Find the signal.
Trace the network.
I'm John. I chase the primary source. The filing, the record, the money trail behind whatever narrative everyone's already accepted. For markets, that work lives at EdenFinTech. This is where I do it for the public record.
AI helps me move faster. The questions are still mine.
Epistemic research
Epistemic research starts with a simple idea: before trusting any data, ask how it was formed in the first place. Markets are full of numbers, narratives and opinions, all of them are built on assumptions, incentives, and sources that can be wrong. Instead of taking information at face value, the focus is on questioning it: separating price from the actual business, distinguishing headlines from evidence, and understanding who benefits from a particular story being told.
In practice, this means thinking about risk first, not last. Cheap assets are ignored without a clear reason for change, and conviction isn’t based on spreadsheets alone but on watching how things actually unfold over time. The goal isn’t to predict perfectly — it’s to stay grounded in reality, avoid unforced errors, and act only when the gap between perception and truth is clear enough to matter.
Most research sells you an answer. Epistemic research sells you a better-formed question, plus the paper trail to defend it.
Case Studies
Every investigation, fully sourced, every link visible. The same approach I use when my own capital is on the line.
AttBid — RMH Holdings Takeover
Maps the mandatory delisting offer for RMH Holdings: the van der Watt brothers' 51% SPV, Atterbury's governance control, and activist minority opposition. Competition Commission cleared unconditionally on 14 April 2026 — deal on track to close, but whether the brothers share in NAV growth remains structurally unresolved.
Read the wiki → Geopolitics · Energy · OngoingStrait of Hormuz — 2026 Crisis
Traces the full sequence of events from US-Israel strikes on Iran through the blockade declaration, ceasefire cycles, and the strategic standoff over who controls the world's most critical oil chokepoint.
Read the wiki → Fintech · Markets · OngoingPayPal — Checkout in Decline?
Maps the competitive network around PayPal's branded checkout: peer processors, Big Tech wallets, BNPL rivals, and the leadership transition that followed Q4 2025's miss.
Read the wiki → South Africa · OngoingSA State Capture & Corruption Network
Maps the state-capture network around the sitting presidency ~ Zuma-era patronage, NPA interference, and corporate enablers.
Read the wiki → Italy · Heritage · DemoAlfa Romeo — A Heritage Network
Traces the organisational, racing, and symbolic network of Alfa Romeo from 1910 to 1987 across people, vehicles, and events.
Read the wiki → South Africa · Private Equity · M&ASA Pest Control — Buy-and-Build
Maps the SA pest control roll-up thesis: 1,000+ fragmented operators, an Act 36 compliance moat, BBBEE procurement advantage, and a first-mover window before global consolidators (Anticimex, Rollins) enter. Identifies platform and bolt-on acquisition targets across Gauteng, KZN, and the Western Cape.
Read the wiki → South Africa · Labour Mobility · Product ResearchWork Abroad Pathway Intelligence
Validates the business case for a signal-based product helping South African professionals identify realistic overseas work routes. Maps five confirmed profession-country wedges (nurses, ICT, engineers, teachers, accountants), automatable official signals (NMC, CSOL, DETE, IRCC), and the trust landscape shaped by endemic fake-job scams.
Read the wiki → South Africa · Franchise · ConsumerXpresso Café — Franchise Model Deep Dive
Maps the unit economics and dual-revenue structure behind SA's first R14 flat-price coffee franchise: establishment fees, 5–7% royalties, vertical integration through Jackass Roastery, and the 500–1,000 customers/day throughput model that makes the numbers work.
Read the wiki → Medicine · Pharma · Clinical EvidenceTirzepatide — T2D Clinical Evidence Vault
Separates clinical T2D evidence from weight-loss marketing across 15 research iterations: SURPASS trial programme HbA1c reductions (1.87–2.58%), insulin displacement data, kidney and cardiovascular outcomes, evidence gaps in advanced CKD and older adults, and real-world patient feedback from Drugs.com, Reddit, and NHS HealthUnlocked.
Read the wiki → South Africa · B2B · Product ResearchRiello UPS — Distributor Catalogue Vault
Maps the Standby Systems Riello UPS catalogue across four topology tiers (offline, line-interactive, online double-conversion, transformer-based three-phase) and six industries (banking, data centres, medical, industrial, telecom, computer rooms). 47+ atomic notes mapping how to size and select the right UPS for a given load.
Read the wiki → South Africa · Safety Tech · Product ResearchVoice-Activated Help — SA Distress App Viability
Assesses commercial, technical, legal, and operational viability of a voice-triggered personal distress app for South Africa. Maps Valor's April 2025 launch and 2023 patent filing, AURA's dispatch API as the response backbone, the Picovoice Porcupine open-source stack, Android/iOS background-audio constraints, POPIA consent architecture, and unit economics at R150–199/month.
Read the wiki → Construction · Global · Evidence MapThe Global Construction Payment Problem
Tests whether subcontractors face a systemic global construction payment problem — mapping delayed payment, retention withholding, pay-when-paid clauses, and contractor insolvency across markets from Australia to South Africa. Verdict: yes, and worsening — with the Carillion and ISG collapses as the canonical cases and statutory adjudication the one measurable counter-measure.
Read the wiki → Quant Finance · Markets · Evidence MapMarkov Trading Models — Do They Actually Work?
Tests whether Markov-based trading models — Markov chains, Hidden Markov Models, regime-switching, Markov Decision Processes, reinforcement learning — produce genuine tradeable edge or just academic backtest artefacts. Verdict across 10 research rounds: a useful regime-detection and risk-management component, not a substantiated standalone profitable system — no surveyed study reaches a strong evidence grade and no credible public live track record exists.
Read the wiki →What a generalist AI misses
I asked Perplexity Pro and SignalTrace the same three-part question about PayPal branded checkout in Q4 2025. Same prompt. Different epistemics.
A 25-line summary.
- Surface citations — Wikipedia, Britannica, aggregator blogs.
- Hedged verdict: “signs of erosion” — no commitment, no frame.
- Market-share stats that conflate different market definitions.
- Timeline stops at end-2025 — misses the Feb 2026 CEO termination.
- No falsifiable forward call. Nothing to test against.
A structured thesis.
- Explicit frame: redirect model = Flash, native wallets = HTML5.
- Hard metrics: transaction margin 2.07% → 0.87%, SMB TPV −$235B, $471B Pay Now TPV.
- Three structural proofs separating cyclical noise from architectural break.
- Current through Feb 3 2026 guidance pull + Mar 2026 CEO succession.
- Falsifiable forward call: Fastlane conversion at May 5 2026 earnings.
Why does a SignalTrace research vault outperform a generalist AI answer? A general-purpose AI optimises for a fast, fluent summary — it aggregates secondary sources, hedges its verdict, and stops at whatever date its index was last refreshed. A SignalTrace vault optimises for epistemics instead. Every claim links to the primary document — the regulatory filing, the court record, the registry entry — so a reader can verify it without trusting the author. Each entity becomes its own node with its own sources, so the research can be queried, extended, and audited long after delivery. The output is not a 25-line summary but a structured thesis: an explicit analytical frame, hard metrics traceable to source, and a falsifiable forward call. Where a generalist AI answers a question and closes it, an epistemic research vault answers the question and opens the next thirty.
A summary answers the question and closes. A wiki answers the question and opens thirty more — every entity (Fastlane, Shop Pay, Stripe Link, Alex Chriss, Enrique Lores) is its own node with its own sources. The corpus deepens every week. You can keep mining.
Same method, twice
EdenFinTech
Ten years running my own capital, published openly. Every trade I recommend is one I've made myself. The research method that runs the portfolio.
Visit edenfintech.com ↗SignalTrace
The same habits, applied to public accountability. What you can't verify, you can't trust. Whether it's a balance sheet or a government contract.
Before you reach out
What does an engagement cost?
Scope-dependent. I price on scope and complexity, not hours. Email to discuss.
How long does a wiki take to build?
A focused entity network — 20 to 40 nodes — takes two to four iterations depending on source complexity.
Can I see the raw sources?
Yes. Every claim in a SignalTrace wiki links to the primary document — the filing, the registry entry, the court record. That's not a courtesy feature. It's the whole point. You don't have to take my word for anything.
Why not just hire a researcher?
Most researchers report what they find. I map why it matters — the network relationships, the belief architecture, the incentive structures underneath the facts. The deliverable isn't a PDF report. It's a queryable knowledge graph you can interrogate yourself long after the engagement ends.
Audit your assumptions
before the market does.
If the research matters, it needs doing properly. Ownership traced, sources verified, the network mapped. Start with a free 30-minute scoping call — I'll help frame the question before any work begins.
Beat fear with information.
Email John →