Monthly update — April 2026

This is the public mirror of an email sent to a private list on 2026-04-24. The body below is the email verbatim. Numbers are point-in-time as of the send date; the corpus is live and these figures will have moved.

Starting a monthly update cadence for Municipal Alpha. Some of you have been hearing about this all month, others haven't heard much in a while, either way this is just an honest snapshot of where things are, no ask attached. This is #1.

What I saw this month

The core product frame keeps tightening. The thing I sell is not a document feed, it is the chain of custody on a signal, so that a buyer sees a municipal procurement move six to eighteen months before it shows up in an RFP, an FAA filing, or a credit downgrade. "We saw it first" is the clearest way to describe it. The clearest walkthrough is the Amherst NH case study, 32 months of public municipal signals before American Tower signed the ground lease:

https://municipalalpha.com/research/amherst-tower

In April two buyer conversations moved that one-off backtest from an artifact into a product shape:

  • Katie Richards at PFAS Laboratories gave substantive feedback that changed the signal filter for environmental consultants. The valuable signals are pre-treatment-decision stage (board of health studies, monitoring permits), not installation stage. A domain expert narrowing the filter for me is more useful than a polished pitch.
  • Maine Trust for Local News signed up to put Municipal Alpha in their grant submission this Friday, 12-month funded pilot covering their Maine newsroom footprint. Decision 2026-06-12. Scott Monroe values the early-signal capability specifically, for the same reason: newsrooms show up to a planning board meeting after the story has already happened. The platform changes that.

Pipeline

  • 13 active buyer conversations across towers, bond credit, civil engineering, PFAS/environmental, alt-data, local news.
  • Warmest: MTLN (grant-funded pilot, decision June 12).
  • Hardest: AMT is warm but slow (Mike Bitto bandwidth, $1K/mo verbal pilot from 4/10 not yet papered), and four tower cold outreach emails are silent at 10 days.

Pipeline value — today

Where a price or grant scope is on the table, ~$20-25K annualized:

  • AMT: $1K/mo verbal from the 4/10 call, MA-only weekly pilot, not yet papered. Annualizes to $12K.
  • MTLN: grant-funded 12-month pilot, funder decision 6/12. Grant cap $5-10K.
  • PFAS Laboratories and SW Cole Engineering are both active unpriced conversations. Segment-typical paid-pilot rates are $400-1,000/mo.

What it could become in the next 12 months

If pilots paper and expand:

  • AMT's MA pilot is the swing variable. Multi-state rollout at similar cadence, or the referral-fee and exclusivity premiums Mike floated on the call, puts that line at $60-150K annualized.
  • MTLN post-grant converting to the $500-750/mo subscription Scott originally pitched would add $6-9K/yr recurring on top of the grant.
  • PFAS Labs and SW Cole are each beachheads for their segments (environmental consultants and NE geotech). Each converting plus one adjacent customer would be $15-35K/yr combined.

Envelope: $20-25K today, $85-200K annualized if the pilots paper and expand at segment-typical rates. All contingent on deals getting signed, expansion decisions, grant outcomes, and conversion timing breaking right. Not a forecast, just the shape of what's in play.

The meta layer

This is where most of my actual work lives.

The bet is that I can run a business this size without staffing up, because every buyer interaction, every correction, every missed signal becomes a rule that the system itself enforces next time. I capped planned headcount at 20 and am building toward that constraint on purpose. Three concrete examples from this month:

  • Katie's pre-treatment-stage feedback is now a deterministic rule, not a human judgment call. The next PFAS buyer gets the narrower filter automatically.
  • Every externally-sent document (prospect sample, investor artifact, cold pitch) runs through a claim-verification gate that refuses to render if a date or dollar amount cannot be traced to its source. Fabrication is a failure mode the system prevents rather than a reviewer catches.
  • Every decision ends up in a per-day log, indexed into a machine-readable ledger. Session bootstrap reads the index. Future Claude instances, or future collaborators, pick up where the last session left off. The operator has fewer things to remember, the system has more memory.

The compounding effect is the interesting part. Capex spent on the meta layer reduces opex on every future customer. That's the business model as much as the product is.

Numbers

  • 2,600+ municipalities configured across 50 states.
  • 472,000+ documents in the corpus, 84,000+ classified signals.
  • 13,000+ tower sites cross-referenced with FAA and FCC data.
  • 43,000+ construction permits tracked, 100+ public tickers linked to municipal activity.
  • Monthly infra cost: around $400 (DigitalOcean ~$45, Claude $200, ScraperAPI $50, Google Cloud ~$80, Google Workspace $16). Revenue: $0 until MTLN grant lands or the AMT pilot papers.
  • Solo operation, my time is the main cost.

Ask

No ask this month, just setting the cadence. If something in here prompts a thought or an intro, reply. If not, see you next month.

More soon.

Matt https://municipalalpha.com