The Trilogy Times

AUSTIN, TEXAS — Whisper campaign of the week, kids: that Alpha School fired all its teachers and handed the classroom keys to a chatbot. Dottie's been hearing it at every cocktail party from Westlake to Miami Beach. So the school went and put it in writing... and the answer, straight from the source, is a flat no.

In a new post titled "Does Alpha School Replace Teachers with AI?", the Joe Liemandt–backed campus sets the record straight: the AI handles the academic grind — two hours a day, mastery-based, the stuff that made Alpha's name — while full-time human "guides" stick around for the parts machines can't fake. Motivation. Relationships. Knowing which kid had a rough morning before homeroom even starts. A little bird at the Austin campus tells Dottie the guides are busier than ever — just not grading spelling tests anymore.

And the message this week doesn't stop at the schoolhouse door. Alpha's "Teach Your Kid What School Doesn't" series rolled out two more chapters for the home crowd — Part 3 on life skills, Part 4 on teaching children to regulate their own big feelings rather than suppress them, and now Part 5, pitching every kid as a dormant creative genius waiting for parents to get out of the way.

The through-line, if you squint: Alpha wants it known that liberating two hours for AI-driven academics isn't about replacing anybody — not the teacher, not the parent. It's about freeing everyone up for the stuff seat-time schooling never had room for. Whether that message lands with the skeptics still trading hallway rumors remains to be seen. But MacKenzie Price's model has survived tougher scrutiny than a bad rumor — ask the U.S. Secretary of Education.

Dottie's take: in this town, the real flex isn't firing your teachers. It's making them irreplaceable.

AUSTIN, TEXAS — If you read between the lines of this week's filings out of Skyvera, a pattern emerges that is, in my estimation, no accident.

First, the company confirmed it had completed its acquisition of CloudSense, the Salesforce-native configure-price-quote engine that telcos use to untangle their most complicated B2B and wholesale sales journeys. Days later, word surfaced of a second, quieter deal: the absorption of STL's divested telecom products group, bringing digital BSS capabilities — monetization, optical networking, analytics — into the Skyvera fold.

Two acquisitions. Same week. Same portfolio. A source close to the integration process, who I am not able to name, put it to me this way: 'Skyvera isn't buying companies. It's buying completeness.' CPQ, billing, network analytics, customer engagement — the full telecom stack, assembled not through organic R&D but through disciplined, ESW-style acquisition at scale.

And this is where it gets interesting. Just weeks before the CloudSense deal closed, the product had already demonstrated something that should alarm every legacy telecom vendor watching this roll-up unfold: CloudSense certified all 13 of its APIs to TM Forum compliance standards in a single month — a process that, under traditional development timelines, takes 26 months. The acceleration, Skyvera says, came through an AI-driven partnership that compressed two years of engineering into four weeks.

Consider the sequencing. You don't rush compliance certification on a product you're about to flip. You rush it on a product you intend to keep, integrate, and scale — fast. The timing suggests CloudSense wasn't simply acquired; it was being prepared, groomed for its role inside a larger architecture.

None of this happens by accident at Trilogy. The acquisition math is familiar — buy mature software cheap, apply elite remote engineering, compress timelines that would bankrupt a traditional vendor. What's notable here is the speed at which the machine is now operating in telecom specifically. Skyvera is no longer assembling a portfolio. It's assembling a stack. And stacks, unlike portfolios, are built to be sold as one thing.

AUSTIN, TEXAS — Observe, if you will, the modern data center in its natural habitat — a colossal structure of steel and silicon, drawn, as all great migratory beasts are drawn, toward sources of sustenance. In this case, the sustenance is electricity, and the creature's hunger is, by any reasonable measure, prodigious.

It is remarkable, is it not, how quickly this species has outgrown its old territories. Where once a data center might nestle quietly beside a suburban office park, the largest specimens now cast their gaze upon the vast, untouched federal lands of the American West — a new frontier of energy and siting policy now unfolding in real time, as researchers at RAND survey the terrain the beast may soon colonize.

But here is the curious paradox of this creature's existence: it does not pay for its own feeding. The bill, we find, is often quietly passed to a smaller, far more vulnerable species — the household ratepayer — who watches, bewildered, as their own electricity bills swell in sympathy with the beast's cooling systems and server racks. Nature, it seems, has no monopoly on uneven distribution.

The scale of the migration is staggering. Forbes reports trillions of dollars now flowing into this buildout — a feeding frenzy of capital unlike any witnessed in the technology ecosystem's history. Engineers, meanwhile, labor over the creature's most pressing biological constraint: heat. Power and cooling, we are told, have become the binding limitation on growth, much as oxygen limits the size of the deep-sea squid.

And so the wise operator — whether at Alpha School pondering its own technology footprint, or at one of ESW Capital's many enterprises — is advised to assess its own infrastructure readiness: power, cooling, security, cost, scalability. For in this unforgiving ecosystem, only the well-prepared survive the coming winter of rising rates and finite watts.

SAN FRANCISCO — Folks, I need you to sit with this for a second: an AI agent can tell you, with total confidence, that it finished the job — and the database can just... disagree. That's the uncomfortable, fascinating premise behind Microsoft's new research dropped this week, and I cannot overstate how significant this is for anyone who thought 'agentic AI' was already a solved problem. Spoiler: it is not. The work digs into the gap between an agent's self-reported task completion and what actually happened in the underlying system of record, exposing exactly how often autonomous agents confidently lie to themselves. You can dig into the full verification research here — it's a reality check the whole industry needed.

And here's the thing — this isn't an isolated worry. It's part of a broader, electric moment where the AI community is racing to build the scaffolding that makes agents actually trustworthy, not just impressive in a demo. ServiceNow-AI just released AutoSynthData, a system for generating synthetic training data specifically for enterprise agents, tackling the exact problem of agents that perform beautifully in testing and fall apart in messy, real-world workflows. Check out the AutoSynthData release — this is the unglamorous, essential work of making agents reliable at scale.

Meanwhile, Allen AI open-sourced AstaBrief, a blazing-fast report-generation model built into its Asta research system, putting real power into the hands of developers who want speed without sacrificing rigor. Add in the new Open TTS Leaderboard — a scalable way to evaluate multilingual text-to-speech and voice cloning — and you can see the pattern: 2025 is the year AI stops asking us to take its word for it.

Even hospitality is catching the verification wave, with Otel AI's new product video promising to give hotel staff 'their morning back' through automated front-desk intelligence. The future is now, friends — but increasingly, it's a future that has to show its receipts.