A founder’s note on the 33-year arc that put me at the bottom of the wireless stack, the pivots that taught me what the human mesh actually is, and what Edge Orbital is building for the customers no one is profitably watching.
The confession
I started Edge Orbital expecting to build a defense company. I’m building something else now.
The pivot wasn’t a strategy reversal. It was the slow recognition — across about eighteen months of customer conversations, technical iteration, and one consumer-safety detour I didn’t expect to take — that the architecture I’d been building has its truest home in a market the defense thesis had taught me to overlook. The rural electric cooperative. The community-solar battery operator. The county-owned water-treatment plant. The substation that sits on a distribution feeder no one is profitably watching because the procurement-consulting-fee economics of the incumbent monitoring vendors don’t reach below the Fortune-500 procurement cycle.
That tier — by line-mile, by population, by exposure to actual physical-security incidents — is the majority of the U.S. distribution grid. Federal infrastructure policy already knows it. The Department of Energy’s Grid Resilience and Innovation Partnerships program directs $10.5 billion in §40101(c) funding through exactly the buyer tier the incumbent vendors don’t reach. The 2025 expansion of NERC CIP-014 brought physical-security planning obligations down from transmission into the distribution-feeder layer. NFPA 855 tightened fire-and-physical requirements on battery energy storage at the community-solar tier. Three regulatory and funding windows converged on the same customer with the same problem at the same time. The procurement options haven’t followed.
That gap is what Edge Orbital is being built into. This is the note on how I got here.
What the atoms taught me
The American Dynamism investing thesis — popularized by Andreessen Horowitz’s dedicated fund and now adopted across the infrastructure-tech VC ecosystem — bets that the next generation of national capability will come from venture-backed startups building hard technology in the physical world. The shorthand inside the category is “atoms, not bits”: a reaction to the prior decade of software-first venture capital, which built the application layer of modern life on top of an industrial base it largely ignored.
The first wave of the bet is the defense and aerospace cohort — Anduril, Hadrian, Varda, Apex, Saronic. Visible, conviction-led, the easy part. The second wave is the layer underneath those bets: the grid that powers them, the substations that distribute that power, the battery storage that makes it dispatchable, and the rural and small-industrial sites where most of the actual atoms-not-bits work happens. The second wave is less visible because the customer tier is less visible. Co-op general managers are not on coastal investor decks. Distribution-feeder substations are not on portfolio-company landing pages. But the physics of the bet — that the United States runs on infrastructure most software people have forgotten about — is the same physics either way.
I spent thirty-three years at the bottom of the wireless stack. That’s where you learn this lesson the hard way.
The thirty-three-year arc
I built the world’s first metropolitan WiFi network in Arizona in 1997, before the industry had a name for it. I ran enterprise point-to-multipoint deployments for fifteen years after that, in geographies where the option of trenching fiber was not actually an option. I built municipal fiber where the budget admitted it. I designed and shipped tactical mesh radios for environments where the assumption of a coordinating base station was a luxury the operator did not have.
The recurring lesson, across all of it: the bottom of every wireless layer is where physical infrastructure meets reality. The application layer assumes the network exists. The network layer assumes the radio works. The radio layer assumes the spectrum is clean, the power is reliable, the site is reachable, and the operator can get to it when something breaks. Those assumptions are where everything quietly fails.
The startups that win at the bottom of the stack are the ones that take those assumptions seriously enough to engineer against them. The ones that don’t — that ship products designed against the assumptions of the application layer — lose to physics. Cellular base stations need backhaul that isn’t always there. Cloud-dependent security cameras need uplinks that fail during the exact incidents they were installed to record. Centralized monitoring needs a SOC contract the operator can’t afford and a SCADA team integration the operator doesn’t have.
I am not the first person to notice this. I am one of the people who has spent enough time at the bottom of the stack to know the shape of the failure mode by feel. That muscle memory is what Edge Orbital is built around — not a piece of IP, although there is IP; not a hardware bill of materials, although the bill of materials is what it is. The shape of the architecture follows the shape of the customer’s actual operating reality, because thirty-three years of getting it wrong in less-honest ways taught me what the operator’s actual operating reality is.
What I thought Edge Orbital was
The original Edge Orbital thesis was tactical mesh for contested-spectrum environments. Defense. Dual-use as the consumer-side de-risking layer. Edge AI at the radio. Mesh that holds together when the cellular layer doesn’t. CJADC2-adjacent, although that was never the lead phrase. The market thesis was that the prime contractors couldn’t build at venture velocity, and the venture-backed startups couldn’t credibly run tactical RF without the kind of founder who’d been at the bottom of the wireless stack for a few decades.
Most of that is still true. The tactical mesh architecture is real. The IP is real. The defense applications are real — and explicitly within the dual-use scope of what we’re building. American Dynamism investors who lead with the defense thesis will see Edge Orbital land cleanly within the first wave of the category.
What turned out to be wrong, or at least incomplete, was the assumption that the defense customer was the right wedge. The defense customer is real, but the defense procurement cycle is slow, and the defense customer’s actual operating environment is one I cannot get to without an OTA pathway and a year of program-office relationships I had not yet built. The architecture I was building had a faster, larger, and structurally-clearer customer waiting elsewhere — I just hadn’t gone looking for it yet.
The consumer-safety detour is what surfaced where the architecture actually fit.
What the consumer pivot taught me
About a year into Edge Orbital, we built Tripwire Recon — a consumer safety app that lives on the iPhone, and is the commercial face of the underlying sensor and mesh architecture. The original framing was straightforward: prove the perception, fusion, and pre-incident detection capability on a real consumer product, generate the kind of revenue and product-market-fit signal that de-risks the venture timeline, and use that proof as the credibility anchor for the defense conversation.
What happened in the consumer build is that the customer didn’t behave like a defense customer at all. The customer behaved like an operator of their own personal safety infrastructure — and the architecture I had built for contested-environment tactical comms turned out to be uncannily well-fit to the operator’s actual reality. Cellular coverage was unreliable in the geographies that mattered most. Always-on monitoring was the cost-prohibitive default the customer was trying to escape. The threat profile was behavioral — casing, scoping, dwelling, returning — rather than discrete-event. The customer wanted the system to fire on the missed check-in, not on a button they might not be able to press.
That word — operator — kept coming back. The campus student walking home at 2 a.m. is operating a personal safety system whether they have one or not. The bar bouncer running a doorman shift is operating a perimeter. The off-grid hiker is operating a personnel-recovery posture. The customer was not buying a “consumer safety app.” The customer was buying an infrastructure layer for their own operating reality.
10-page PDF: faction breakdowns, zone strategy, mesh tech explained. Yours free.
I started calling it the human mesh — the network of people who would notice the wrong thing in the moment a system designed for them could act on it. The phrase wasn’t marketing. It was the literal architecture: a sensing layer that surfaces what the human mesh would already see, made proactive through pre-incident detection.
The defense framing dissolved against the consumer framing — but not because defense had stopped being real. Because infrastructure for the human mesh turned out to be a more honest description of what the architecture actually was. Defense was the contested-environment edge case. The human mesh was the system.
From the human mesh to the grid edge
Once the architecture had a name that wasn’t defense-shaped, the next customer surface became visible.
An operations manager at a rural electric cooperative is operating an infrastructure system. The system has perimeter exposure: substations, distribution feeders, line equipment, leased pole assets. The threat profile is behavioral — the recon window before copper is pulled, the multi-night reconnaissance pattern that preceded Moore County and similar incidents, the equipment status anomalies under conditions a perimeter camera alone would miss. The operator does not have a SOC contract. The operator does not have a SCADA-team integration budget for a new sensor stack. The operator does have a CIP-014 obligation that just expanded into the distribution-feeder layer, a federally-funded resilience program designed for exactly their tier, and a copper-theft loss curve that is now a baseline operating cost rather than an exception.
The same architecture that surfaces what the human mesh would already see in a campus walk-home corridor turns out to surface what an operations manager would already see in a substation perimeter. Different operator, same architectural pattern: edge-resident perception, sensor fusion, behavioral pre-incident detection, alerting that routes to the operator’s actual workflow rather than to a remote analyst contract.
NERC E-ISAC recorded substation physical-security breaches up 10× since 2016, with 3,500 breaches in 2024 alone — roughly two outages a week. The Department of Energy estimates copper theft alone costs American businesses about $1 billion per year, with damage multipliers running 10–28× the value of the stolen copper. The 2022 Moore County incident — 45,000 customers offline for days — was not on a transmission-tier substation. It was on a distribution-tier substation operated by a co-op. The parts of the grid most consequential to national interest infrastructure sit on co-op-owned distribution feeders. And the parts of the grid the legacy monitoring market has structurally failed to serve are exactly those parts.
This is the second American Dynamism wave. The first wave funded the defense and aerospace cohort. The second wave funds the layer that powers them — and the rural, distributed, operator-budget tier that funds the layer that powers them.
The nimble pirate
The trap inside the second wave is to read it as “compete with the F500 monitoring vendors at lower price points.” That framing loses on first principles. The incumbent monitoring vendors are not architecturally wrong — they are architecturally tuned for a customer tier whose math works at their cost structure. The math that works for an enterprise utility CFO is not the math that works for a co-op operations manager. The enterprise sales motion does not fit a line item. The line item does not carry an enterprise sales motion.
The right framing is structural. There is a buyer tier the incumbents cannot profitably serve. The buyer tier owns most of the distribution wire and most of the land area in the country. The buyer tier has a real and growing threat exposure. The buyer tier now has federal funding aligned with its specific cost structure. And the buyer tier has no security platform — not because no one has built one, but because the architecture the incumbents have been priced to support is not buildable for that customer at their cost structure.
This is what I’ve come to think of as the nimble pirate position. Not competing with the F500 platform vendors. Operating below the price floor they cannot cross without breaking their go-to-market. Architecturally distinct: edge-resident, sensor-fused, operator-priced, designed against the operations-manager spreadsheet from day one rather than the enterprise-procurement RFP. Federally funded at the buyer-tier match. Dual-use by design, in the precise sense the term means in American Dynamism investing — the underlying perception, fusion, and pre-incident-detection capability serves both civilian critical-infrastructure and adjacent dual-use applications without re-engineering.
Reindustrialization at the pole is not an abstract phrase to me. It is the literal demand profile of a 12-county rural electric cooperative trying to harden its substation footprint against a budget that hasn’t moved, with a federal funding program designed for exactly that customer, and a security architecture that has been architecturally absent from their reachable procurement options.
What I’m building, capability-level
Edge Orbital is pre-pilot, and we are explicit about that. What we are building toward is field deployment with the architecture above and the customer tier above as the explicit design constraints. We do not have a deployed customer count. We do not have an LOI. We do not have a hardware vendor brand to point at. We have the architecture, the IP, the founder credibility, and the customer-development conviction that the buyer tier is real, that the threat profile is real, that the federal-funding alignment is real, and that the architectural shape we have been building has a structurally-honest home in that tier.
The capability is three properties that work together.
First, each site stands alone. Every monitoring node runs full perception and pre-incident detection locally. No central cloud dependency. No SCADA-team integration cost. No always-on monitoring fee. A node continues to do its job during a regional internet outage, during a sustained attack on backhaul, during the kind of contested conditions where infrastructure monitoring matters most. Decentralized resilience as an architectural property, not a marketing word.
Second, operator-budget pricing. The system is designed against the operations-manager spreadsheet first. Per-site economics live at the tier where a 12-county co-op can deploy across its full footprint, not just one site. Without that, the buyer tier does not open. The architecture follows the unit economics, not the other way around.
Third, pre-incident detection on the patterns that matter. Copper-theft signatures (extended after-hours dwell, irregular movement against the fence line, vehicle stage-and-return patterns). Physical-attack precursors (the multi-night reconnaissance pattern that preceded Moore County and similar incidents). Equipment status anomalies under conditions a perimeter camera alone would miss. Detection is sensor-fusion-driven and runs at the edge — the alert fires before the failure mode completes, not as a forensic record afterward.
This is what I mean by infrastructure for the unguarded. Not metaphor. The unguarded edge of the grid. The unguarded perimeter of the small industrial site. The unguarded approach to the regional water-treatment plant. The places critical-infrastructure policy now obligates someone to monitor, that the incumbent stack was never priced to reach, and that adversaries have already identified.
What I’m not building
I am not building an enterprise platform priced down for small customers. The architecture is genuinely different.
I am not building a video management system. I am not building an alarm aggregator. The category is pre-incident detection, adjacent to during-incident acoustic detection, to CCTV’s post-incident review, and to all-incident defense platforms operating at national scale — and structurally below all of them in cost-per-site.
I am not building a defense vendor. The architecture is dual-use, by design, and the defense application is a real and explicit part of what we will sell into. But the wedge is the civilian critical-infrastructure tier, and the credibility I owe my investors and my customers is to say so plainly rather than to wrap the civilian thesis in a defense wrapper because defense is the easier conversation to have on a coastal Zoom.
I am not, finally, building anything I cannot demonstrate. Capability before claim is the standing rule for what we say in public — to investors, to operators, to the press. Pre-pilot means pre-pilot. Pre-hardware means pre-hardware. What I have built — the architecture, the IP, the customer development, the founder credibility — is what I have built. What I have not built — the deployed pilot count, the customer LOI, the verified field data — I will tell you I have not built, until I have built it.
The work is the work
The American Dynamism investing thesis is, fundamentally, an argument that the venture capital industry’s prior decade of software-first conviction missed a generation of physical-world technology that the United States now needs at scale. The argument is correct. The portfolio is being built. The first wave is visible.
The second wave — the layer underneath, the customer tier the first wave depends on — is being built more quietly because the customer tier itself is quieter. Rural electric cooperative general managers are not on coastal investor decks. Distribution-feeder substations are not on portfolio-company landing pages. The 832 rural electric cooperatives serving 42 million Americans across 56% of the U.S. landmass are not the visible part of the American Dynamism portfolio. They are the part the visible portfolio depends on.
Edge Orbital is being built for them. The architecture, the unit economics, the federal-funding alignment, the dual-use breadth — all of it. Infrastructure for the unguarded, built at the bottom of the wireless stack, by an engineer who has spent thirty-three years learning what the operator’s actual operating reality looks like. Supporting the national interest by being honest about which customer the work is actually for.
If the second wave aligns with your portfolio mandate, the investor overview walks through the technical differentiation, market opportunity, and current raise. If you operate a co-op, a community-solar BESS site, a small industrial perimeter, or any operations footprint that does not quite fit a CIP-014 budget, the critical-infrastructure thesis walks through the architecture. If you write about American Dynamism, distributed energy, rural electrification, or the second wave of the category, my email is on the about page and the lights are on.
The work is the work. Capability before claim. Atoms at the pole.
— CJ