Contact map
Stage 0This is not a sales list. It's a learning map: which domains and which people to learn from in order to test the thesis and find where it breaks. It sets the targets for Stage 0, step 2 — the 10–20 expert calls.
“The goal is not to pitch everyone. The goal is to learn what would make the thesis wrong or stronger.”— the thesis, Stage 0 plan
The outreach playbook
Method before names: eight principles, then the first message, the sequencing, the prep, and the pitfalls.
Lead with a specific question, never a pitch
Open the first message with the one thing only this person can tell you (e.g. 'Have you ever seen a process transfer fail because of tacit tooling knowledge rather than labor cost?'), not with what Moduloa is. Experts answer sharp questions; they ignore visions. The pitch, if any, is one sentence of context so they know why you are asking.
Warm intros beat cold outreach, and Norway is your warmest edge
A one-line intro from a shared contact converts an order of magnitude better than a cold email. Spend Wave 1 in your own network and language specifically to earn the intros that carry you into robotics, Asia, and capital later. End strong calls with 'who else should I be talking to?' and make the intro trivial to forward by writing the blurb yourself.
Ask explicitly what would break the thesis
The stated goal is to learn what makes the thesis wrong or stronger, so make that the literal question: 'Where does this fall apart?' 'What am I underestimating?' Frame disagreement as the outcome you want. Polite validation is a failed call; a precise reason you are wrong is the best possible result at Stage 0.
Come prepared enough to be worth their time
Read their work first and reference it specifically. Know the thesis cold so you can spend the call on their expertise, not explaining yourself. Bring a falsifiable claim they can react to (e.g. 'I claim qualification time, not labor cost, is the real bottleneck — true or false?') rather than an open-ended 'what do you think of my idea?'
Respect and cap their time
Ask for 20-30 minutes and mean it. State up front you are not selling and not raising. Send 2-3 questions in advance so they can decline or prepare. End on time even if it is going well; leaving them wanting more is what earns the second call and the intro.
Close the loop with what you actually learned
Follow up within a day with a short note stating the one or two things their input changed in your thinking — not a generic thank-you. This proves you listened, converts a one-off into a recurring sounding board, and is what earns light advisory relationships. When a public prediction or journal entry shifts because of them, tell them.
No NDA theater, and be honest you are pre-factory
You are building in the open at Stage 0 with no factory — say so plainly. No NDAs, no secrecy, no pretending to be further along. Openness lowers the cost of them helping and is what makes the learning honest; the moat is the framework you will build over years, not a slide you must protect today.
Separate learning from selling in every interaction
These are learning interviews, not pilot pitches or fundraising. Say it explicitly, especially to investors and operators, because it is what unlocks candor. A pilot lead or advisor is a bonus that emerges from a good conversation, never the ask. The moment it feels like a pitch, the corrections stop.
Short — 5-8 sentences, readable on a phone without scrolling. Structure
- one line of genuine, specific connection — their paper, their transfer case, the intro path, not flattery;
- one sentence of who you are and the honest frame — solo founder in Norway, building in the open, pre-factory, not selling, not raising;
- the single specific question only they can answer, phrased as something to react to rather than an open prompt;
- the concrete, small ask — 20-25 minutes, and offer to send 2-3 questions in advance;
- the public thesis link so they can vet you themselves. One question, not five. No attachments, no deck, no NDA, no 'quick chat about an exciting opportunity.' The subject/first line should name the specific topic, not the company. Write it so the reply is easy even if the answer is 'you're wrong, here's why' — that is the reply you want.
Think in three waves across the 3-6 month window, not a flat 10-20.
stay local and warm — Norway/Nordics industrial plus one or two reachable Manufacturing & operations academics (Netland-type). Goal is not coverage but calibration: let these first calls rewrite your questions before you spend colder relationships. Rewrite your one-paragraph ask and per-person questions after every 2-3 calls.
go to the load-bearing and operational domains — Robotics & humanoids, Supply chain & production routing, Factory software — where you most need falsification of the labor-cost and portable-blueprint claims. Deliberately seek people likely to disagree.
once the thesis has survived or been reshaped, take it to Industrial investors & capital (commercial pressure-test, now credible because it is partly falsified) and Asia manufacturing ecosystems (the hardest-to-reach benchmark, best approached via intros earned in earlier waves). Leave ~30% of the call budget unbooked so warm intros generated along the way can be spent where the thesis is proving weakest. Cadence of roughly one call per week keeps synthesis ahead of volume; batching more than two per week early tends to waste the sharpest corrections because you have not yet processed the last one.
- A one-paragraph ask: who you are, that you are a solo Norway-based founder building in the open at Stage 0 with no factory, the single thesis claim you want their view on, and an explicit '20 minutes, not selling, not raising.'
- The public thesis link (and the /predictions and /journal pages) so anyone can check you are real and serious in 30 seconds without you having to explain — let the public artifact do the credentialing.
- One falsifiable version of the core claim, stated so an expert can simply agree or disagree (e.g. 'Automation offsets the Norway-vs-China wage gap only for process families X and Y' — true, false, or where?).
- A specific, person-by-person question tied to what only that individual knows — drafted before you write the outreach, not improvised on the call.
- A lightweight, consistent way to record and synthesize learnings (one doc per call: what they said, what changed, follow-ups, who they pointed to) so patterns across 10-20 calls are visible and can feed the dated predictions and journal.
- A short, honest bio and a ready-to-forward intro blurb someone else can paste to connect you onward — remove all friction from the warm intro.
- A running list of the process families / sub-claims where the thesis is strongest vs weakest, updated after each call, so you always know which question is now the most valuable to ask next.
- Pitching instead of asking. The moment the message or call becomes about how good Moduloa is, the expert stops correcting you and starts being polite — and polite agreement teaches nothing. Optimize for being told you are wrong.
- Fishing for validation and hearing it. It is easy to ask leading questions and count agreement as evidence. Ask the disconfirming version, seek out the people most likely to disagree, and treat a call where nobody pushed back as a call that failed.
- Front-loading the coldest, highest-stakes domains. Going to Asia ecosystems or top investors first — before the thesis has survived local, warmer contact — burns scarce, hard-won relationships on questions you were not yet sharp enough to ask well.
- Batching too many calls before synthesizing. Running five interviews in a week means the sharp correction from call one never reshapes calls two through five. Let each call rewrite your questions before the next.
- Treating intros and advisors as the scoreboard. Counting calls, contacts, or a maybe-advisor as success while the actual goal — a partly-falsified, sharper thesis — stalls. Measure learning and changed predictions, not headcount.
- Ghosting after the call, or thanking without substance. Failing to close the loop with what specifically changed wastes the relationship: no recurring sounding board, no onward intro, and the expert never sees that their time mattered.
Who to learn from, by domain
Seven domains, ordered by suggested outreach sequence — warmest and highest-learning first. Open a domain for the full guide.
01 Norway / Nordics industrial Now
This is the founder's nearest, warmest network and the single most credible real-world test of the core thesis: Norway is a high-wage country where researchers and firms have already documented cases of automated production beating low-cost-country factories (a car-parts maker found robotized production in Norway was faster and cheaper than its Chinese plant). If the "labor cost advantage erodes" argument is true anywhere, evidence for or against it already exists in Raufoss, SINTEF and NTNU reshoring studies. Being Norway-based, Sondre can reach these people in person and in his own language, and their feedback carries institutional weight for later credibility, funding and pilots.
Tests the foundational assumption that physical AI + automation erode the labor-cost advantage enough to make production location-flexible in a high-wage economy — and the harder second claim that the value is in a routable framework (standards, certification, routing) rather than in owning capacity. Norwegian reshoring researchers can confirm or break the "automation offsets wage gap" premise with hard cases (which reshoring succeeded, which failed, and why), and cluster/catapult operators can stress-test whether "production blueprints moved between certified hubs" is realistic given how real qualification, tooling and tacit process knowledge actually behave.
Warm, in-person and in Norwegian is the strong play here — this is a small, high-trust ecosystem where a good intro travels fast. Route 1 (best): use the Norwegian Catapult / Siva / Innovation Norway front door. SMEs can apply for small development-project support with a node guiding you start-to-finish; that application is itself a legitimate reason to book a call and get routed to the right SINTEF/Raufoss people. Route 2: email academics directly — Norwegian researchers reply to specific, well-read cold emails, especially from a local founder who has clearly read their paper; open with one sharp question tied to their exact finding, not a pitch. Route 3: show up. Attend a Raufoss / SINTEF Manufacturing / MIL open day or industry event and talk to people at the lab. Register matters: with academics be precise, curious and cite their work; with cluster operators and factory managers be concrete, numbers-first and humble about not yet having a factory. Across all: lead with "I'm testing whether this thesis is wrong" — the learning framing disarms and Nordic culture rewards honesty over salesmanship. Keep first asks tiny (20–30 min, one or two questions).
English is fully viable — Norwegian professional and academic life operates in English daily and every named institution publishes in it. But Sondre is Norwegian, so writing the first cold email in Norwegian to Norwegian recipients is a warmth and credibility signal (it says "local, serious, one of us") and costs nothing; switch to English the moment there's a non-Norwegian in the room. Nordic directness is real and helpful: be brief, skip flattery and hype, state exactly what you want and why, and expect equally blunt feedback — if the thesis has a hole they will say so plainly, which is the point. Flat hierarchy means a solo founder can email a senior researcher or cluster director directly without a gatekeeper; substance beats titles. Match register: academic (precise, evidence-first, hedged) vs. operator (practical, cost/lead-time/yield-first).
- Read Henrik Brynthe Lund / NTNU / SFI Manufacturing reshoring work, esp. 'Make at home or abroad? Manufacturing reshoring through a GPN lens' (ScienceDirect) — it studies nine Norwegian firms that reshored and names the conditions under which automation beats low-wage offshoring; know those conditions cold.
- Understand the Raufoss model: NCE Raufoss cluster (5 core firms + ~46 SMEs, ~5000 employees, 95% export share, light-metals + automation), and that SINTEF Manufacturing runs the Manufacturing Technology Norwegian Catapult Centre (MTNC) with modular 'mini-factories' — this is a living example of shared, certified capacity, close to your routing idea.
- Map the funding/access layer: Norwegian Catapult is run by Siva with Innovation Norway and the Research Council; SMEs can get small development-project support via a node. Know which of the ~5 centres / 8 nodes fits (MTNC/Raufoss for automation & metals, MIL/Grimstad for mechatronics & robotics pilot testing).
- Read SINTEF Manufacturing's advanced-manufacturing and robotics pages (NextGenRob, Robot Manipulator Lab, SFI Manufacturing) so you can name specific programs and speak to where the science actually is on gripping/handling/flexible automation vs. the hype.
- Have your own one-page thesis and a short list of the exact assumptions you want tested ready to send — Norwegians will ask 'what specifically do you want from me?' within the first two minutes.
- Of the Norwegian firms that reshored on the back of automation, which ones actually beat their low-cost-country alternative on total landed cost — and what conditions had to hold (volume, part complexity, competence, tooling) for it to work? Where did it fail?
- How much of a production process is genuinely transferable as a 'blueprint' versus locked in tacit knowledge, tooling and local qualification? If I moved a certified process from Raufoss to another hub, what breaks first?
- Does the Catapult / mini-factory model prove that shared, certifiable, modular capacity is real — or does each line end up too bespoke to route work between hubs the way I'm assuming?
- Where does high Norwegian labour cost still win despite automation, and where does it still lose decisively? What's the honest boundary of the 'automation erases the wage gap' claim?
- If the moat is the framework — standards, certification, data, routing — who in Norway already owns pieces of that (certification bodies, SINTEF, cluster standards), and does that make my framework redundant or fundable?
- For a solo founder with no factory, what's the smallest credible proof that would make SINTEF/Raufoss/a manufacturer take the routing thesis seriously enough to co-run a pilot?
A short list of concrete corrections to the thesis (the conditions under which automation does and does not offset wage gaps, and where 'routable production' breaks) grounded in real Norwegian reshoring cases — plus, ideally, one senior SINTEF/NTNU/SFI Manufacturing researcher willing to be an informal advisor or sounding board, a warm intro to a Raufoss cluster firm or MTNC/MIL operator who has actually moved a line, and a live route into the Catapult/Siva/Innovation Norway support pathway that could later fund a small pilot. Success is a sharper, partly-falsified thesis and 2-3 named people who will keep taking your calls.
02 Manufacturing & operations Now
This domain owns the single hardest claim in the thesis: that a production process can be abstracted into a portable, certifiable "blueprint" and reliably re-executed in another certified hub with comparable quality, yield, and cost. Production-system academics and shop-floor operators know exactly where the tacit knowledge, tooling, ramp-up curves, and process variation live — the things that make transfer easy or impossible. They are also the people who have actually run plant-to-plant transfers and global lean rollouts, so they can tell you empirically whether "routable capacity" is a real phenomenon or a slide.
Core assumption under test: that manufacturing capacity can be commoditized and routed like compute — i.e. a blueprint plus a "factory OS" can move production between hubs without losing yield, quality, or economics, and that humanoids/physical AI genuinely erode the labor-cost advantage rather than just shaving a slice of it. Operators can confirm or break: (a) how much of a working line is tacit/undocumentable, (b) real ramp-up time and yield loss on process transfer, (c) whether labor is actually the binding constraint versus capex, tooling, qualification, and supply chain, and (d) whether standardization across sites (the lean-rollout analogue) has ever worked at the fidelity the thesis needs.
Segment by register — the opener that works for an academic will fall flat with an operator and vice versa.
- Academics (Netland-type): warm, specific, deferential-but-peer. Cold email works if you cite a specific paper and ask ONE crisp, researchable question they'd find interesting. They reply to intellectual curiosity, not to founders pitching. LinkedIn is fine as a secondary channel; a comment on their blog (Netland runs better-operations.com) or a thoughtful reply to a post can precede the email. Frame yourself as building an open thesis and wanting to know where it's wrong — academics love being asked to falsify something.
- Operators / plant & ops managers: warm intro beats cold every time; they are time-poor and allergic to consultant-speak. If cold, be concrete and humble — "I'm trying to understand how real a plant-to-plant process transfer is; you've done it, I haven't." Offer to share what you learn. Short messages, no jargon, no deck.
- Execs / cluster & institute leaders (SINTEF Manufacturing, NCE Raufoss, TotAl-gruppen): approach via the institution's project/partnership channels — these bodies exist to talk to industry and often host open events. For a Norway-based founder, the Raufoss cluster is the single best warm surface: geographically local, industry-facing, and used to fielding outside inquiries. Attend an event first, then follow up with named people you met. Across all three: lead with learning, name the specific assumption you want tested, keep the first ask to 20-30 minutes, and never send a pitch deck.
English is fully viable across this domain in Europe — Netland and SINTEF/NTNU researchers publish and work in English, and Norwegian industry runs on English for anything technical. For the Norwegian operators and the Raufoss cluster specifically, the founder's native Norwegian is a genuine advantage: use it for warm rapport, switch to English for precision on technical terms. Register matters more than language: academics expect a literate, citation-grounded, hedged tone (claims framed as hypotheses to test); operators expect blunt, concrete, no-buzzword talk and respect directness — Nordic directness is an asset here, not a liability, so skip the American-style hype and over-enthusiasm. Avoid "disrupt/platform/routable capacity" marketing framing with operators; translate the thesis into plant language (transfer, ramp-up, yield, qualification, standard work). For any later Asia outreach (TPS/Toyota lineage), assume Japanese-language and introduction-mediated norms — that is a later, intro-dependent effort, not a cold-email one.
- Read 2-3 of Netland's actual pieces before emailing him — his blog better-operations.com and at least one paper on corporate lean-program rollout across global plant networks (his ETH/POM work on standardization across sites is directly the 'routing' analogue); cite the specific finding you want to probe.
- Understand the Raufoss cluster concretely: SINTEF Manufacturing (~85 staff, JV with six manufacturers), NCE Raufoss (17 companies, ~5000 employees, 85% export ratio, lightweight materials + automated production for automotive/defense). Know who does what before you reach out — sintef.no/en/manufacturing and raufossindustripark.no.
- Learn the vocabulary of process transfer and qualification: PPAP/APQP, first-article inspection, process capability (Cpk), ramp-up curves, tacit vs codified process knowledge, standard work — so operators take you seriously in the first three sentences.
- Read enough Toyota Production System / lean canon (Ohno, Womack & Jones, plus Netland's take on why lean programs succeed or fail on transfer) to speak credibly about what standardization across sites has and hasn't achieved.
- Get concrete on the humanoid/physical-AI labor claim before testing it — have real numbers on what fraction of manufacturing cost is direct labor in the target processes, so the 'labor advantage erosion' question is grounded, not hand-waved.
- When you've transferred a running process from one plant to another, how much of the working line turned out to be tacit or undocumented — and how long was the real ramp-up to matching yield?
- In your experience, is direct labor actually the binding cost constraint, or is it capex, tooling, qualification, and supply chain? If labor went to near-zero via humanoids, what would still stop production from being 'portable'?
- Corporate lean rollouts try to standardize practice across a global plant network. Where does that standardization break down, and what does that tell us about whether a 'blueprint' can be routed between certified hubs?
- What would a 'process blueprint' actually have to contain to let a different, certified site hit the same quality and cost — and is that even fully expressible, or is some of it irreducibly tacit?
- Which product/process types could plausibly become portable and routable in the next 5-10 years, and which never will — where's the real boundary?
- If you wanted to prove this thesis wrong in one experiment, what transfer would you run and what would you expect to fail?
Concrete good outcomes, in priority order: (1) A sharp correction to the transferability/labor claim from someone who has actually run process transfers — either a documented failure mode that reshapes the thesis or credible confirmation of where routing is plausible. (2) An informed academic ally (Netland-type) willing to react to the thesis periodically and point to literature — a light advisory relationship, not a formal one yet. (3) A warm intro onward into the Raufoss cluster or a specific plant willing to describe a real transfer, which later becomes a candidate pilot/certification test site. (4) A crisp list of the process families where the thesis is strongest vs dead-on-arrival, so predictions can be dated and falsifiable.
03 Robotics & humanoids Now
This domain is the load-bearing wall of the entire Moduloa thesis: the claim that humanoids + physical AI erode the labor-cost advantage only holds if general-purpose robots actually reach reliable, unsupervised manipulation at a cost that undercuts human labor in manufacturing. These are the only people who know how far real deployments are from that point, versus the hype. If they tell you the "hand is still the bottleneck" and most deployments are teleoperated data-collection dressed up as autonomy (which the Q2 2026 evidence strongly suggests), your whole timeline — and therefore Moduloa's Stage-0 sequencing — has to move.
The core enabling assumption: that within a credible horizon, humanoid/physical-AI systems become general and cheap enough that labor cost stops being the reason production sits where it sits. Specifically it tests (a) whether dexterous manipulation and task-generalization are close or a decade out, (b) whether robot "capacity" is genuinely portable/re-deployable between sites and tasks (the premise behind routable capacity), and (c) whether the real bottleneck is hardware/AI or integration, safety, and operations — which would shift the moat away from robots and toward exactly the framework layer Moduloa is betting on.
Register-match, because this domain is three different cultures.
- Academics: cold email works if it is short, references a specific recent paper of theirs, and asks one genuine research-shaped question — not a pitch. Email the senior PhD/postdoc, not just the PI; they are more reachable and often more candid about what does not work. Do NOT lead with your startup.
- Operators/deployment engineers: warm intro or LinkedIn beats email. Lead with a concrete, non-obvious observation ("your BMW deployment loaded ~90k parts over a year — I'm trying to understand how much of that was autonomous vs supervised"), signal you are not press and not selling, and offer to keep it to 20 minutes.
- Execs/investors: only worth it once you have a sharp framing; they reply to a crisp thesis they can react to, and can give warm intros onward. Across all three: frame yourself honestly as a solo founder building a thesis in the open who wants to learn what would make it wrong — that disarms the "what are you selling me" reflex and is genuinely your goal. Conferences (CoRL, ICRA, Automate, Hannover) are the highest-yield channel because a 5-minute hallway conversation converts to a call far better than any cold email. Publish your thesis and predictions first so there is a public artifact that makes you legible and gives people a reason to engage.
English is fully viable for the Western academic and industry side — it is the working language of robotics research and of Figure/Apptronik/1X/Agility. Academic register: precise, hedged, citation-aware; overclaiming instantly loses credibility, so match their measured tone (which also fits the Moduloa voice). Industry/operator register: plainer, outcome-focused, allergic to buzzwords like "routable capacity" until you have earned them — describe the concrete thing first. China is the real language gap: for Unitree/AgiBot and the broader ecosystem, Mandarin and WeChat are the actual channels; senior researchers often read English but cold outreach in English to Chinese operators has low hit rates — go through a bilingual analyst or diaspora researcher instead. Nordic directness is an asset here: being straightforwardly honest about being pre-factory and wanting to be proven wrong reads as refreshing to engineers, not as weakness. Avoid Norwegian; keep everything in English.
- Read Figure's Helix posts and the BMW Spartanburg deployment details, and Apptronik's Apollo / Google DeepMind Gemini Robotics setup — know the specific numbers (parts loaded, uptime claims, precision) so you can ask what was autonomous vs teleoperated.
- Read one or two recent VLA / robot-learning papers from whoever you're contacting (arXiv cs.RO, or the person's own Scholar page) so your email cites their actual work, not a generic compliment.
- Understand the current consensus bottlenecks — dexterous manipulation and tactile perception, data scarcity/cost of good demonstrations, and that the pilot-to-production gap is largely an integration/operations problem (Gartner: <20 firms scaling beyond pilots by 2028) — so you can probe rather than be educated from zero.
- Learn the China vs West bifurcation: unit volumes and price points (Unitree G1 ~$16k vs Western $50k+, ~80% of 2025 installs in China) so you can ask credible questions about where cost curves actually bend.
- Read your own thesis critically enough to state, in one sentence, the exact robotics assumption you most want them to attack — so the conversation is about learning, not validation.
- Of your real deployments, what fraction of task-time is genuinely autonomous versus teleoperated or human-supervised — and how is that ratio actually trending, not projected to trend?
- Is dexterous manipulation the true bottleneck, or is it integration, safety-certification, and operations? If a plant hands you a new task tomorrow, how long until the robot does it reliably — days, months, or 'not yet'?
- How portable is a trained robot's capability between sites and tasks? If I move the same hardware to a different line/product, how much of the learned behavior transfers versus needs re-collection and re-training?
- At what real, fully-loaded cost per hour (including integration, maintenance, downtime, teleoperation) does a humanoid actually undercut human labor for a given manufacturing task today — and where do you see that number in 3 and 5 years?
- Where does the durable value/moat sit: in the robot hardware, the foundation model, the data, or the certification/integration/operations layer around it? What are you happy to outsource?
- What would have to be true for 'production blueprints routed between certified robot-run hubs' to be real rather than a slide — and what's the single biggest reason you think it won't happen the way I'm describing?
A calibrated, evidence-based revision of the thesis's robotics timeline and its central labor-cost claim — ideally confirmation or (more valuably) correction of how fast autonomy and cost-parity are actually arriving. Concretely: 2-3 candid conversations that pin down the autonomous-vs-teleoperated reality and the true bottleneck; at least one relationship with a deployment engineer or robot-learning researcher willing to be a recurring sanity-check or informal advisor; one or two warm intros onward (e.g. from a researcher to an operator, or from an analyst into the China ecosystem); and a sharper public position for Moduloa that has survived contact with people who actually ship robots. A pilot lead is a bonus, not the target at Stage 0.
04 Supply chain & production routing Now
This domain is the operational heart of the thesis: if production really can become portable, certified, routable capacity, then supply-chain and network experts are the people who know whether qualification, MOQs, lead times, tariffs and dual-sourcing make "route a blueprint to the best certified hub" feasible or fantasy. They have already lived the closest real-world analogues — resilience/reshoring research (MIT), the Triple-A framework (Stanford), and live capacity-routing marketplaces (Xometry, Fictiv) — so they can tell Sondre where his model bends reality. Getting this wrong means the whole "moat is the framework, not the factory" claim collapses on contact with real qualification friction.
The core assumption that manufacturing capacity can be commoditized into interchangeable, certifiable, routable units — that a production "blueprint" plus a certification standard is enough to move volume between hubs without re-qualification killing the economics. Specifically it tests: (1) whether supplier/part qualification is fast and portable enough for dynamic routing, or whether it is the actual bottleneck (each new hub = months of PPAP/first-article/audit); (2) whether the labor-cost-arbitrage-erosion premise changes sourcing decisions that today are driven far more by tooling amortization, MOQ, logistics and tariffs than by wages; and (3) whether a routing/OS layer captures durable value or just becomes a thin pass-through margin like today's manufacturing marketplaces.
Segment the outreach by register, because these three groups reply to different things.
- Academics (Simchi-Levi, Sheffi, Lee): cold email works if it is short, specific, and shows you read their actual work. Open with one sharp intellectual question that engages a named paper/framework, position yourself as a founder-researcher testing a thesis (not selling), and explicitly say you want to know how it might be wrong. Academics respond to good problems and to people who cite them correctly; they ignore pitches. A 20-minute call ask is realistic; a PhD student or postdoc in their lab is often the faster, higher-yield first contact.
- EMS/marketplace operators: LinkedIn (a personalized connect note, not InMail spam) beats cold email; lead with a concrete operational puzzle from their world ("how long does re-qualifying a board at a second site really take you?") rather than your vision. Warm intros from mutual connections convert far better here — mine your network and the Nordic cluster/Katapult ecosystem.
- Nordic/SINTEF/NTNU: go in person or via a single warm email in Norwegian; Nordic directness means a plain "I'm a solo founder in Norway testing this thesis, can I get 30 minutes to have you poke holes in it" lands well. Across all groups: name-drop the seed thinkers to show you've done the map, keep the first ask tiny, share a one-page thesis link rather than a deck, and always close by asking who else you should talk to (the warm-intro chain is the real prize).
English is fully viable for the entire academic and EMS-executive layer — MIT/Stanford faculty, and senior ops leaders at Flex/Jabil/Foxconn/Xometry all operate in English, and Norwegian-accented professional English is a non-issue. For deep Asia-based EMS shop-floor and mid-tier supplier contacts, Mandarin (mainland/Taiwan) and Japanese matter and a language barrier is real below the executive tier, so start at the English-speaking corporate/executive level and treat the shop-floor layer as later, intro-dependent. Register: academics want intellectual precision and correct citation (industry-consultant tone reads as shallow to them); operators want blunt, numbers-first practicality and are allergic to vision-speak; execs want signal that you respect their time. Nordic directness is an asset — with SINTEF/NTNU and Norwegian industry, skip the American hype register entirely; plain, honest, "help me find the flaw" framing is culturally native and builds trust. Match "measured, honest, no hype" to every segment.
- Read Simchi-Levi's Risk Exposure Index work (the 'Risk Exposure Index Revisited', Operations Research 2019 / SSRN 2875596) so you can ask about qualification/recovery-time as the real routing constraint, not just node risk.
- Read Hau Lee's 'The Triple-A Supply Chain' (HBR 2004) and be ready to argue whether a routing OS delivers Agility/Adaptability/Alignment or breaks Alignment by commoditizing suppliers.
- Read Sheffi's reshoring take (MIT Sloan 'Reshoring, restructuring, and the future of supply chains') — he is skeptical reshoring happens fast; have a crisp answer for why humanoids/physical AI change his timeline, or concede where he's right.
- Study Xometry's and Fictiv's actual business models (Xometry investor materials + the 'network vs marketplace' critiques) so you can ask operators why pass-through margin stays thin and whether certification could change that.
- Learn the concrete qualification vocabulary — PPAP/APQP, first-article inspection, IATF 16949 / AS9100 / IPC standards, MOQ and tooling amortization — so operators take you seriously in the first two minutes.
- What actually stops a company from second-sourcing a given part today — is it wages, tooling amortization, MOQ, logistics, tariffs, or qualification time? Rank them, because the thesis bets on wages eroding.
- If humanoid/physical-AI labor removes the wage gap, does re-qualification (PPAP, first-article, audits) become the new binding constraint on moving production between hubs — and how many weeks/months is that in practice?
- Is a production 'blueprint' even a transferable object for real parts, or is too much tacit process knowledge, fixturing and tuning locked to a specific line and workforce?
- Xometry and Fictiv already route capacity across thousands of shops and still earn thin pass-through margin — what would a certification/OS layer have to own for the routing intelligence to capture durable value instead of being commoditized?
- Where does dynamic routing break on regulated/safety-critical goods (auto, aero, medical) where the certification is site-specific and change control is deliberately slow — is the addressable market only low-criticality parts?
- What would you need to see in the next 2-3 years to believe portable, routable certified capacity is real — and what single piece of evidence would convince you it will never work at scale?
Concrete, ranked outcomes: (1) a sharp correction — a clear verdict on whether qualification-time, not labor cost, is the true bottleneck, which either sharpens or breaks the thesis and should feed directly into the Stage 0 predictions/journal; (2) one or two recurring sounding-board relationships (a lab postdoc, a Nordic SINTEF/NTNU researcher, or a mid-level EMS ops person) willing to react to future iterations; (3) at least one credible advisor candidate from the academic or EMS-operator layer; (4) warm intros onward down the chain (academic -> their industry contacts, or operator -> a plant willing to describe real re-qualification numbers); and eventually (5) a pilot lead — a certified hub or EMS site willing to let you test what 'routing a blueprint' actually requires. Success is measured in falsification or reinforcement of the routing assumption, not in headcount of calls.
05 Factory software (MES / digital twin / physical AI) Now
The Moduloa thesis rests on a "factory OS + routing intelligence" that can move production blueprints between certified hubs and coordinate portable capacity. The people who build MES, digital twins, and physical-AI platforms already own the software layer that would have to become that OS — they know exactly how far today's tools are from a blueprint that is truly portable across sites, machines, and vendors. They are the reality check on whether "routable capacity" is a software problem that is nearly solved, decades away, or structurally impossible given how fragmented the shop-floor stack really is.
The core testable assumption: that a production process can be captured as a certifiable, portable "blueprint" and re-instantiated at another certified hub with predictable output — i.e. that the factory OS + routing layer is the real moat rather than any single factory. These people can confirm or break the sub-claims: (1) how much of a working line is actually transferable vs. tacit/site-specific tuning; (2) whether MES/digital-twin data models are interoperable enough (OPC UA, MTConnect, OpenUSD) to move a process without a rebuild; (3) whether physical AI (learned robot policies) genuinely reduces re-commissioning cost across sites, or just shifts it. If most value is locked in site-specific integration and human tuning, the "portable blueprint" premise weakens sharply.
Segment by persona — the register differs sharply.
- Academics / NIST / standards people: cold email works well if you cite a specific paper and ask one sharp conceptual question; they reply to genuine intellectual engagement, not pitches. Lead with "I'm testing a thesis and I think your work bears on whether it's wrong."
- Founders/CTOs (Tulip, integrators): warm intro via MIT/startup networks beats cold; if cold, LinkedIn with a 3-sentence note that shows you understand their product's actual limits (not marketing) and asks a learning question, not a demo.
- Big-vendor execs (NVIDIA, Siemens, Rockwell, PTC): hard to reach directly; instead target their developer advocates, field engineers, and PMs who speak at GTC/Hannover — meet them at the booth or in session Q&A, then follow up. For all: explicitly say you are pre-factory, building in the open, and want to learn what breaks the thesis — that honesty disarms the "is this a sales call?" reflex and is on-brand. Keep first ask to 20-25 minutes. Offer to share your dated public predictions/thesis as a credibility artifact.
English is fully viable for this entire domain — it is the working language of MES/digital-twin/physical-AI globally, and NVIDIA/PTC/Siemens/Rockwell/Tulip all operate in English. Academic register (NIST, university labs, standards bodies) rewards precision, citations, and hedged claims; avoid startup hype words ("disrupt", "OS for X") with this group or you lose credibility. Operator/integrator register is blunt, practical, allergic to buzzwords — talk in commissioning time, downtime, changeover, MOQ of engineering hours. Exec register is outcome- and ROI-framed. Nordic directness is an asset here: state plainly "I might be wrong and I want you to tell me why" — it reads as serious, not weak, to both academics and operators. Asia (FANUC/YASKAWA Japan, many robot OEMs) matters for physical AI, but their public-facing platform teams and GTC presence operate in English; Japanese/Mandarin only becomes relevant for deep OEM relationships later, not for Stage-0 learning calls.
- Read Tulip's own framing of 'composable MES' and its integration/connector list (SAP, Teamcenter, SCADA, OPC UA) so you can ask what genuinely does NOT transfer between sites, not what the brochure says does.
- Watch or read one NVIDIA GTC 2026 physical-AI session and the Omniverse DSX / Mega Blueprint and Isaac GR00T material, so you can speak precisely about sim-to-real and virtual commissioning rather than generically about 'AI'.
- Skim the OPC UA / MTConnect companion-spec and one NIST digital-twin interoperability paper to understand where the data-model seams actually are — this is the crux of 'portable blueprint'.
- Understand PTC's Onshape CAD-to-Isaac-Sim (OpenUSD) bridge and what FANUC/ABB/KUKA integrating Isaac/Omniverse into virtual commissioning actually changes about line bring-up time.
- Have a one-page, honest statement of the Moduloa thesis and its testable claims ready to share, plus your dated public predictions, so experts can attack specifics.
- When you move a working process from one plant to a new site, what fraction of the value is in the transferable configuration vs. tacit, site-specific tuning and integration that has to be rebuilt by hand?
- Is today's MES/digital-twin data model (OPC UA, MTConnect, OpenUSD) interoperable enough that a 'production blueprint' could be re-instantiated at another vendor's line without a full re-integration — and if not, what is the hardest seam?
- Does physical AI / learned robot policy genuinely reduce re-commissioning cost when the same process runs at a different site, or does it just move the cost from mechanical setup to data collection and validation?
- If routable, portable capacity became real, what breaks first — the software, the certification/liability, or the economics of keeping a line idle enough to be 'available' to route to?
- Who would actually own the 'factory OS' layer — do you see incumbents (Siemens/Rockwell/NVIDIA) foreclosing a neutral routing layer, or is there room for an independent standards/routing player?
- What is the strongest reason you think 'production as routable capacity' will NOT happen this decade — where is the thesis naive?
Concretely: (1) at least two blunt corrections that sharpen or falsify the 'portable blueprint' claim — ideally a numeric feel for how much of a line is truly transferable; (2) one map of where the real interoperability seams are (which standards, which gaps), so Moduloa's certification/routing framework targets a real problem; (3) one warm intro onward — to a systems integrator who has actually relocated a line, or to a standards-body contact; (4) over time, one credible technical advisor from this domain (an ex-MES/digital-twin or physical-AI person) who can pressure-test the factory-OS architecture. A pilot lead is a bonus, not the goal at Stage 0.
06 Industrial investors & capital Next
This domain is the only one that can pressure-test whether Moduloa's core commercial claim holds: that "routable, certified production capacity" is a venture-scale platform business, not just a nice engineering idea. Investors who have actually underwritten marketplace/network-effect economics AND hard-tech manufacturing have seen the failure modes (multi-tenanting, thin take rates, cold-start liquidity, capex-heavy moats that don't compound) up close. Getting their read early — while it's free and low-stakes because you're not raising — is the cheapest way to learn whether the framework-as-moat thesis is financeable before you sink years into building it.
The specific assumption under test: that the moat is the FRAMEWORK (standards, certification, data, routing) rather than any single factory, and that this framework can capture durable, compounding value as a platform/capacity-marketplace — with defensible unit economics (take rate, who pays, switching costs, network effects between certified hubs) — rather than collapsing into a low-margin brokerage or being commoditized by the hub owners themselves. Secondarily, whether "portable, certified, routable capacity" is investable at Stage 0 by a solo non-technical-capital founder in Norway, or whether the capital structure (deep-tech capex + platform patience) is fundamentally mismatched.
Lead with the fact that you are NOT raising — this is disarming and unusual, and it changes the entire dynamic from "another founder wants money/free advice" to "someone did the work and wants a sharp reader." Segment by type
- Nordic/EU seed investors — cold email or a warm intro via NTNU/SINTEF/Oslo Innovation Week works; Nordic directness is an asset, so be concise and specific, no hype.
- US thesis-driven partners (Eclipse/Schematic/a16z) — do NOT cold-pitch; instead engage substantively with their published thesis (a thoughtful reply to an essay, a specific disagreement, one genuinely new data point from your predictions/tools), because the reference class is 'nobody remembers the hundredth founder asking for free advice.' Offer something: your dated predictions, the humanoid-race tracker, a specific dataset.
- Marketplace thinkers (Andrew Chen/NFX/Tirole) — read and apply their frameworks first; only reach out with a sharp, framework-specific question, not a general ask. Universal rules: get the RIGHT person (the partner covering industrials/supply-chain, not the firm generically); first two sentences must state why it's relevant to what they think about; keep it under ~150 words; one polite follow-up after 5-7 business days is expected. Warm intros through their portfolio founders convert far better than cold — spend effort there. Register: with academics use precise, citation-aware language; with operators be concrete and outcomes-focused; with execs be brief and respect their time.
English is fully viable and standard across the entire investor domain — Nordic, European, and US VC all operate in English, and Nordic investors expect it. As a Norwegian founder, Nordic directness (concise, no overselling, honest about unknowns) is a cultural advantage here, not a liability — it matches the 'measured, honest, no hype' Moduloa voice and reads as credible to serious hard-tech investors who are allergic to founder hype. Register shifts matter more than language: academic marketplace economists (Tirole and that school) expect rigor and correct use of terms like two-sided markets, multi-homing, take rate; US thesis VCs write in a mission/narrative register (American Dynamism, reindustrialization) and respond to founders who speak that frame fluently; operators/angels want plain, concrete, numbers-first talk. No Mandarin/Japanese needs for THIS domain — Asia-language realities belong to the supply-chain/manufacturing domains, not industrial capital, which is Western-English-dominated.
- Read the target's actual thesis in their own words: Eclipse's 'digital transformation everywhere except where it matters,' Schematic's 'digital industrial' framing, and a16z's American Dynamism essays — and be able to say precisely where Moduloa fits or breaks their model.
- Internalize the marketplace-economics canon before talking to anyone: Andrew Chen / a16z on network effects and marketplace defensibility, and the NFX network-effects framework — specifically multi-homing/multi-tenanting risk, take-rate dynamics, and cold-start liquidity, because these are exactly where 'routable capacity' will be attacked.
- Map the reference class of what already exists: manufacturing marketplaces (Xometry, Hubs/Protolabs) and high-labour-cost robotics winners (AutoStore) — know why they worked or stalled, so you're not re-pitching a solved/failed pattern.
- Have your own Stage-0 artifacts ready to offer as proof-of-work: the dated predictions, the humanoid-race tracker, and a crisp one-line articulation of the framework-as-moat claim — investors reward founders who bring data, not decks.
- Check each fund's stage, ticket size, and geography (e.g., via Dealroom/Crunchbase and their portfolio pages) so you approach only people whose mandate could plausibly touch this — approaching the wrong-stage/wrong-sector investor signals you didn't do homework.
- Where does a 'routable, certified capacity' marketplace most plausibly die — cold-start liquidity, multi-homing by hub owners, or a take rate too thin to support the certification/data cost base? Which have you personally watched kill a marketplace?
- Is the framework (standards + certification + routing data) actually a compounding, defensible moat, or does it collapse into a low-margin broker once hubs can transact directly — and what would have to be true for it to compound?
- Who realistically pays, and for what — the hub for certification, the buyer for routing, or a transaction fee — and which of those has the pricing power to be venture-scale rather than a services business?
- Does the capital structure even work: can one venture carry deep-tech/capex patience AND platform-liquidity patience simultaneously, or is that a fundamental mismatch that means Moduloa should be structured differently (e.g., asset-light framework only)?
- What single piece of evidence at Stage 0 — before any factory — would move you from 'interesting thesis' to 'financeable company,' and what evidence would make you write it off entirely?
- Is a solo non-capital-native founder in Norway the wrong shape of founder for this bet in the eyes of the capital that could fund it, and if so what's the honest fix — co-founder, geography, or narrative?
A good outcome is primarily CORRECTION and framing, not validation: a clear read on which part of the commercial thesis is weakest (moat, take rate, capital structure, or founder-market fit) so you can either strengthen it or kill it cheaply at Stage 0. Concretely: (1) one or two thesis-driven investors who agree to be periodic sounding boards / eventual advisors on the commercial model; (2) a sharpened articulation of the moat-and-monetization story in language capital actually finds financeable; (3) at least one warm intro onward — ideally to an operator who has lived the marketplace/certification problem, or to the specific EU non-dilutive program that fits a blended capex+platform venture; (4) an honest verdict on whether this is fundable as one company or should be restructured (e.g., asset-light framework first). Not a term sheet — you're not raising, and saying so is what makes the learning honest.
07 Asia manufacturing ecosystems (China / Taiwan / Japan / SE Asia + India) Next
This domain is where the Moduloa thesis either dies or gets proven. Asia — especially the Shenzhen–Dongguan electronics cluster and the Yangtze River Delta — is both the incumbent low-labor-cost manufacturing base the thesis claims will be disrupted AND, right now, the fastest deployer of the very humanoids the thesis rests on (Chinese firms hold ~90% of global humanoid installations, with output forecast to grow ~94% in 2026). These are the only people on earth actually running physical-AI-plus-humanoids at scale in real factories, so they hold the ground truth on whether robotics erodes the labor-cost advantage or simply deepens Asia's lead through vertical integration, co-located component clusters, and 10-14 day prototype cycles. If production is going to become portable and routable, this ecosystem is the benchmark to beat.
The thesis's core causal claim: that humanoids + physical AI erode the labor-cost advantage enough to make production PORTABLE and ROUTABLE between hubs. Asia ecosystem insiders can confirm or break the harder, quieter assumption underneath it — that the advantage is mainly labor cost. The rival hypothesis they can voice: the real moat is agglomeration (thousands of co-located suppliers, tooling density, tacit process knowledge, guanxi-based supplier trust, 2-hour logistics radius for actuators/reducers/sensors), none of which move just because labor gets automated. If that is right, humanoids make Shenzhen MORE dominant, not more routable, and "certified routable capacity" is solving the wrong bottleneck. They also test whether a neutral routing/certification framework could ever sit above ecosystems that are themselves racing to own the full humanoid stack.
Segment your outreach by persona — the register that lands is completely different for each.
- Academics/analysts: cold email works well and is the highest-yield, lowest-friction entry point. Reference a specific paper/report, ask ONE sharp question, keep it to 5 sentences, be explicit you are learning not selling. Nordic directness is an asset here.
- Taiwan/Japan EMS and OEM execs: LinkedIn is viable and normal; Taiwan management is English-comfortable and internationally networked. Lead with a specific, informed observation about their operations, not a generic connect.
- Mainland-China operators (Unitree/UBTech/AgiBot etc.): cold email/LinkedIn is weak; the ecosystem runs on guanxi and WeChat, and a warm intro is worth 100 cold messages. This is exactly where Innovation Norway/NORBA/EU Chamber and HAX/SOSV alumni earn their keep — get introduced.
- Sourcing agents/consultants: transactional and responsive; a paid intro call (even a modest consulting fee) is a legitimate, fast way to buy 90 minutes of ground truth and often a chain of onward intros. Universal rules: come as a curious founder building in the open (your public thesis, tools, and journal are your credibility — link them), never as a vendor; offer to share what you learn back; be patient and relationship-first with Chinese contacts, punctual and prepared with Japanese ones, fast and specific with Taiwanese ones. What gets a reply everywhere: a genuinely good question they haven't been asked, and evidence you did your homework.
English is fully viable for: academics/analysts region-wide, Singapore, Taiwan EMS management, and Japanese executives at multinationals (though prepare for more formality and slower, consensus-driven responses in Japan — punctuality and a proper introduction matter more than speed). English is a real barrier for: many mainland-China factory-floor operators, mid-level engineers, and smaller robotics-firm staff, where WeChat + Mandarin dominate — expect a sourcing agent, a bilingual intermediary, or a diaspora contact to be the bridge. Do NOT expect meaningful cold reach into Unitree/AgiBot-tier engineering teams in English without a warm Chinese-speaking connector. Cultural register: with Chinese contacts, invest in the relationship before the ask (guanxi is literal relational capital — favors, informal contact, reciprocity); with Japanese contacts, hierarchy and a credible referrer matter; Taiwanese contacts are the most English-fluent and the easiest first entry into Asian hardware. Norwegian bluntness reads as refreshing honesty to academics but can feel abrupt to mainland contacts — soften the opener, keep the question sharp.
- Read the current supply-chain landscape so you don't ask beginner questions: MERICS on UBTech, the Tianxia Gongchang / TrendForce / Omdia humanoid supply-chain and shipment data (China ~90% of installs, ~50k units 2026, YRD vertical integration), and the actuator/reducer split (Japan's Harmonic Drive/Fanuc/Yaskawa ~60-70% of precision reducers/servos vs Chinese parity at 30-40% lower cost).
- For each specific person: read their actual paper, patent, IR deck, or conference talk and be able to reference one concrete point — this is the single biggest reply-rate lever.
- Study the geography and agglomeration argument (Shenzhen–Dongguan electronics cluster, Yangtze River Delta 2-hour logistics radius, 10-14 day vs 12-week prototype cycles) so you can pressure-test your own routability claim against it in the conversation.
- Learn the diversification map (Foxconn/Quanta/Pegatron moving AI-server and EV production to Mexico, Guadalajara/Monterrey; India PLI scheme + Tata Electronics; Vietnam/Thailand) — this is real-world 'production moving between hubs' and either supports or complicates your routing thesis.
- Set up and warm up a WeChat account, and identify your institutional bridges (Innovation Norway China desk, NORBA, EU Chamber Shenzhen) before you need them, so a warm intro is one message away.
- When a humanoid or a lights-out line removes most of the direct labor from a task, what fraction of your true landed-cost advantage actually disappears — and what stays because it's tooling density, co-located suppliers, and process know-how that can't be shipped?
- You already run the most vertically integrated humanoid supply chain on earth; does automation make you MORE locked to the cluster (in-house motors/reducers/sensors, 2-hour logistics) or does it genuinely let capacity move to, say, Norway or Mexico?
- Your own firms are diversifying to Mexico and India for tariff and geopolitical reasons — what actually broke when you tried to replicate a Shenzhen line elsewhere, and how long did it take to reach parity?
- If a neutral third party offered a certification + routing layer that let a customer move a production blueprint between certified hubs, would you ever plug into it — or is capturing the full stack yourself the whole strategy?
- What is the real current bottleneck to humanoid mass deployment in your factories — hardware cost, dexterity/reliability, integration/reprogramming time, or something operational I'm underweighting?
- For a portable 'production blueprint' to actually run identically in two hubs, what tacit knowledge or supplier relationship has to travel with it that a spec sheet and a certification never capture?
Concretely: (1) At least one honest correction or validation of the labor-cost-erosion premise from someone who runs real production — ideally quantified (how much cost is labor vs agglomeration). (2) A clearer, evidence-based read on whether automation makes Asian ecosystems more sticky or more routable — this directly sharpens or kills the 'routable capacity' framing. (3) One or two ongoing relationships: an analyst/academic willing to be a sounding board (potential advisor), plus a warm intro chain into an operator or EMS team. (4) Real diversification case studies (Mexico/India line replication) as evidence for or against portability. A great outcome is being told precisely WHY the thesis is wrong by the people best positioned to know — that's more valuable at Stage 0 than polite agreement.
Know someone in one of these domains?
This map is public precisely so people can suggest better contacts, correct it, or make an introduction. One warm intro beats any number of cold emails.
Make an intro or a correction →Where this came from
Synthesized from the thesis's own list and public research. Public figures and institutions only — no private contact details are compiled.
pom.ethz.ch · better-operations.com · mtec.ethz.ch · leanblog.org · sintef.no · sintef.no · raufossindustripark.no · effra.eu · euroma-online.org · poms.org · shingo.org · en.wikipedia.org · idss.mit.edu · ssrn.com · ctl.mit.edu · mitsloan.mit.edu · gsb.stanford.edu · gsb.stanford.edu · karkhana.io · umbrex.com · theassemblystudio.com · rapiddirect.com · valueaddvc.com · cornfordandcross.com