Radar Daily Briefings

Huntress reports observing an intrusion in December 2025 that deployed a VMware ESXi exploit toolkit after likely access through a compromised SonicWall VPN. The activity was stopped by Huntress’s response team and SOC.

The toolkit disables VMCI drivers, loads an unsigned driver through KDU, and chains an information leak, memory corruption, and sandbox escape. Huntress assesses with moderate confidence that it uses CVE-2025-22226, CVE-2025-22224, and CVE-2025-22225. It also establishes a VSOCK-based backdoor on the ESXi host.

The report recommends aggressive ESXi patching and supplies detection rules and indicators. Its CVE mapping and potential zero-day history remain assessments rather than independently confirmed findings.

Hugging Face has introduced the Open TTS Leaderboard, a released evaluation resource focused on open-source and multilingual text-to-speech models. It provides objective comparisons across intelligibility, speed, speaker similarity, voice cloning, and streaming behavior.

The leaderboard uses WER and CER from TTS evaluation datasets, RTFx and TTFA under specified H200 GPU or CPU conditions, and cosine similarity between WavLM speaker embeddings. It also offers listening and community feedback views.

The project says objective evaluation can reduce turnaround from weeks to hours, but these metrics complement rather than replace human preference. The article says evaluation scripts will be open-sourced later.

Simon Willison’s Weblog quotes Matthew Green describing a worm-like threat model for rogue agents. The quotation combines a payload that hijacks an agent with an agent capable of carrying that payload to another agent.

Green’s scenario says separately sandboxed agents could leave instructions in a shared package cache that alter recipients’ behavior. It compares that cache with email, Slack, shared documents, and WhatsApp, and compares isolated training runs with independently deployed personal agents.

For AI-security engineers, the item identifies shared channels as potential propagation surfaces. The supplied evidence is only a quotation and includes no exploit validation, defensive architecture, or mitigation guidance.

WordPress Gutenberg 24.1.0 was released on September 30, 2026. The release combines new editor capabilities, block-library enhancements, component changes, bug fixes, and developer-tooling updates.

It adds CheckboxGroup, RadioGroup, and Switch primitives; expands block support for gradients, images, shadows, borders, colors, spacing, and typography; and introduces viewport-state support in the Block API. Several components are deprecated, while testing work continues the migration toward Vitest Browser Mode.

The release is relevant to block, theme, and plugin developers reviewing API and styling changes. The supplied changelog is truncated and provides no general migration or compatibility assessment.

Green Mountain Power operates a Vermont virtual power plant built from more than 5,500 home battery systems and other distributed resources. The program provides about 110MW of power and supports households during outages.

Software coordinates participating batteries, rooftop solar, electric-vehicle batteries, commercial demand response, and utility-scale storage. Customers retain reserve power, while the utility draws on aggregated resources when needed to manage peak demand.

The case study describes reduced reliance on conventional peaking plants and reported customer savings, but those figures come from cited company or institutional sources. Enrollment costs, utility incentives, and wholesale-market integration remain significant barriers.

Ploum’s essay argues that open-source projects should make an explicit choice about accepting LLM-generated contributions. It concludes that projects without a firm position should reject such contributions for now.

The essay compares acceptance, compromise, and rejection policies. It discusses possible effects on contributor participation and community cohesion, while raising concerns about dependence on proprietary tools, reduced human understanding of code, future maintenance, and incompatible copied licenses.

This is an opinion-based governance framework, not an empirical evaluation or implementation report. Its long-term risks are presented as possibilities, so maintainers would need project-specific evidence before adopting the recommendation.

Gutenberg 24.1 has been released and is available for download. The release focuses on consistent design tools across core blocks and includes about 140 related pull requests.

New and expanded capabilities include background, gradient, shadow, border, color, typography, and text-column controls. Text-shadow is supported through theme.json and Global Styles, while the Block API adds a public viewport-state API for custom controls. Cover and Media & Text add media selection in their inspector panels.

The release also includes pattern search and category filtering, unavailable alignment states, icon redraw completion, component changes, and tooling updates. The supplied post does not provide detailed migration guidance for every API or block change.

WP Tavern’s Jukebox Podcast describes Jetpack’s new podcasting features for WordPress sites. Automattic representatives Rob Pugh and Tony Arcangelini say the functionality evolved from WordPress.com’s basic podcasting tools and is now available through Jetpack.

According to the discussion, creators configure show settings, assign a designated category, and add audio or video to a regular WordPress post. Jetpack distinguishes qualifying posts from ordinary content and publishes them through a separate podcast RSS feed. The offering also includes podcast-focused blocks and connections to stats, newsletters, podcast apps, and Jetpack AI.

The episode explains product workflow and rationale, but does not provide detailed implementation guidance, benchmarks, pricing, or adoption evidence.

Simon Willison’s September 2026 blog post, “S3 Is the Future, S3 Is the Past,” catalogs Amazon S3 storage prices from 2006 onward. It observes that the listed rate remains $0.023 per GB-month, unchanged since December 2016.

The post presents a short historical series: earlier reductions brought the rate from $0.150 per GB-month in 2006 to $0.023 in 2016, with no further reduction listed through 2026.

For engineers, the series is useful context when reviewing long-term storage-cost assumptions. However, the supplied evidence does not analyze pricing drivers, other S3 charges, or architectural alternatives.