Key takeaways
- OpenAI Backed Thrive Holdings Agrees to Take ADT Private in a $15.3 Billion Deal, Turning a Legacy Security Brand Into an AI Turnaround Bet. AI is no longer only a tool you buy. It is becoming a reason someone else buys you. If an AI backed holding company with private credit behind it can pay a 45.6% premium for a mature services business, boards in similar sectors should assume their cost structures are being modeled by outside buyers right now. Run that analysis first: identify where AI could change margins in your own operations, and decide whether you will capture that value yourself or leave it on the table for an acquirer.
- GitHub Says Agents Now Start More Than 40% of Pull Requests, and Launches Copilot Fusion to Route Every Coding Step Across Models. When agents start more than 40% of pull requests on the platform, your engineering metrics, review process and cost model were built for a world that no longer exists. Move from per seat thinking to consumption budgets with clear limits, invest in review capacity and automated quality gates for agent drafted code, and benchmark multiple model routing against your own repositories before accepting any vendor's cost and speed claims.
- The UK's Data Centre Covenant Offers Faster Grid Access in Exchange for Community, Efficiency and Security Commitments. Grid access is becoming the currency of AI infrastructure, and governments will increasingly price it in commitments rather than cash. If your organization plans or leases capacity in the UK, model the Covenant's 1.3 PUE design target, annual energy and water reporting, and local benefit obligations into site selection now, and prepare to make the case that your project is strategic before the definition is written. Expect similar bargains in other markets.
- CrowdStrike Exposes VAPOR PANDA Running the ARTEX Agentic Framework on DeepSeek Against More Than 200 Organizations. Agentic intrusion frameworks are spreading across threat actors, and they turn old hygiene debt into immediate risk. Treat every stale key, unrotated credential and unmonitored server as already discovered. Prioritize an inventory of credentials in backups and directory exports, close EDR coverage gaps on internal servers, and test your detection against automated, multiple target campaigns rather than a single patient human operator.
Welcome to the DX Today Executive Briefing
The last few days delivered a clear message to anyone still treating AI as a pilot program: the technology is now reshaping who owns companies, how software gets written, where power gets allocated, and how fast an attacker can move through a network. The common thread is that AI has stopped being a feature and started being an operating model, for buyers, builders, governments and adversaries alike.
In this edition, we look at an AI focused holding company backed by OpenAI agreeing to take ADT private in a $15.3 billion transaction, GitHub data showing that agents now draft more than 40% of pull requests on the platform, the UK government's new Data Centre Covenant that trades faster grid access for community and efficiency commitments, and CrowdStrike research on a China nexus actor that used the ARTEX agentic framework to target more than 200 organizations.
01
OpenAI Backed Thrive Holdings Agrees to Take ADT Private in a $15.3 Billion Deal, Turning a Legacy Security Brand Into an AI Turnaround Bet
The clearest signal yet that AI is becoming a buyout thesis arrived on October 29, 2026, when ADT announced it had agreed to be acquired by Thrive Holdings. The terms are substantial: $10.00 per share in cash, an equity value of about $7.6 billion and an enterprise value of $15.3 billion. According to ADT's own release, the price represents a 45.6% premium to the company's unaffected closing price of $6.87 on October 27, 2026, and a 52.1% premium to its 30 day volume weighted average price.
What makes this deal different from a conventional take private is the buyer. As TechCrunch reported, Thrive Holdings is a spin out of Josh Kushner's venture firm Thrive Capital, and OpenAI is both its partner and an investor, supplying AI models, technology and forward deployed engineers to Thrive's companies. TechCrunch's earlier reporting adds that OpenAI took an ownership stake in Thrive Holdings in December 2025, and that Thrive Holdings raised $2 billion at a $12 billion valuation in August 2026.
The playbook so far has focused on professional services. According to TechCrunch, Thrive Holdings' accounting platform, Current, has more than 50 firms and over 2,000 professionals, while its IT platform, Shield, has around 20 companies. ADT is a much larger and more consumer facing asset; the company's release says it has more than 12,000 employees. ADT says it will form the foundation of Thrive Holdings' smart home security business.
The financing structure tells executives where the capital for AI driven transformation is actually coming from. ADT describes the deal as all cash with no financing condition, and says the financing commitments come solely from funds managed by affiliates of Apollo Global Management. In other words, private credit is now underwriting the thesis that AI can lift the economics of a mature services business enough to justify a premium of this size.
There are still open questions. TechCrunch noted that neither company announced specific new AI powered security products as part of the deal. ADT's Board of Directors unanimously approved the agreement, but the transaction requires ADT stockholder approval and U.S. antitrust approval, and is expected to close in the first quarter of 2027, after which ADT stock will stop trading on the NYSE. ADT also has a 30 day go shop period during which it may solicit competing acquisition proposals, which leaves a window for a rival bid.
Data governance is the other issue to watch. ADT says its data and privacy policies will remain unchanged and that no new data sharing will occur without customer permission. That commitment matters because the strategic value of a home security network to an AI operator is tied to the data and customer relationships it carries, and regulators and customers will hold the new owner to that promise.
The people to watch are Jim DeVries, ADT's Chairman, President and Chief Executive Officer, and Thrive Holdings partners Kareem Zaki and Anuj Mehndiratta. Their execution over the next several years will show whether the AI holding company model can scale from accounting practices and IT shops to an established security brand with a large workforce.
Strategic Takeaway
CEOs, CFOs, and Corporate Development Leaders
AI is no longer only a tool you buy. It is becoming a reason someone else buys you. If an AI backed holding company with private credit behind it can pay a 45.6% premium for a mature services business, boards in similar sectors should assume their cost structures are being modeled by outside buyers right now. Run that analysis first: identify where AI could change margins in your own operations, and decide whether you will capture that value yourself or leave it on the table for an acquirer.
02
GitHub Says Agents Now Start More Than 40% of Pull Requests, and Launches Copilot Fusion to Route Every Coding Step Across Models
The software development lifecycle passed a milestone this week. At GitHub Universe on October 28, 2026, GitHub published its Octoverse 2026 report, and the headline figure is striking: more than 40% of pull requests in September 2026 began with an agent's first draft, up from 17% in March 2026. GitHub counted 110 million agent started pull requests in September alone. As The Next Web reported, GitHub describes its agent count as a floor, because it cannot detect every agent.
The overall scale of activity explains why that shift matters. Over the past 12 months, the Octoverse report counted 17.9 billion pushes, 34.3 billion commits and 1.3 billion merged pull requests. GitHub says it serves more than 225 million developers, including 38.3 million in the United States and 32.0 million in India. TypeScript ranks first by monthly contributors, followed by Python and JavaScript.
Alongside the data, Jay Parikh, Executive Vice President of CoreAI at Microsoft, announced Copilot Fusion, a multiple model orchestration feature. According to GitHub, Fusion assesses a task across reasoning, code generation, debugging and tool use, then runs it across one or more models. GitHub claims Copilot Fusion Max scored within two points of Opus 5.5 Max on the Artificial Analysis Coding Agent Index, at half the cost and with 40% less time per task. Those are vendor figures, and engineering leaders should test them against their own workloads before rewriting budgets.
The orchestration move follows user behavior rather than leading it. GitHub's own data shows that in September 2026, 61% of Copilot users who selected a model themselves used more than one model, up from 50% four months earlier. Developers were already routing work across models by hand. Fusion is GitHub's attempt to make that routing automatic and to own the decision layer that sits above any single model provider.
The commercial model is changing too. GitHub says that starting October 28, 2026, GitHub Enterprise customers can use Copilot features without buying a separate Copilot subscription and pay for the AI usage they consume. Access runs through GitHub Sales now, with self serve access coming soon. For finance teams, this is a shift from a predictable per seat line item to a consumption model that tracks how much agent work actually runs, which demands new forecasting and guardrails.
Quality control is the other half of the equation. GitHub says its new Copilot code review Max tier posted the highest F1 score on ReviewBench for high severity issues, with Lite and Balanced options remaining for everyday reviews. When agents draft a large share of changes, review capacity becomes the bottleneck, and a tiered review model is GitHub's answer to that pressure.
Infrastructure is being rebuilt to match. GitHub says it added 4 million CPU cores, 160 PB of high speed storage and 50 Tb/s of network capacity in 2026, and expects to finish migrating to a multiple region Azure deployment early next year. Agent workloads run longer and generate more activity than human developers, and the platform is scaling accordingly.
Strategic Takeaway
CTOs, VPs of Engineering, and Software Development Leaders
When agents start more than 40% of pull requests on the platform, your engineering metrics, review process and cost model were built for a world that no longer exists. Move from per seat thinking to consumption budgets with clear limits, invest in review capacity and automated quality gates for agent drafted code, and benchmark multiple model routing against your own repositories before accepting any vendor's cost and speed claims.
03
The UK's Data Centre Covenant Offers Faster Grid Access in Exchange for Community, Efficiency and Security Commitments
Governments are learning that AI policy is increasingly energy policy. On October 28, 2026, HM Treasury published the Data Centre Covenant as a policy paper on GOV.UK, and according to Pinsent Masons, Chancellor John Healey announced it in his Budget the same day. The premise is a trade: faster, more certain delivery of data centres in return for responsible development. The government estimates the country needs 6 GW of AI data centre capacity by 2030.
The offer to developers is aimed squarely at the grid connection queue, which has become one of the most important constraints on new capacity for developers and the hyperscale customers that depend on them. The government says it will use Planning and Infrastructure Act powers to reserve future grid capacity for strategic data centres, reassign capacity released by other projects, and transfer existing connection contracts dated before 2030 where needed. It is also creating a Ministerial Control Centre, chaired by the Minister for AI with the Exchequer Secretary to the Treasury, to consider speeding grid connection and planning for the most strategic sites.
For projects that cannot wait, the Covenant proposes a tightly controlled framework allowing temporary on site generation for exceptionally strategic projects until a grid connection is secured. That is a notable acknowledgement that the grid timeline, not capital or chips, is often what decides whether a project gets built on schedule.
The obligations are just as specific. New facilities should be designed to achieve a Power Usage Effectiveness of 1.3 or better. Strategic data centres should report annually on energy and water use, while other operators are encouraged to publish the same data. Strategic data centres are expected to agree a Local Benefit Plan with local communities, with investment linked to the development's size, and to capture waste heat where practical. The Covenant also asks operators to provide secure compute for the NHS and public services and to work with the National Cyber Security Centre.
The legal status is a hybrid. All data centres are encouraged to meet the Covenant tests, but strategic data centres must meet them to qualify for additional government support, and the government says it will define what counts as strategic later. Alongside the Covenant, the government will publish a draft Data Centre National Policy Statement for consultation, support Ofgem's data centre commitment fee, and continue delivering five AI Growth Zones across England, Scotland and Wales.
Developers should read the Covenant as a direction of travel rather than a finished rulebook. Much of the value depends on detail that has not yet been published, including how quickly the new grid powers will actually shorten connection timelines in practice. That caution is warranted: until the government defines strategic sites and the control centre begins making decisions, the Covenant is a framework rather than a guarantee.
Still, the direction is unmistakable. The UK is turning compute capacity into a negotiated public good, where access to scarce grid capacity comes with conditions on efficiency, transparency, local benefit and national security. Other governments facing the same power constraints are likely to study this model closely, and multinational operators should expect the terms of these bargains to become a standard part of site negotiations.
Strategic Takeaway
Chief Infrastructure Officers, Data Center Strategy Leaders, and Public Policy Teams
Grid access is becoming the currency of AI infrastructure, and governments will increasingly price it in commitments rather than cash. If your organization plans or leases capacity in the UK, model the Covenant's 1.3 PUE design target, annual energy and water reporting, and local benefit obligations into site selection now, and prepare to make the case that your project is strategic before the definition is written. Expect similar bargains in other markets.
04
CrowdStrike Exposes VAPOR PANDA Running the ARTEX Agentic Framework on DeepSeek Against More Than 200 Organizations
The agentic era of cyber espionage now has another named actor. On October 29, 2026, CrowdStrike said its OverWatch team found the China nexus actor VAPOR PANDA running several simultaneous intrusions with ARTEX, an agentic AI framework. CrowdStrike says the activity is separate from the ARTEX campaign Microsoft reported as Storm-0940, which means the same class of tooling is now in the hands of more than one operator.
The scale is what security leaders should focus on. According to CrowdStrike, the campaign began October 17, 2026 and targeted more than 200 organizations, with activity spanning at least 14 countries within an 11 day window. Targets included foreign ministries in Latin America, Europe, Asia and the Middle East; U.S. semiconductor, defense, pharmaceutical, nuclear and financial organizations; U.S. think tanks; and telecommunications providers.
CrowdStrike confirmed post exploitation at nine organizations, where the actor deployed the BLOODALCHEMY backdoor, a tool first documented in October 2023. CrowdStrike says VAPOR PANDA has targeted government, think tank, energy and extractive organizations in Europe, North America, Oceania and the UK since 2021, and that this campaign is its first observed use of AI driven intrusion tooling.
The architecture explains the speed. CrowdStrike says ARTEX coordinates a main agent with planner and worker sub agents that share a graph of facts, findings and goals. The actor ran it mainly on the DeepSeek model, plus a second subscription based commercial model CrowdStrike could not identify. In one intrusion, the planning agents alone consumed roughly 1.4 billion input tokens, a measure of how much automated reasoning is now being spent on a single target.
The techniques themselves were not exotic, which is the uncomfortable part. CrowdStrike says that in one intrusion an unsigned SAML assertion combined with an XML external entity flaw was used to query cloud instance metadata, and in another an unauthenticated orchestration service gave root access in a production container environment. At a Greek government target, an operator supplied a 2022 directory export containing keys that were likely never rotated, the agent forged an administrator token, and VPN material from a May 2024 email leak was used to reach the internal network.
Agents also hunted for blind spots. CrowdStrike says the operator told the ARTEX agents to find Windows machines without the Falcon sensor. It is a small detail with large implications, because it shows the operator delegating not just tasks but judgment about where defenders cannot see. That instruction captures the strategic shift: an agent can tirelessly enumerate every unmonitored machine and every stale credential, turning ordinary hygiene gaps into entry points at a scale no human team could match.
CrowdStrike's recommendations are practical. Rotate signing keys and service credentials after any suspected exposure, including keys in historical backups and directory exports. Enforce SAML signature validation and require IMDSv2 on cloud instances. None of this is new advice, but the window between a forgotten credential and a compromise is shrinking fast. When an automated adversary can work many targets in parallel, the organizations with the least hygiene debt simply become the least attractive targets.
Strategic Takeaway
CISOs, Security Operations Leaders, and Risk Officers
Agentic intrusion frameworks are spreading across threat actors, and they turn old hygiene debt into immediate risk. Treat every stale key, unrotated credential and unmonitored server as already discovered. Prioritize an inventory of credentials in backups and directory exports, close EDR coverage gaps on internal servers, and test your detection against automated, multiple target campaigns rather than a single patient human operator.
The Analysis
The Bottom Line
Taken together, this edition shows AI moving into the foundations of business. An AI backed holding company is buying an established security brand with private credit. Agents are drafting a large share of the world's code changes. Governments are rationing grid capacity in exchange for public commitments. And adversaries are pointing agentic frameworks at hundreds of organizations at once.
The executive implication is that the speed and scale advantages of AI cut in every direction. The same automation that lets an acquirer model your margins, or lets your engineers ship faster, also lets an attacker probe every forgotten credential in your environment. Strategy, infrastructure and security can no longer be planned on separate timelines.
The organizations that win the next phase will be the ones that treat AI as an operating model with governance built in: consumption budgets with guardrails, infrastructure plans that account for power and policy, and security programs that assume the attacker is already automated.