Edition, Wed, 16 Sep 2026.

Google ships Gemini 3.8 Live, its answer to GPT-Live voice agents

Google DeepMind released two speech-to-speech models on Tuesday, betting that fluid, near-real-time voice is the next frontier product category.

Google DeepMind released two speech-to-speech models on Tuesday, betting that fluid, near-real-time voice is the next frontier product category.

Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking target production voice agents: the Extended Thinking variant reasons and speaks simultaneously and tops Artificial Analysis' Speech to Speech Quality Index with a score of 82.6, while also posting 97.7% on Big Bench Audio, 68.6% on τ-Voice, and 35.1% on Sierra's τ-Voice-banking. The base model switches among 97 languages mid-conversation and runs tool calls in the background while the chat continues, with all generated audio watermarked by SynthID.

The models are live in the Gemini API and AI Studio, in Search Live, and for consumers in Gmail and Keep, with Workspace Docs access for Google AI Pro and Ultra subscribers and private preview in Gemini Enterprise.

Meta One puts a price on social media and AI

Meta launched its new global subscription bundles Monday, pairing app perks with extra AI usage starting at $7.99 a month.

Meta launched its new global subscription bundles Monday, pairing app perks with extra AI usage starting at $7.99 a month.

Meta One spans Facebook, Instagram, WhatsApp, and Meta AI with 50+ features at launch, including 48-hour Instagram stories, a bold follow button, and more media generation powered by Meta's Muse models. Individual bundles run Core at $7.99 per month and Premium at $19.99, a few dollars cheaper than the three standalone Plus subscriptions at $11 per month, while creator and business tiers run from Essential at $14.99 to Max at $499 and add Meta Business Agent capacity on WhatsApp. Meta says the core experience on all its apps stays free and that bundles will expand to Edits and AI glasses over time.

Starship aims for its first true orbit on September 22

SpaceX's 14th test flight is finally trying to put the upper stage into Earth orbit, and it will deploy 26 V3 Starlink satellites in the same flight.

SpaceX's 14th test flight is finally trying to put the upper stage into Earth orbit, and it will deploy 26 V3 Starlink satellites in the same flight.

The 75-minute window opens at 7:15 a.m. CT, and success would make this the first revenue-generating Starship launch for SpaceX, which went public in June in the largest IPO in history. Musk says Falcon 9 and Falcon Heavy will be retired once Starship is flying reliably several times per week. SpaceX is not attempting to catch the Super Heavy booster this time after continued engine-relight problems on Flight 13, and it walked back plans to catch the upper stage as well, calling an explosion too big a setback right now.

AI's top executives are all asking for regulation, and Nvidia's Huang is pushing back

Altman, Amodei, and Hassabis have each publicly called for guardrails in recent days, while Jensen Huang insists safety can be left to the companies building the systems.

Altman, Amodei, and Hassabis have each publicly called for guardrails in recent days, while Jensen Huang insists safety can be left to the companies building the systems.

The shift follows a period of heightened safety anxiety, including a senior Anthropic safety researcher's resignation in protest, in which he cited a 10% chance AI could kill all humans by the end of the decade. Amodei argues "the most effective method of pacing is via regulation that targets all US frontier AI companies," while Zuckerberg favors individual company responsibility, warning that any policy slowing American model releases "even by a month" adds risk to US leadership. Huang told Dreamforce that "safety is an engineering problem, not a legal one," and The Verge notes the industry has been making the regulation pitch for years with few results, under a White House that wants AI barreling full speed ahead.

AWS says some Middle East cloud capacity is gone for good

Wartime damage from Iranian strikes has permanently knocked out part of AWS's Middle East footprint, including one UAE availability zone.

Wartime damage from Iranian strikes has permanently knocked out part of AWS's Middle East footprint, including one UAE availability zone.

The strikes overwhelmed the region's redundancy in Bahrain and left one UAE availability zone inaccessible, AWS says. It is a rare case of a geopolitical conflict permanently reshaping cloud infrastructure, and a reminder for customers running workloads in the region that wartime risk now factors into availability planning.

ServiceNow ships out-of-the-box MCP servers for Claude, Copilot, and Gemini

The Now Platform now exposes governed ITSM, ITOM, CMDB, and SPM tools that major AI clients can call natively.

The Now Platform now exposes governed ITSM, ITOM, CMDB, and SPM tools that major AI clients can call natively.

ServiceNow made its out-of-the-box MCP Servers for ITSM, ITOM, CMDB, and SPM generally available on September 10, 2026, letting supported MCP-compatible clients such as Claude, Copilot, and Gemini invoke purpose-built ServiceNow tools instead of every customer rebuilding the same domain logic on top of generic table access. The company stresses these are outcome-optimized tools rather than API wrappers: in a change management benchmark, the MCP path delivered roughly 52% lower Claude-side cost, 72% faster execution, and about 6.5 times fewer output tokens than direct Table API access.

Governance stays on the Now Platform. Every request runs in the user's context with ACL and role enforcement, while domain logic, workflow steps, validations, and supported actions execute server-side instead of being reconstructed by the model. Supported jobs include incident and change work (pulling related records, finding similar incidents, assessing change risk and conflicts), alerts-to-service-impact analysis, CMDB search with topology and blast-radius context, and portfolio status reviews.

ServiceNow vs. Palantir: two very different bets on enterprise AI

With AI ACV past $1 billion and a 26x forward multiple, one analyst argues ServiceNow screens more attractively than Palantir at 65x sales.

With AI ACV past $1 billion and a 26x forward multiple, one analyst argues ServiceNow screens more attractively than Palantir at 65x sales.

Both companies posted jaw-dropping Q2 numbers, but their models are radically different. ServiceNow reported revenue of $3.99 billion, up 24% year over year, with subscription revenue of $3.877 billion, 123 deals over $1 million in net new ACV (up 40%), and customers with agentic AI in production up 9x over nine months. CEO Bill McDermott frames the company as "the control business," and the AI Control Tower has 500 plus customers live within six months of launch, with a 98% renewal rate. Palantir grew revenue 92.8% to a Rule of 40 score of 155, with U.S. commercial revenue up 149% to $764 million and net dollar retention at 157%.

The valuation gap drives the pick: ServiceNow trades at 26x forward earnings and 9.3x price-to-sales with a $4.2 billion buyback remaining, versus 74x forward earnings and 65x sales for Palantir. The watch items for NOW are whether Q3 subscription growth of roughly 20.5% holds after federal on-premise revenue was pulled forward into Q2, and a GAAP subscription gross margin that has slipped to 73.5% from 80% with the Otto and Moveworks integration to justify it.

SolarWinds' State of ITSM: AI saves hours, but teams say the work still feels heavier

84% of IT professionals say AI met or exceeded ROI expectations, yet 52% report an increased overall workload.

84% of IT professionals say AI met or exceeded ROI expectations, yet 52% report an increased overall workload.

SolarWinds surveyed 844 IT professionals across North America, EMEA, and APAC for its 2026 State of ITSM report, and after an average of 16 months of AI usage the results cut two ways. Respondents reported saving close to nine hours a week: 3.3 hours on detecting and flagging issues, 3.0 hours responding to end-user requests, and 2.9 hours triaging tickets. Average resolution time in earlier generative AI research fell from 27.42 hours to 22.55 hours, a 17.8% reduction, and 86% said AI improved employee productivity. Yet 52% said overall team workload increased, while only 30% saw a decrease.

The report traces the added work to hidden costs rather than the tools themselves: 48% of respondents cited managing and maintaining AI tools and integrations, 47% reviewing and validating AI-generated outputs, and 37% training and fine-tuning models. 83% spend at least three hours a week on AI monitoring and maintenance, and organizations using activity-oriented metrics such as ticket deflection were 2.4 times more likely to report increased workload than those using outcome-oriented ones. Sean Sebring, Solutions Engineer Manager for IT Service Management at SolarWinds, points to the data foundation and integration complexity as the two factors that separate AI success from failure in ITSM.

State and local governments lean on ServiceNow to modernize citizen services

From a Texas health agency to Raleigh, N.C., public sector customers are using the platform to unify contact centers and automate tickets with AI.

From a Texas health agency to Raleigh, N.C., public sector customers are using the platform to unify contact centers and automate tickets with AI.

State Tech Magazine profiles how state and local agencies are using ServiceNow to consolidate constituent-facing operations. The Texas Department of State Health Services folded more than 20 public phone numbers into a single omnichannel contact center tied to the ServiceNow platform, giving the agency what Fusting calls the first 360-degree view of its customer. Generative AI call summarization reduced after-call wrap-up time by 55%, roughly 90 seconds per call, in its pilot and stabilized at about 25% across the broader Tier 1 team. Adoption has been the biggest surprise: the agency now has about 1.6 million customer portal accounts and is adding roughly 20,000 a month.

In Raleigh, N.C., CIO Mark Wittenburg says the city uses ServiceNow for IT asset inventories, incident management, and AI-driven service automation, with generative AI automating ticket summaries and documentation and saving up to two minutes per IT ticket. The city has deployed an AI-powered chatbot and is developing what Wittenburg describes as a "zero-touch help desk agent" that can handle routine service requests under human supervision, with plans to extend the capabilities to citizen services through its Ask Raleigh platform.