless ops | ~5–10 closes/yr across referrals + advisor deals |
| Realistic Year 1 | $60–150k profit (learning curve, fees, dead stock) | $0 in a bad quarter to $300k+; variance is the cost of the upside |
| Chad's edge | Ambaum tech ops + Shopify storefront skill | Live Mercor partnership, CPA credibility with seller channels, Seattle buyers |
| Main risk | Price normalization (memory 2027–28), deflating GPUs, operational grind | Market window: labs are hiring in-house teams; hunger is peaking now |
| Controllability | High — your effort sets the pace | Lower — buyers' budgets and legal cycles set the pace |
Recommendation: licensing as the lead workstream (start now, while the window is open and his Mercor proof is fresh), physical goods as the second engine (steady, capital-backed, builds while licensing deals cycle). They share one asset: a reputation with owners of things AI needs.
Fresh market confirmation (Oct 13, 2026, MIT Technology Review + Fast Company): SimpleClosure's Asset Hub has handled 150+ data deals and expects $10M+ returned to startups in 2026; typical small-startup payouts $10–100k (one 2.5-year-old crypto startup netted ~$25k for its codebase + ~$15k pending for workspace data; Cielo24's full Slack/email/Jira archive went for hundreds of thousands). Fast Company reports labs have largely exhausted defunct-company supply and are now buying live-company data at $100k–$2M per company, with micro1 running an "unlimited" $50k referral bonus. Translation: the finder market is hotter than the Oct 7 artifact described, and the growth is in operating companies — Chad's network.
ContentRunner.com — potentially valuable, rights review decides
Facts: Seattle content-writing marketplace, established 2013, matched freelance writers with businesses needing content (Adweek covered its writer "Offerings" launch). The site today returns a server error (checked Oct 13, 2026) — dormant. Chad used to own it and still has access to the files. Why it could be valuable for data licensing: - A content marketplace's archives are writing-workflow data: client briefs → writer drafts → editorial revisions → final copy, at scale, over years. That is the long-horizon work-process data labs pay premiums for, plus a licensable text corpus. - Even its internal ops data (Slack/email/tickets of an operating company) sits in the same category as the Mercor/Ambaum deal. The make-or-break questions (all answerable from the files + a lawyer): 1. Chain of title: did Chad retain the assets on exit, or did they transfer to a buyer/successor entity? 2. Writer and customer terms: did ContentRunner's TOS retain any license to content produced on the platform, or did all rights pass to customers on payment (the usual marketplace pattern)? Briefs and revision histories may be platform-owned even where final articles are not. 3. Scale: number of orders, articles, writers, years of records — size sets the price band. 4. PII/consent posture: writer and customer personal data needs the same de-identification treatment as any dataset. Action: inventory the files (counts, date ranges, systems) and locate the historical TOS/writer agreement. Zero cost, and the answer prices the asset. Do not pitch it to any buyer before the rights review.
Buying old companies / agencies for their data — viable, with one structural warning
Marketplaces: Flippa (5,000+ listings; agencies are a standing category — current examples: a software dev agency asking $725k on $910k revenue/$317k profit; smaller digital agencies at $83–130k asks; micro agencies under $10k), Acquire.com (startups/SaaS), Microns (micro-startups), Empire Flippers and Quiet Light/FE International (brokered, larger). Going businesses trade at ~1.6–3.5x annual profit on Flippa examples — expensive if data is the only prize. The economics that actually work: - Dormant/winding-down companies, not operating agencies. Asset sales, ABC (assignment for benefit of creditors) processes, and shutdown channels price data as a leftover, not a business. SimpleClosure's whole model proves sellers exist at $10–100k data payouts. - Buy the data asset, not the company, where possible — cleaner chain of title, no inherited liabilities. - The structural warning (from Chad's own artifact): agencies are poor data assets because their archives are mostly client-owned work — the same Shopify-partner/client-data problem as his Mercor negotiation. A dev agency's repo history usually belongs to its clients. Target instead: product/SaaS companies whose workflows and data are their own IP, content platforms (the ContentRunner pattern, where a platform TOS may hold rights), and companies in regulated/expert verticals where archives command premiums. - Sellers increasingly know the value (47% of SimpleClosure shutdowns now include a data sale), so the pure-arbitrage window narrows — the durable edge is diligence speed and rights screening, which is Chad's CPA skill set. Sequencing: this is a variant of the licensing leg, not a third business. Start as finder/advisor (no capital), learn what buyers actually pay for which archives, then deploy capital into data-asset purchases only with that pricing knowledge in hand.