The intelligence layer arrives first
ResearchKorthos is a market-intelligence site for the machine economy: funding, factory ramps, deployments, and an evidence-graded supply-chain graph for humanoids and robotics — seven months old, run by someone anonymous, updated near-daily. We checked five of its load-bearing claims against primary sources before writing this note; all five held. A field note on what it is, how far it can be trusted, and why its existence is itself a data point for the thesis.
2026-07-26 · Field notes · 6 min read · By Sondre Hegerland KristiansenA tracker for the machine economy
Korthos calls itself "market intelligence for the machine economy." Half of it is a structured-data product: a funding monitor, a verified-events feed, a public-markets tracker, an index binding recent arXiv research to specific commercial robots, and — most interesting from where this brand sits — a supply-chain graph of 200-plus supplier-to-customer relationships in which every edge carries an evidence grade: verified public, widely reported, or inferred. The other half is a publication: near-daily articles and a weekly "State of the Machines" briefing, the strongest of which carry fifteen to twenty primary citations each.
The operator is anonymous. No names, no team page, no legal entity in the terms; the only jurisdictional hint is a reference to UK copyright exceptions, and a founder account posts under initials. The domain was registered on 2025-12-17 — the whole operation is roughly seven months old, though its archive carries records backdated to 2021. Its robots.txt explicitly welcomes AI crawlers, which is a deliberate strategy: this is a site built to be read by machines as much as by people. The publishing volume — twenty-plus articles in a two-week stretch — strongly suggests heavy automation behind the "not a raw scraper" claim in its methodology. None of that is disqualifying. It is context for the only question that matters: can the work be trusted?
Five claims, traced to the source
The rule this journal applies to any potential source: check the citations against what they actually say, and treat uncited claims as red flags until independently verified. So before citing Korthos anywhere, five of its most load-bearing recent claims were traced to primary sources. Humanoid's USD 152 million Series A at a 1.35 billion valuation — confirmed by Reuters, Forbes, and the company's own release, with Korthos's granular details (total raised, the Schaeffler unit commitment, beta timing) matching. Unitree's approval for a USD 619 million Shanghai IPO — confirmed by Caixin, down to the 51.5% humanoid revenue share and 5,500 units shipped. Figure's BotQ ramp — one robot per hour, a 12,000-per-year line — confirmed against Figure's own newsroom, with Korthos correctly framing the 1,000-unit milestone as manufactured units, not deployments. The SKF–Leaderdrive joint venture for humanoid joint components — confirmed by SKF's release. And AGIBOT's 15,000th robot — a claim Korthos published without a citation, which is exactly the pattern the rule exists to catch. Checked independently: it holds, via PR Newswire and The Robot Report.
So: five for five, including the uncited one. Two honest caveats temper the pass. Nearly everything Korthos publishes ultimately traces to company announcements — it relays accurately and structures well, but it does not independently verify, and it faithfully transmits industry hype even while flagging it. And where sources conflict (competing humanoid market-share figures from different vendors' filings), it relays both without reconciling them. The verdict that follows is the sourcing policy this journal will use: Korthos is a competent, disciplined pointer — usable as a curated feed and a map, never as a terminal source. Every claim that matters still gets clicked through to the primary.
The data layer forms before the factories
The Vooma note argued that the goods economy's coordination layer is automating before its physical layer. Korthos is the same pattern one level up: the intelligence layer of the machine economy is forming before the machine economy's factories exist at scale. Someone anonymous, in seven months, assembled an evidence-graded map of who supplies whom in humanoid manufacturing — while the humanoid factories themselves are still counting first thousands. Information about the buildout is compounding faster than the buildout.
That matters to this thesis in two specific ways. First, routing — the model's core decision — runs on exactly this kind of data: comparable, evidence-graded information about capability, supply relationships, and capacity. A supply-chain graph with confidence grades on every edge is a primitive ancestor of a routing input. That such a thing is being built commercially, now, by an outsider, is evidence the data layer of routable production is plausible — not in 2036, but today. Second, Korthos's production-claims taxonomy — separating achieved output from sustained rate from capacity target from marketing — independently converges on the same discipline as this brand's register: claims graded, dated, and held apart from hopes. When two unrelated efforts arrive at the same epistemic hygiene from opposite directions, that is weak but real evidence the hygiene is the right response to this market's noise.
The moat argument cuts both ways
The learning page argues the durable moat is the framework and its data loop, not the robots — P-18 on the register. If that is right, then whoever structures the machine economy's public signals first gets a head start on part of that moat — and right now an anonymous aggregator is moving faster on public-signal structure than any certifier or manufacturer. Korthos has no operating model: nothing on portable blueprints, certification, tiers, or routing. It watches the buildout; this brand proposes an operating system for it. Complementary today. But a tracker with a supply graph is one strategic hire away from a routing opinion, and the honest reading is that Korthos demonstrates how cheaply the observational half of an intelligence moat can be assembled by someone who never touches a factory.
The other discomfort is our own: citing an anonymous source is a risk this journal normally would not take. The mitigation is the policy above — Korthos gets cited as a map, and every load-bearing fact gets verified at the primary before it enters this site's pages. On that basis, one Korthos reference now stands on the execution page, next to the humanoid-race tracker. If their citation discipline slips, the reference comes out — that is what it means for a source to be on probation. This note scores nothing on the register; it is evidence about the information layer, not a test of any dated claim.
Where this came from
Korthos pages read in full, and the primary sources used to verify its claims — vendor-claimed figures are marked as such throughout. Key references:
Korthos — main site · Korthos — methodology · Korthos — supply-chain intelligence graph · Korthos — humanoid supply chain moves briefing · Korthos — production claims taxonomy · Forbes — Humanoid raises USD 152M at 1.35B · Caixin — Unitree STAR Market IPO approval · Figure — ramping F.03 production · SKF — Leaderdrive joint venture release · The Robot Report — AGIBOT's 15,000th robot
Add to this
Corrections, evidence, and disagreement are welcome — this is knowledge in the open. Anyone can read; sign in with GitHub only to post or react, and it appears here instantly.