Monthly update — June 2026

This is the public mirror of an email sent to a private list on 2026-06-17. The body below is the email, lightly edited. Numbers are point-in-time as of the send date; the corpus is live and these figures will have moved.

Quick June update, and the headline is a real one. With two segments now under paid contract, and excluding my own comp, Municipal Alpha is operationally in the black, the revenue from the two contracts already covers what it costs to run the whole company. The clearest proof the model works is that one of them, Maine Trust for Local News, converted from a pilot into a twelve-month MSA, a real annual contract, not a toy trial.

The two contracts

They span two segments. On the media side, Maine Trust for Local News is on a twelve-month MSA, roughly $500 a month, redline in flight. On the economic-development side, Our Katahdin plus One North, the Great Northern Paper mill-site redevelopment up in Millinocket. Together that is about $13.2K of booked ARR. These aren't logos for show, they pay for the product and they're shaping it, calibration sheets, coverage cleanup and MSA redlines all in motion this month.

I'll say the $100K number out loud since some of you will ask. I'm driving toward $100K ARR by the end of August, even with the summer slowdown. It's an ambitious benchmark I use to set the pace, not a forecast I'm pegging myself to or a cliff I fall off of, but I'm going after it.

The part I think you'll appreciate most

With those two contracts, minus my comp, we're in the black, roughly $1,100 a month coming in against about $410 a month going out. The striking part is what that $410 actually is: the company's entire team is a small stack of AI subscriptions.

The AI workforce is about $360 a month, and it splits cleanly by role. Claude Max, running Opus, is $200 a month, the builder and the operator, it writes the code, runs the pipeline, drafts the outreach, maintains the whole rule system. Gemini is about $159 a month across roughly 114,000 calls, the high-volume analyst, it bulk-classifies the municipal documents and serves as the adversarial second model that reviews anything before it leaves the building. The metered Anthropic API line is basically zero, the Claude work runs through the Max subscription, not pay-per-call.

The rest is about $50 a month, ScraperAPI plus the domain, and the DigitalOcean server is free, covered through the Roux Institute affiliation. So the whole company runs on roughly $410 a month of tooling, no payroll, no office, no seat licenses. That's the 98 to 99 percent margin story in concrete terms, a path to a million in revenue that runs on a stack you could put on a personal credit card.

The irony cuts the right way. The capital efficiency that is the entire point is the thing the standard programs can't categorize, Google's startup-credit program keeps bouncing us with "you need VC backing." We're too efficient to fit the funding-dependent mold. Still working that one, so if anyone has a line into Google for Startups it wouldn't hurt.

And to be straight about it, this isn't purely solo. The biggest pieces of leverage are things I'm not paying for. The Roux Institute covers the server, has been the platform behind every one of those speaking slots, and is the network where the conversation I mention below started. And this list has pulled real weight, New North surfaced the capital-efficiency framing I'm now building the whole company around, and several of you have sharpened the thesis or opened a door without my asking. The $410 is the cash cost, the leverage around it is the bigger number, and most of it is generosity I'm trying to be worth.

That conversation is with Devang Pethkar, a Northeastern student with a modeling background. Nothing's formalized yet, but it's leaning toward him coming on to stress-test the unit economics, the per-vertical ACV and margin models and the path to a million, and he's already engaging at exactly that level. His first question pulled on a real seam, that the highest-ACV vertical is also the smallest market, so scale has to run through the larger markets at a repeatable price, and we've got a working session this week to go deeper. If it comes together it's real analytical help at zero cash, and either way it's a good preview of how having the whole decision-making layer encoded lets a second set of eyes plug straight into the strategy.

The wedge, and where the scale lives

Worth being clear about something: the EDO and media contracts aren't the revenue play, they're a paid wedge. They cover the costs, but their real value is distribution. Being the data partner to local-news organizations puts us in front of their coverage and their audience with visible attribution, when our intelligence shows up in a story it runs with our name on it, which is branded visibility in the exact markets we serve, and domain authority too, these are long-standing, well-respected domains, so an attributed mention or backlink carries real SEO weight back to us. And economic-development boards are full of regional business leaders who sit one relationship away from the people who actually buy the higher-value product. Those EDO relationships already opened our two hottest conversations, Ready.net on the broadband side, in flight toward a commercial-tier contract around $36K, and Cianbro, one of the largest construction and infrastructure firms in the Northeast.

On where the scale comes from, Devang's question framed it well. The volume path runs through the AEC firms, architecture, engineering and environmental, somewhere between 120,000 and 160,000 US firms, of which maybe 12,000 to 25,000 are mid-and-large firms with a public-works practice that could actually buy, a market big enough to support the goal with a repeatable motion. The infrastructure developers, the Ready.net and DESRI and Syncarpha kind of buyers who site projects through local government, are the high-ACV end, per-deal valuable, a mis-sited project costs millions. The part the early feedback has made clear is that these two aren't a choice. They overlap so heavily, same ecosystem, same municipal signals, same prospecting surface, that chasing the high-value developers adds essentially no operational load on top of the AEC motion. So we run both at once, AEC for the volume and the developers for the ACV, off one engine.

For sizing on the two we're already in: the EDO market is roughly 8,000 to 12,000 organizations depending on how you count, with a serviceable core of a few thousand. Media is around 7,900 local-news outlets, and the consolidation everyone reads as a death spiral is actually a tailwind for us, ownership is concentrating into nonprofit trusts, and MTLN is exactly one of those, so a single trust contract covers a whole portfolio of papers.

The corpus is doing double duty

A few things on how the corpus works for us beyond being the product. Coverage growth is now demand-driven, we onboard a new region only when a lead pulls for it, Texas for Ready.net, New Jersey for Syncarpha on the solar-storage side, so every new region is already attached to a live deal and the spend tracks real demand instead of speculation. The outreach itself runs on the corpus, every cold email leads with a real, verified capital signal pulled from our own documents in the recipient's own backyard, a specific town's wastewater upgrade before the RFP, not a generic pitch, the product is the outreach. And it's working, Syncarpha is the clearest warm hit so far, a solar-storage developer engaging off exactly that motion, on the New Jersey coverage we stood up for them. The meta-layer keeps turning too, every buyer conversation feeds a rule that sharpens how I present the product the next time, the running thread has become "intel, upstream before the decision," earned conversation by conversation and encoded so it doesn't reset.

Speaking and ecosystem

Three speaking slots since the last update, all downstream of the Roux relationship. Startup Maine in mid-May, where a lot of these EDO and investor relationships started. NENS, the Northeast National Security Conference, on June 3, which is where the capital-efficiency thesis came up with New North Ventures. And AI in Action at the Roux Institute on June 5, on the municipal-innovation track, which is also the relationship now bringing us the co-op. The Roux throughline keeps compounding, same root, more branches each month.

Numbers (point-in-time 2026-06-17)

3,300+ municipalities indexed across all 50 states.
700,000+ source documents in the corpus.
200,000+ classified signals across a dozen-plus verticals. Cross-customer learning is the defensibility, and we extract from sources others find uneconomical.

One ask

And only if it's easy. If you know a business that would benefit from seeing local-government activity coming before it happens, anyone in site selection, permitting, infrastructure, or selling into local government, I'd welcome the intro. No pressure though, I'm making good progress on my own.

One last thing, and it's the most on-thesis part of all of this

I didn't write this update. The system did. I gave it the topics, the guidance and the judgment calls, what to highlight, what the numbers mean, where to push, but the data, the structure and the voice all came from the system working as designed. The thing you just read is itself the product, the clearest proof I can hand you that one person and a system that compounds can out-produce a team.

The reason I can hand it that much and trust what comes back is that the whole thing runs on an encoded set of principles, twelve of them, that govern every decision it makes. They're the closest thing the company has to a constitution, they're how the data stays honest and the voice stays mine even when I'm not the one typing. Here they are.

The 12 Commandments

I. Every claim traces to a source.
II. Guard the trust.
III. The system runs without you.
IV. Market signal outranks internal signal.
V. Ship, then compound.
VI. Match the weight of the solution to the weight of the problem.
VII. Diagnose before you prescribe.
VIII. Fix at the most general level the evidence supports, but no further.
IX. Be wrong fast.
X. Capture what you learn or lose it.
XI. Every repeated judgment is a failure of encoding.
XII. The work compounds.

Matt