Q2 2026: 184 Subscribers, a Solo Quarter, and the Number That Actually Matters
An honest accounting of Q2 2026: 184 email subscribers, 728 followers, a solo quarter with no viral moments, and the Agentic AI Security Stack delivered.
Every article published between April and June was already a chapter in the same document
Q1 closed with a specific commitment: the Agentic AI Security Stack would ship in Q2, freely available, no portal. That commitment sat on top of 150 email subscribers and a manuscript running well past its original seven-module scope.
Looking back at Q2 from the end of it, the organizing logic is clearer than it felt from inside. April moved through governance gaps, attack surface taxonomy, adversarial inputs, and prompt injection mechanics. May pushed into trust boundary failures: orchestrator-to-orchestrator propagation, authority boundary problems, shadow AI detection, memory poisoning as a persistence mechanism. June shifted to evaluation and testing methodology, asking what tools actually exist for assessing LLM security properties and where those tools have themselves become part of the attack surface.
Every topic selected, every article drafted, was a chapter being tested in public before it locked. Reader questions and gaps that surfaced in published articles found their way into the corresponding chapters. The book launched June 26. Article 070 covers what is in it. This piece covers what the quarter around it looked like.
The cadence shifted mid-quarter. The practical result is in the June open rates: the four Wednesday posts averaged 25.7%, with cleaner read-through figures than the compressed twice-weekly pairs that preceded them.
The algorithm started driving discovery before the subscriber count reflected it
The quarter ends with 184 email subscribers, up from 150 at the close of Q1. That is 34 new subscribers, 23% growth, in a quarter with no collaboration event, no viral moment and no traffic spike attributable to a single piece.
That net figure masks two-way movement. Some of the audience acquired through Q1’s collaboration spikes did not stay for a solo technical quarter. Others arrived precisely because of it. The 184 at the end of Q2 are disproportionately the readers who showed up for deep AI security and stayed when that was all there was. A churned audience that self-selects toward the content direction is a better outcome than a retained audience that is only loosely aligned with it.
The honest way to read those numbers requires a distinction Substack does not always make obvious in how it presents growth data.
Email subscribers receive posts directly in their inbox. They are the readers who opted into a recurring claim on their attention. They represent actual reach in the sense that any given post lands in front of them regardless of what the platform’s algorithm decides that day. Followers operate differently: they see content when they happen to be scrolling the platform feed, visibility contingent on the algorithm surfacing it. Every email subscriber automatically follows the publication, but the reverse does not hold.
The follower count went from 333 to 728 over Q2, a 119% increase that reads dramatically on a chart. That growth is real at the awareness level and useful as a top-of-funnel signal, but it is not the metric that reflects inbox access or reader depth. The 34 email subscribers are the ones who crossed the threshold from light interest to active subscription.
How they arrived matters. Substack’s network data shows 76% of new subscribers over the past 90 days came through the Substack network itself: recommendations, restacks, and platform discovery. Q1’s growth inflections were driven by the Wyndo and ToxSec collaborations. Q2 had no equivalent event. The Substack algorithm replaced the function that cross-newsletter partnerships served in Q1.
That shift is worth examining closely, because it also explains why the follower count grew faster than the subscriber count. The algorithm converts curious browsers into followers efficiently. Converting followers into email subscribers requires a reason specific and timely enough to cross the inbox threshold. Without a collaboration providing that reason, follower accumulation outpaced subscription.
The residual from Q1’s partnerships is still visible in the Q2 traffic data. The AI Maker’s publication sent 206 views to this newsletter in Q2, four months after the collaboration that generated them. A single well-executed partnership in February is still driving discovery in June. That is the data point I keep returning to when thinking about what Q3 should prioritize.
One referral source tells a more specific story: view.hashicorp.com sent 35 views during Q2. HashiCorp’s official LinkedIn newsletter, HashiBits, featured “The Secret That Shouldn’t Exist”, the January piece on how modern agents access APIs without carrying static keys. Thirty-five readers followed the link from a major infrastructure vendor’s editorial feed. That is not organic discovery; that is the content landing in front of a Vault and Terraform audience through a channel that specifically curated it. It did not produce a subscriber spike, but it is the kind of third-party validation that does not show up in any metric the Substack dashboard tracks.
On the content side, the Q2 posts that drew sustained engagement shared a consistent profile. Treat Coding Agents as Privileged Build Participants had the highest combined open rate (31.6%) and view count of any Q2 post, alongside the highest number of signups within 24 hours of publication. Distillation at the Projection Layer, covering how model weight distillation can function as an extraction vector, pulled a 30.2% open rate. AI Agent Memory Poisoning drew the most views and the quarter’s highest read-through rate.
The pattern holds from Q1: posts that arrive at an uncomfortable conclusion about infrastructure teams have already shipped outperform posts that provide comprehensive coverage of a well-documented problem. Specificity and a defensible technical position travel further than thoroughness.
The newsletter now reaches readers in 35 countries. The US accounts for 86 total signups, Brazil for 22, and Canada, Spain, and Singapore each sit in the 6 to 8 range. The distribution is widening without an explicit international strategy.
A solo quarter is the baseline a collaboration-driven Q3 can actually measure against
My read of Q2 is that a disciplined solo quarter establishes a ceiling. Email subscriber growth of around 20 to 25% per quarter without a deliberate acquisition mechanism is the structural reality at this stage. The Substack algorithm drives follower accumulation. Converting followers into email subscribers requires a specific reason at the right moment.
Q3 is designed to provide those reasons. A collaboration is in the works. The Q1 pattern already demonstrated what a well-executed partnership does to subscriber conversion; Q3 will test whether that repeats.
The book now functions as a reference anchor the newsletter did not have before. Content that extends, applies, or challenges the stack’s framework has a concrete base to build from. Readers arriving at the newsletter for the first time have a complete reference to orient against rather than an implied promise of one. That change in the newsletter’s architecture is the most durable output of the quarter.
The metric to watch going into Q3 is email subscribers, not followers. The follower count at 728 is useful context, a signal that the content is reaching audiences beyond the existing subscriber base. The email subscriber count at 184 reflects actual reach: the readers who have given this newsletter a recurring claim on their inbox. Q3 growth will be measured against that number, not the larger one.
184 to start. More to come.
Peace. Stay curious! End of transmission.