The Trilogy Times

AUSTIN, TEXAS — Buried beneath the bigger headlines this year was a one-line announcement that, if you read between the lines, tells you almost everything about where Skyvera is headed: the ESW Capital telecom portfolio company has acquired STL's telecom products group, picking up digital BSS functionality spanning monetization, optical networking, and analytics.

On its face, it's a modest deal. No purchase price disclosed. No press tour. Just a terse notice that the assets now belong to Skyvera. But modest deals are how this portfolio always moves — and this is where it gets interesting.

Consider the sequence. Earlier this year, Skyvera completed its acquisition of CloudSense, the Salesforce-native configure-price-quote engine that handles the front end of telecom sales — the quoting, the configuration, the B2B2X complexity that trips up most carriers. Now Skyvera has absorbed STL's monetization and optical networking stack, which handles what happens after the quote closes: billing, charging, network provisioning, analytics.

Put those two pieces on the same table and you don't see a grab-bag of telecom software. You see a vertically integrated BSS stack being built one acquisition at a time, from quote to cash, under a single roof. A source close to the portfolio — who, for obvious reasons, can't be named — described the strategy as "filling in the gaps before anyone notices there was a map."

It's a classic ESW maneuver, and it tracks with the broader Trilogy playbook outlined in this paper before: buy mature, undervalued enterprise software cheap, staff it with Crossover's global talent bench, and consolidate. Skyvera has now done this with CloudSense, with Kandy, with VoltDelta, with ResponseTek — a growing shelf of telecom-adjacent products that increasingly look less like a portfolio and more like a single operating system for telcos trying to modernize legacy infrastructure.

Whether STL's optical networking and monetization tools get folded directly into CloudSense's CPQ flows, or run as a parallel offering, remains unannounced. But nothing in this portfolio happens by accident. Every acquisition is a sentence in a longer paragraph — and Skyvera, quietly, is still writing it.

AUSTIN, TEXAS — Marin Software spent two decades selling marketers tools to manage digital ad spend. It spent its final act in bankruptcy court, and it took ESW Capital just 66 days to walk out with the keys, according to court filings reported this week. For a firm whose stated playbook is acquiring mature enterprise software at 1–2x ARR and squeezing it toward 75% EBITDA margins, a distressed asset moving through Chapter 11 is not a cautionary tale. It's a discount rack.

Marin becomes the latest name in a portfolio that already runs past 75 companies and roughly $1.14 billion in total deployed capital — a number that keeps climbing while most of the market waits to see what AI-driven M&A actually looks like in practice. Boston Consulting Group's mid-2026 outlook describes a tentative recovery in dealmaking, heavy on caveats. ESW does not appear to be waiting for clarity. Speed is the strategy: identify sticky customers who can't easily migrate off a platform, acquire before competitors notice the opportunity, staff the surviving operation with Crossover's global remote talent, and raise support pricing on the way out.

The same week, Scott Alexander of Astral Codex Ten published a long, skeptical-but-engaged review of Alpha School, Joe Liemandt's two-hour-curriculum experiment in Austin. The two stories do not share a dateline, a sector, or a courtroom. They share an operator. Liemandt's instinct — find the slack in a system, strip it out, call the resulting margin proof of efficiency rather than extraction — shows up whether the system is a school day or a software license agreement.

Separately, Resmed's sale of MatrixCare to private equity suggests the broader software-and-healthcare buyout market is heating up around exactly the kind of mature, sticky assets ESW has spent two decades specializing in. Whoever is shopping next should expect company.

KYIV — Observe, if you will, the data center. A curious beast: windowless, humming, fed on electricity and cooling water, its only outward sign of life a blinking amber light. For years it has grown unchecked, colonizing floodplains and deserts alike, confident in its apex status. This week, nature reminded it otherwise.

In Russia, drones dispatched by Ukrainian forces found their way into a facility belonging to Yandex — colloquially, 'Russia's Google' — and in a single strike disabled supercomputers used to train the company's flagship AI model. An extinction event, localized but instructive: even the mightiest compute cluster is, in the end, a structure of concrete and silicon, vulnerable to anything that can fly and carry a payload.

Half a world away, in the Asia-Pacific basin, a gentler but no less consequential die-off is underway. Firmus and PLDT, two operators poised to go public, withdrew their IPOs this week, as governments from Manila to Jakarta paused new data-center permits amid concerns over power grids and water tables. Call it a drought — the gold rush slowing not from predation, but from the ecosystem's own carrying capacity asserting itself.

And yet, dear viewer, what is destroyed does not always vanish. In Spain, archaeologists have found Neanderthal wooden tools preserved for tens of thousands of years — not by the wood surviving, but by dissolved minerals seeping in to cast its shape before it rotted away, leaving stone where timber once was. A species' labor, fossilized by accident.

Fittingly, it is also in Spain — at the Museo de Historia de la Computación — that a very different fossil has been assembled: a faithful replica of the Cray-1 supercomputer, built from thirty "obsolete" Mac Minis, its red vinyl bench recreated with evident affection. One generation's apex predator, lovingly taxidermied by its descendants.

The lesson, as always in these matters, is patience. Every dominant organism believes itself permanent. The fossil record — whether stone, silicon, or scorched server rack — tends to disagree.