The Trilogy Times

AUSTIN, TEXAS — The school bell just rang worldwide, kids. Word is Alpha School’s latest move, Alpha Anywhere, takes its AI-powered academics model beyond its brick-and-mortar campuses and beams it straight into the family kitchen, the spare bedroom, the wherever-you-can-focus corner of the house.

Call it homeschool with a pit crew. Call it private school without the car line. Call it Joe Liemandt’s second act getting a passport.

Alpha School, the Austin-born K-12 outfit co-founded by Liemandt and MacKenzie Price, has made its name on a simple, eyebrow-raising promise: core academics in two hours a day, powered by adaptive learning software, with students advancing only after mastery. The school says its students learn 2.3 times faster than national norms and routinely test in the top 1–2% on NWEA MAP Growth assessments. Not exactly bake-sale chatter.

Now comes Alpha Anywhere, the remote version of that model. A little bird from the education wing tells me the pitch is not “more screen time,” but “better screen time” — a distinction Alpha has been hammering in recent posts aimed at parents worried their children are being raised by glowing rectangles.

Here’s the trick: Alpha is not selling passive digital babysitting. The model pairs AI-driven academic work with human coaching, accountability, and the larger Alpha gospel that school should leave room for life skills, entrepreneurship, leadership, athletics, public speaking, coding — the stuff that does not fit neatly into a worksheet.

The timing is rich. The Austin American-Statesman is asking whether Austin tech can follow Trilogy’s example to replenish the talent pool. Alpha’s answer appears to be: why wait for college? Start the talent factory in fifth grade.

Blind item: which legacy school operator looked at Alpha Anywhere this week and suddenly discovered the word “AI” for its fall brochure?

This is classic Trilogy logic in a blazer and sneakers. Automate the repeatable. Reserve humans for judgment, motivation, and taste. Crossover does it for labor. ESW does it for enterprise software. Alpha is doing it to the school day.

Parents may still ask the obvious question: does an AI tutor at the kitchen table replace school, or merely unbundle it? Alpha’s bet is that the old bundle was bloated, expensive, and slow.

And if the numbers hold outside Austin? Then the next hot campus may not be in Miami, New York, or Palo Alto. It may be next to the fridge.

AUSTIN, TEXAS — If you read between the lines of Skyvera's recent acquisition activity, a very deliberate picture begins to emerge. The Trilogy International telecom software unit has completed the acquisition of CloudSense, a Salesforce-native CPQ and order management platform built specifically for telecom and media providers — and this is where it gets interesting.

CloudSense is not a random bolt-on. It slots directly into Skyvera's stated mission of bridging legacy on-premise telecom infrastructure to cloud-native systems. Telcos and media companies have spent decades building configure-price-quote workflows around systems that were never designed for modern commerce. CloudSense, native to Salesforce, offers a way out — and Skyvera now owns that exit ramp.

But the CloudSense deal is only part of the story. Around the same time, Skyvera also absorbed the divested telecom products group from STL, picking up digital BSS capabilities including monetization tools, optical networking functionality, and analytics. A source familiar with the portfolio, who asked not to be named, described the STL acquisition as "the kind of deal that only makes sense if you're building toward something much larger."

Set those two moves alongside Skyvera's existing holdings — Kandy for cloud-based real-time communications, VoltDelta for customer engagement, ResponseTek for experience data, and Mobilogy Now for device lifecycle management — and the pattern becomes impossible to ignore. Skyvera is assembling a full-stack telecom software suite, one divested asset at a time.

Nothing here is accidental. The ESW Capital playbook — acquire at a discount, staff globally through Crossover, push toward 75% EBITDA margins — requires products that are genuinely sticky. In telecom, BSS and CPQ systems are about as sticky as it gets. Carriers don't rip out billing infrastructure on a whim.

The question isn't whether Skyvera is building something consequential. It clearly is. The question is how far this consolidation goes — and who in the legacy telecom software market hasn't yet gotten the call.

KANSAS CITY — In the fluorescent ecosystem of the modern American school, a new sentinel has taken its perch. It does not breathe, does not tire, and does not look away. It watches through cameras, parsing the passing herd of students for the unmistakable outline of a gun.

But in one Missouri school, according to a new lawsuit, the sentinel failed to cry out.

A school shooting survivor has sued an AI gun detection company after its system allegedly did not identify a weapon before shots were fired, a case that now places one of the technology industry’s most delicate creatures under the legal microscope: the safety algorithm. As reported by Ars Technica, the complaint raises a deceptively simple question with profound consequences: how accurate must an AI system be when its purpose is to prevent catastrophe?

Here, the familiar rhythms of enterprise software meet something far more primal. In ordinary habitats, a false negative may mean a missed sales lead, a delayed invoice, a cloud bill left untrimmed. In schools, hospitals, and airports, the same computational lapse can become a wound.

AI detection companies often describe their models as tireless assistants, augmenting human attention across sprawling camera networks. The pitch is beguiling. Human guards are fallible mammals, distracted by fatigue and repetition. Machines, by contrast, are presented as vigilant owls in the rafters, forever scanning. Yet the lawsuit reminds us that these systems are trained organisms of a kind, shaped by their datasets, bounded by their assumptions, and vulnerable to rare events outside their learned experience.

The case may also test the language used to sell such systems. If an AI product is marketed as a protective shield, courts may ask whether customers and students were invited to expect shield-like behavior. If it is merely an advisory tool, the burden may shift back to the humans who chose to rely upon it.

Across the broader technology savannah, this is becoming the central migration of AI: from clever helper to entrusted guardian. Models are moving into finance, medicine, education, and public safety, where failure is no longer an inconvenience but an event with witnesses.

And so the machine sits beneath the legal canopy, its promise and its limits exposed. Not evil. Not alive. But consequential — and, perhaps for the first time, asked to account for the moment it did not see.

CAMBRIDGE, MASSACHUSETTS — A confluence of recently published research — spanning predictive policing, healthcare simulation, and the broader sociotechnical architecture of machine learning systems — has, it could be argued, crystallized what this correspondent would term the 'fairness paradox': the proposition that artificial intelligence systems, designed ostensibly to remove human subjectivity from consequential decisions, may in fact encode, amplify, and institutionalize that very subjectivity at a scale previously unattainable by purely human bureaucratic apparatus (a troubling irony that would not be lost on students of Weber).

The thesis is, on its surface, well-established. Research from the Human Rights Research Center documents with considerable specificity how predictive policing algorithms erode procedural fairness by operationalizing historically biased arrest data as though it constituted objective ground truth — a methodological conflation that, preliminary evidence suggests, disproportionately burdens communities already subject to elevated surveillance intensity.

The antithesis, however, demands equal consideration. Scholars publishing in *Frontiers* advance the position that the formal-mathematical tradition of fairness (demographic parity, equalized odds, calibration) and the sociotechnical tradition (examining power, context, and institutional embedding) are not merely complementary but are, in fact, incommensurable in ways that resist easy synthesis. One cannot, the argument proceeds, simply 'add fairness and stir.'

The synthesis — insofar as one exists — may reside in simulation. A *Nature*-published case study in AI-assisted healthcare attempts, with admirable if incomplete success, to connect algorithmic fairness metrics to actual patient outcome distributions — demonstrating that a system technically satisfying formal fairness criteria may nonetheless produce population-level disparities when embedded within existing clinical workflows (a finding that should, one ventures to suggest, give considerable pause to procurement officers across the health sector).

What emerges from this body of work, taken in aggregate, is not a prescription but a provocation: that the question 'is this AI fair?' is, in point of epistemological fact, unanswerable without first specifying fair to whom, by what measure, and under whose institutional authority — questions that are, it could be argued, irreducibly political rather than technical in nature.