Digital Plumber

Selector Foundry represents a shift from traditional AIOps dashboards to agentic NetOps, where specialized AI agents (Rosetta, Loom, TicketMaster, Herald, Artist, Sculptor) collaborate on incident investigation and remediation within one correlated data model. Artist specifically solves a workflow friction point: in war rooms and post-incident reviews, teams repeatedly request custom dashboard views (one incident, one application path, one device context, executive summaries) that arrive too late or require SQL expertise. Artist accepts requests like "show me edge-rtr-3 dependencies during this incident" and builds an interactive dashboard matching the request, refinable in conversation. The platform consolidates raw telemetry (metrics, logs, events, flows, streaming telemetry) through a harmonized data model supporting cross-domain correlation; Selector claims 1,412 signals across three alert clusters resolve to a single incident with root cause (device reboot) visible in the view. This moves network operations from "detect and search" to "detect and context-render automatically," critical for NOCs handling multivendor, multi-domain environments where alert storms obscure root causes across L1–L7.

OcNOS-DC is a network operating system purpose-built for AI and HPC fabric operations with runtime support for RoCEv2, Priority Flow Control, ECN-based congestion control with DCQCN, dynamic load balancing (DLB) for avoiding hash-based flow collisions, and UEC compliance. Extension to Tomahawk 6 represents significant milestone: Tomahawk 6 delivers 2x throughput jump from Tomahawk 5 (51.2 to 102.4 Tbps), supports native 1.6T ports (up to 64 native 1.6T or 512 × 200G radix for scale-out), and includes options for 200G and 100G PAM4 SerDes depending on copper reach and port density. OcNOS-DC already ships on Tomahawk 4 and Tomahawk 5 in production. Tomahawk 6 support extends OcNOS to 40+ open hardware platforms, giving operators choice in switch ODMs (Micas, Arista, Edge-core, Delta, Wistron) while maintaining single network OS and fabric-control paradigm. Particularly valuable for neocloud providers building GPU data centers at scale who need standardized OS to reduce training and automation burden across heterogeneous hardware deployments.

RPKI-valid routing jumped from 54% to 70.3% in a year, driven largely by China's adoption. This marks significant acceleration in cryptographic routing security deployment. Of all routed IPv4 space not in RPKI-valid routes, nearly 60% is in the United States, and about one in four addresses belongs to the US Department of Defense. Nearly half of all invalid routes reach only a single vantage point; when wrong networks announce addresses, 71% reach only one. This isolation of invalid routes demonstrates ROV's practical effectiveness at blocking hijacks early. However, IPinfo found thousands of records flagged as failed that were present in every independently validated set they checked, with their routes still RPKI-valid, meaning a count of failing records is not a measure of lost protection. Path validation via ASPA is emerging as the next frontier, with deployment accelerating but still requiring broader router and validator support.

The Timor-Leste Internet Exchange Point (TL-IXP) will allow ISPs and other networks to exchange traffic locally rather than routing through international connections. Plans include route servers, DNS root mirrors, monitoring infrastructure, and routing security measures such as RPKI validation. Timor-Leste has made progress in strengthening routing security through RPKI, with approximately 60% of announced address space covered by valid ROAs. ROAs allow network operators to declare which Autonomous Systems are authorized to originate specific IP address prefixes, and Route Origin Validation uses this information to identify and reject invalid route announcements. Over the seven days ending July 10, 2026, APNIC Labs measured invalid-route filtering in Timor-Leste at 34.42%, compared with 14.46% across Southeast Asia and 26.69% globally. While measurements indicate encouraging progress, continued ROA creation and broader implementation of ROV among network operators will be important as Timor-Leste's Internet ecosystem grows and becomes more interconnected.

CVE-2026-0322 affects PA-Series and VM-Series firewalls and Panorama (virtual and M-Series); Cloud NGFW and Prisma Access unaffected. Very limited exploitation observed with three IPs identified in advisory. Fixed releases available for PAN-OS 10.2, 11.1, 11.2, 12.1, 12.2 with 11.2.16 expected October 15. This continues 2026's pattern of firewall management interfaces as primary attack surface, compounding risk for teams still managing unpatched internet-facing PA deployments already exposed to CVE-2026-0257 (GlobalProtect auth bypass), which ransomware groups Qilin and Settra actively exploit.

Ruddarraju, previously serving as OpenAI's CTO of compute since July 2026, has been promoted to lead the overall compute organization, replacing Sachin Katti. Ruddarraju, who ran infrastructure at Elon Musk's xAI until July 2025, takes over from Katti, who had been overseeing infrastructure and led the $30B Georgia data center build, and leaves the role within weeks to stay on as an adviser. OpenAI has appointed Ruddarraju to lead its infrastructure and compute businesses, running its Stargate and other data center efforts, after he joined OpenAI last July as head of compute and infrastructure and was named OpenAI's CTO of Compute this July. Ruddarraju's career spans large-scale infrastructure work, including nearly six years at Robinhood rising to Senior Director of Engineering and Head of Infrastructure, over five years at eBay, and foundational work at Amazon Web Services. The consolidation of compute and infrastructure leadership signals OpenAI's prioritization of infrastructure scale during a critical period of capital expansion and technical execution for frontier AI development.

Acronis has appointed Prakash Panjwani as its new chief executive officer, effective October 15, 2026, replacing Jan-Jaap Jager, who will remain a shareholder. Based in Schaffhausen, Switzerland, Acronis sells software combining cybersecurity, data protection, IT operations, and hybrid infrastructure management, with distribution heavily channel-focused through managed service providers. Panjwani previously served as CEO of WatchGuard Technologies and SafeNet. During his decade at WatchGuard, he oversaw the company's evolution from a firewall-focused business into a broader cybersecurity platform incorporating AI-powered endpoint detection and response (EDR), extended detection and response (XDR), and managed detection and response (MDR). The leadership change comes as Acronis looks to expand its AI-based cybersecurity and data protection business and strengthen its position in the MSP market where service providers use the platform to support over 750,000 businesses globally.

John Capobianco from Itential developed NetClaw over 18 months after recognizing early that MCP was essential for networking operations. While Itential was connecting AI requests to workflows, validation, and approvals, he independently explored MCP for networking—matching the core question of how to make agents useful on real infrastructure. By February 2026, that became NetClaw, an open network engineering agent integrating pyATS (Cisco's gold-standard network testing framework), MCP, network APIs, observability platforms, security constraints, and documentation into a single open-source project. The agent is organized around skills, tools, evidence, and constraints—enabling it to reason about network state deterministically while respecting operational guardrails. This is a hands-on example of operationalizing agentic network operations: grounding agent reasoning in deterministic tools, connecting to management systems through standard protocols, and enforcing operational constraints. The project is publicly available for teams starting their own network agent implementations.

The v0.163.0 release continues OpenTelemetry's incremental standardization of telemetry collection. The addition of release-candidate semantic conventions for host metrics in the hostmetrics receiver marks progress for infrastructure observability within OTel's standards process. This brings parity to application, infrastructure, and business metrics under one specification framework. The Prometheus exporter compatibility layer sees three stabilizations: content negotiation, native histogram mapping, and resource constant labels configuration. No OTLP wire protocol changes or trace semantics shifts occurred, keeping this release low-risk for existing deployments. For platform teams managing collector fleets at scale, this represents incremental progress toward complete semantic convention coverage without breaking existing telemetry pipelines.

Anthropic shipped Claude Haiku 5.5 on October 7 at $0.10/$0.50 per million tokens up to 100K context, positioning it as a low-cost competitor to GPT-6 Luna's sub-$1 tier economics. Google's Gemini 4 Argon reached general availability on October 13 after launching to vetted cyber defenders, now broadly available across paid API tiers and consumer subscriptions. Meta Superintelligence Labs shipped Muse Wonder on October 12 as Meta's first frontier models positioned head-to-head with OpenAI and Anthropic at $4/$20. Mistral Large 4 was previewed October 6 behind a guardrailed Studio with no published price, licence or architecture, and weights due end of month—marking a shift from September's volume releases to gated frontier entries. xAI released Grok Voice 3.0 on October 14. This reflects consolidation: after September's 20+ releases in two weeks, October pivoted to targeted launches with stricter access controls and pricing tiering that requires production cost recalibration.

The survey found that 80% of EMEA executives experienced AI-related incidents reaching their board or C-suite in the past year, with only 40% changing AI governance policy or frameworks in response. This disconnect between incident frequency and governance response reflects a critical maturity gap. 75% of respondents expect AI agents to have some level of autonomy with respect to decisions by 2027, yet only 37% say their governance is designed to scale with agentic AI deployment. Only 37% of respondents have a formal, board-reviewed quantification of the financial risks created by gaps in their AI governance. The research highlights that enterprises are deploying autonomous AI faster than their governance frameworks can adapt, creating exposure to regulatory and operational risk. Supply chain visibility is also weak: 54% of respondents acknowledge gaps or limitations in how their governance framework covers AI systems and models used by key vendors and technology partners.