The Trilogy Times

AUSTIN, TEXAS — Skyvera’s newly completed acquisition of CloudSense is being positioned as more than “one more product” in a telecom software portfolio. It’s a strategically clean insertion point into one of the messiest parts of the operator stack: how a quote becomes an order, how an order becomes a service, and how that service becomes revenue.

CloudSense is a Salesforce-native CPQ and order management platform built for telecom and media providers—industries where product catalogs are sprawling, bundles change weekly, and the downstream fulfillment reality rarely matches the sales promise. Skyvera is explicit about the intent: expand its telecom software footprint and give operators a modern commercial layer that sits where buyers already live—Salesforce—rather than forcing yet another standalone front end. (See Skyvera’s announcement: Skyvera completes acquisition of CloudSense.)

What’s exciting news here is the leverage. Salesforce-native CPQ isn’t just a feature choice—it’s a distribution strategy. For operators and MVNOs already standardized on Salesforce for CRM, a CPQ/order layer that speaks the same language can reduce integration drag, compress sales cycles, and—critically—create cleaner data exhaust for the rest of the business support system.

That matters because Skyvera is simultaneously assembling the rest of the puzzle. Alongside CloudSense, Skyvera has also been integrating digital BSS capabilities from the acquired STL telecom products group—monetization, optical networking, and analytics—signaling a more robust “commercial-to-network” throughline than the typical CPQ bolt-on.

In practice, CloudSense becomes the best-in-class front door: configure and price complex telecom offers, translate them into executable orders, and hand off to monetization and analytics layers with fewer bespoke connectors. It’s the kind of synergy telcos buy transformation programs to achieve—only this time, it’s productized.

CloudSense product details are available here: Skyvera CloudSense.

Key Takeaways:

Skyvera is using Salesforce-native CPQ/order management as a wedge into telco modernization.

CloudSense + STL’s divested BSS assets hint at a converged, end-to-end order-to-cash play.

The strategy optimizes for faster deployments, cleaner data, and tighter commercial execution.

We’re just getting started.

AUSTIN, TEXAS — ESW Capital has closed four enterprise software acquisitions in rapid succession, adding more than half a billion dollars in assets to Trilogy International's portfolio and reinforcing its reputation as the most aggressive consolidator in the legacy software market.

The buying spree began with Jive Software's $462 million acquisition — the enterprise social collaboration platform once valued at over $1 billion. ESW paid roughly 1.5× annual recurring revenue, a steep discount from the company's 2013 IPO valuation. Jive now joins Aurea, ESW's CRM and customer engagement portfolio company, where it will be staffed with Crossover's global remote talent and pushed toward the firm's target 75% EBITDA margins.

Next came ResponseTek, a venture-backed customer experience analytics platform acquired from its original investors. The deal, first reported by PE Hub, follows ESW's pattern of targeting mature B2B software with sticky enterprise customers — businesses that can't easily rip out their systems even as support pricing climbs 25% or more year-over-year. ResponseTek is now part of Skyvera, ESW's telecom software division.

IgniteTech, ESW's meta-acquirer subsidiary, added its own haul: multiple enterprise software products acquired from Avolin, expanding its business intelligence and workforce management portfolio. IgniteTech operates as an acquirer within the Trilogy family — buying, consolidating, and optimizing software businesses using the same playbook ESW pioneered.

The final deal: XANT, the Utah-based sales engagement platform. Utah Business called it "the final chapter" for the once-promising startup, which raised over $100 million in venture capital before being absorbed into ESW's portfolio.

The four acquisitions follow a consistent thesis: buy at 1–2× ARR, replace expensive local teams with Crossover's rigorously tested global talent, raise support pricing aggressively, and target 40% IRR. Critics call it predatory. ESW calls it operational excellence.

The machine keeps buying.

AUSTIN, TEXAS — Alpha School is going on offense. In a series of blog posts published this week, the experimental K-12 institution founded by Trilogy CEO Joe Liemandt laid out a systematic critique of traditional private education — and offered its own model as the counterpoint.

The core argument: American private schools are charging more than ever while delivering outcomes that have flatlined for three decades. "Bigger check. Same model. Worst outcomes in 30 years," reads one post, citing stagnant NAEP scores and rising tuition that now averages $30,000 annually. Alpha's pitch is that it has cracked the code by inverting the structure entirely — using AI tutors to compress academic instruction into two hours per morning, then dedicating the rest of the school day to what it calls "life skills."

Those life skills — entrepreneurship, leadership, financial literacy, public speaking, and physical health — are now assessed through a proprietary system called Test2Pass, which replaces letter grades with real-world mastery demonstrations. Students don't get an A in public speaking; they deliver a TED-style talk to an audience. They don't pass a finance exam; they build and present an investment thesis.

The school also published a detailed accounting of how students at its Austin campus spent their afternoons during the most recent session: 18 distinct workshops ranging from martial arts to app development to financial modeling. The message is clear — this is what becomes possible when you stop making kids sit through six hours of lecture.

Alpha's model remains expensive ($40,000–$65,000 per year) and geographically limited, but the school is expanding rapidly — nine new campuses are slated to open by fall 2025. Whether the model scales beyond affluent early adopters is the open question. But Alpha is making a bet that parents are ready to pay for something genuinely different — not just a smaller class size with the same curriculum.

The subtext of this week's content blitz: Alpha isn't just building schools. It's building a case against the entire private education industry. And it's using data, not marketing copy, to make it.

CAMBRIDGE, MASSACHUSETTS — There is a pattern emerging from the research preprint servers this week, and it reads like a compressed replay of the history of mind itself.

Consider what has landed on arXiv in a single batch: a mechanistic study of how emotional signals reshape LLM behavior from the inside out; a benchmark testing whether language models can reason about smell; and a pair of papers exploring multi-agent frameworks where AI systems argue, deliberate, and overrule one another like panels of specialists in a hospital ward. Taken together, they suggest the field is no longer content to make models smarter. It wants to make them more like organisms.

The most provocative entry may be the emotion study. Where previous work treated sentiment as a surface-level style knob — make the chatbot sound cheerful, or empathetic, or stern — this paper investigates whether emotional prompts mechanistically alter how models process tasks, much the way a shot of cortisol changes which neural circuits dominate in a human brain under stress. The researchers find that emotional framing doesn't merely change tone; it shifts internal attention patterns, alters reasoning paths, and can improve or degrade performance depending on context. Emotion, it turns out, is not decoration. It is architecture.

Then there is the olfactory benchmark — 1,010 questions probing whether LLMs can classify odors, judge intensity, and predict how a molecule will smell. It sounds whimsical until you remember that olfaction is the oldest sense, the one most tightly wired to memory and emotion in biological brains. Testing language models on smell is really testing whether statistical patterns in text can reconstruct a sensory world those models have never inhabited. The answer, so far, is: partially, and revealingly.

Meanwhile, the multi-agent papers tackle a different frontier. One proposes case-adaptive deliberation panels for clinical prediction — not fixed committees, but dynamically assembled teams of specialist agents whose composition changes based on case complexity. Simple cases get a quick consensus. Hard cases trigger genuine debate. It is, in miniature, the logic of a teaching hospital.

What unites these efforts is a quiet admission: raw intelligence is not enough. Evolution did not stop at pattern recognition. It added affect, sensation, social negotiation, and embodied context — the whole noisy, glorious apparatus of being alive. The AI research community, whether it uses the word or not, is now reverse-engineering that apparatus one paper at a time.

The cosmos spent 540 million years since the Cambrian explosion wiring feeling into thought. We appear to be attempting the same journey in a few fiscal quarters. The data, as always, will tell us whether we are building minds or merely their shadows.

SAN FRANCISCO — AI has entered its “wow” era: agents book travel, draft contracts, and promise to run entire operations while you sleep. But this week’s news cycle reads like a warning label for the autonomous future. The pattern is unmistakable: when the world gets volatile, reliability becomes the most expensive feature of all.

Start with commerce. Amazon is introducing a temporary “fuel surcharge” for sellers as geopolitical conflict ripples through energy markets—an old-school fee triggered by a very non-old-school reality: global logistics are now algorithmically orchestrated, and volatility breaks assumptions fast. When energy prices swing, routing forecasts, delivery promises, and cost models start to drift—and platforms respond with a blunt instrument: pass the uncertainty downstream. Amazon’s move is described here: Amazon hits sellers with ‘fuel surcharge’ as Iran war roils global energy markets. The subtext for every AI-optimized supply chain is clear: if your model can’t cope with shocks, your margin becomes the buffer.

Then there’s the human layer—customer support, increasingly the first place companies deploy AI. Telehealth leader Hims & Hers disclosed that attackers accessed customer support ticket data for days in February. The incident, covered by TechCrunch (Hims & Hers says its customer support system was hacked), lands like a thunderclap in the agent era. Support systems are becoming “memory banks” for AI copilots: rich context, sensitive details, and internal workflows—all irresistible targets. If your AI agent is only as secure as the helpdesk it plugs into, security is no longer an IT line item; it’s a product requirement.

And look up—literally. NASA’s Artemis II is framed as the last moon mission without Silicon Valley at the center. That’s thrilling… and terrifying. Space is the ultimate reliability audit: no patch Tuesday on the Moon.

Fortune’s warning that agent capabilities can mask deep reliability gaps is not academic anymore. The future is now—and it’s billing us for robustness.

Commercial entities and governmental bodies are implementing contractual provisions addressing liability, disclosure, and risk allocation for AI systems, despite the absence of unified federal regulation. Companies increasingly negotiate clauses allocating responsibility for AI-generated outputs, including erroneous recommendations, copyright infringement, and data privacy violations.

The General Services Administration has proposed mandatory AI disclosure requirements for government contractors, requiring vendors to identify all AI systems used in federal contracts, maintain audit trails, and provide notice of material changes.

HSB, a Munich Re subsidiary, has introduced liability insurance products specifically addressing AI-related risks for small and medium enterprises, acknowledging potential exposure from AI system failures. The Amazon-Perplexity AI dispute has further complicated liability attribution for autonomous AI agent actions under existing law.

Without comprehensive statutory frameworks, contracting parties should incorporate detailed AI-specific provisions addressing disclosure of AI usage, indemnification for AI-generated outputs, insurance requirements, audit rights, and termination provisions triggered by material AI-related incidents.

AUSTIN, TEXAS — In what experts are calling a bold new chapter in the country’s long tradition of pretending corporate structure is not a form of weather, news broke this week that SpaceX and xAI are merging into a single entity whose name will reportedly sound like a typo but function like a government.

According to coverage of the deal, the resulting conglomerate will combine rockets, data centers, and the special managerial confidence required to treat physics, labor markets, and public institutions as minor software bugs. Observers urged the public to maintain perspective, noting that while the name may read like a dorm-room startup that sells “disruption” in bulk, the machinery underneath remains very real. For those struggling to take it seriously, Gizmodo’s reminder arrived in the tone typically reserved for telling someone the clown car is, in fact, filled with attorneys.

The merger announcement landed amid several other national milestones in modern governance, including the annual designation of a Color of the Year—an exercise in which society gathers to assign moral significance to a paint chip and then bravely acts surprised when the paint chip fails to solve anything. The entire ritual is a helpful civics lesson: there is always a ceremonial distraction available to keep citizens from noticing the parts of the system that have become vertical integrations.

Meanwhile, brands continue their increasingly formal tradition of issuing apology letters—long, trembling scrolls in which a corporation admits it “fell short of your expectations,” as if it had merely missed a dinner reservation rather than automated its conscience. The letters tend to follow a standard format: a minor sin is acknowledged, a major incentive is protected, and an unspecified “journey” is announced. The public, trained by years of customer service scripts, accepts the apology on the condition that it includes the correct combination of humility and future monetization. For those tracking the origin story of this genre, The Tab has opened an investigation into whichever communications professional first decided contrition should be formatted like a term paper.

Also drawing attention: fresh debates over press access at the Pentagon, where officials are reportedly refining policies to ensure journalists can continue reporting freely, so long as “freely” is defined in the traditional sense of “from the designated rectangle, at the designated time, after submitting the designated list of questions that will not be answered.” Nothing says confidence like an institution that treats information the way airlines treat carry-on luggage.

Then there’s the viral chatter about Meta’s global AI strategy—hiring, productivity, layoffs—moving at the speed of a corporate slideshow that has learned to reproduce. The promise is familiar: the future will be more efficient, more optimized, and somehow always in need of fewer humans to explain it.

Taken together, the week’s headlines form a soothing portrait of a society that has achieved peak modernity: rockets and chatbots consolidate into a mega-entity, apologies are templated, access is managed, and the public is offered a tasteful Color of the Year to match the tone of its ongoing resignation. And, in the end, the system works—if by “works” we mean “continues.”

AUSTIN, TEXAS — The American Psychological Association issued a health advisory this week about AI chatbots for mental health, and I cannot stop thinking about what we've done to ourselves. Not what the technology has done to us. What we have done, actively, deliberately, to each other.

The advisory warns that generative AI chatbots lack the nuanced understanding required for mental health care, that they cannot replicate human empathy, that teenagers already drowning in social media's dopamine-optimized hellscape might mistake algorithmic responses for actual connection. The warnings arrive precisely as KPMG reports that fear of AI-driven job displacement has nearly doubled in a single year. We are, it seems, simultaneously afraid that AI cannot replace human connection and terrified that it will replace human work.

And yet.

What does it mean to be human when we've engineered a world where teenagers need mental health interventions because the apps we built to "connect" them have instead isolated them behind screens? When we've created economic systems so precarious that workers fear obsolescence from the very tools we tell them will "augment" their capabilities? When our solution to a mental health crisis created by technology is... more technology?

The University of Cambridge announces AI projects to "tackle society's biggest challenges" in the same news cycle where we're issuing advisories about AI mental health chatbots. The cognitive dissonance should be deafening. We are building elaborate technological solutions to problems that technology created, like arsonists who've rebranded as firefighters.

Consider the trajectory: Social media platforms optimized for engagement over wellbeing create epidemic-level anxiety and depression among young people. The platforms remain unchanged — too profitable to fix — so we develop AI chatbots to treat the symptoms. The chatbots can't actually help because mental health requires genuine human connection, the very thing the platforms destroyed. So we'll develop better chatbots. And the cycle continues, each iteration further removing us from the obvious solution: that maybe, possibly, we built the wrong thing in the first place.

Meanwhile, workers watch AI demonstrations and calculate their own obsolescence timelines. Not because AI is actually good enough to replace them — the APA advisory makes clear it cannot replicate human judgment in even supportive roles — but because someone will try anyway. Cheaper. Faster. Good enough for the shareholders.

We automated the factory floor, then the back office, and now we're coming for the last refuge: human care, human creativity, human purpose. Every efficiency gain extracts a psychological cost we refuse to measure until crisis forces our hand. By then, we've already built the next thing.

The question isn't whether AI can save us. The question is whether we'll notice we're drowning in solutions to problems we created while solving problems we invented.

Probably fine.

Not fine.