#Deployment/Engineering

AIComcast and Booz Allen used Claude Mythos Preview to find vulnerabilities that arise from interactions across code, configuration, and deployment rather than single-file bugs. Comcast identified a critical authentication flaw across 258 systems and about 170 million lines of code before any exploitation was observed. Booz Allen reported that one analyst reviewed eight production systems across 138 repositories in twelve days, a review its team estimated would have taken several months without the model.

Why it matters: The case studies show how security teams validate and remediate model-found exploit chains, a workflow relevant to anyone managing large codebases.

AIGoogle has made Gemini Nano Banana 2.1, identified as gemini-nano-banana-2.1, generally available as an image generation and conversational editing model. It improves visual quality, prompt adherence, multi-turn character consistency, and text rendering, and adds panoramic aspect ratios such as 1:4, 4:1, 1:8, and 8:1 at 1K, 2K, and 4K resolutions. The gemini-3.1-flash-image model is deprecated with no shutdown date announced, and developers are told to migrate to the new model.

AIAnthropic is launching an expanded Cyber Verification Program with three access tiers for qualifying security professionals, giving each tier different cyber capabilities and reduced blocking classifiers. On CyScenarioBench, Claude Opus 5.5 was blocked on 46 of 50 trials in the Defense Access tier, while the Red Team Access tier had no blocks and completed 34 of 50 tasks. Existing Project Glasswing members will move to the Specialized Access tier, and data retention is required for enrolled organizations.

Why it matters: The program lays out three verified access tiers with different cyber blocks, and its CyScenarioBench figures show how safeguards change what defenders can do.

AIThariq says planning with HTML is much more token efficient than generating raw HTML. The model does not need to recreate components or logic for common elements such as state machines, diagrams, and code snippets. Background from the quoted post says he is building a Claude Code skill that generates HTML plans, with linting to reduce common failures.

AIMicrosoft has added citation links to Copilot replies in Word, letting users click through to original web pages or internal documents to verify information. The company says the change improves transparency about where Copilot's information comes from. The feature targets AI hallucinations, which are errors or fabricated sources produced by AI tools.

AIGoogle released EmbeddingGemma 2, an open embedding model under the Apache 2.0 license that maps text, code, images, video, and audio into a shared 768-dimensional space. Developers can load a 270M-parameter text and code setup, or add vision and audio encoders up to a 740M-parameter full multimodal model. Matryoshka truncation to 256 or 128 dimensions reduces vector storage, with the guide noting quality losses on image, video, and speech retrieval at lower dimensions.

Why it matters: The guide gives concrete encoder sizes and dimension-storage tradeoffs, showing how to choose a configuration for text, code, image, video, and audio retrieval.

AITogether AI is running a large dedicated inference cluster of NVIDIA B300 GPUs on IBM Cloud, backed by NVIDIA Spectrum-X Ethernet networking, and is the first customer on it. Together AI operates the inference layer, IBM provides the cloud, and NVIDIA supplies the silicon and networking. The companies say the setup aims to deliver enterprise-grade, open-model inference at scale.

AICursor's iOS app now lets users see and reply to local agents running on their computer. Remote control is on by default except for Enterprise organizations, and agents keep running on the computer rather than moving to the cloud. The computer must stay on and online, and users can enable Keep this computer awake in desktop settings.

AIGoogle has been accused of clearing 300 hectares of Finnish forest for an AI data center before an environmental-impact assessment was finished, according to the Finnish Association for Nature Conservation. Critics also say Microsoft's wetland and garden restoration around its Texas data centers masks their effects, with Public Citizen calling the plan "lipstick on a pig" and noting the sites would draw power from gas plants. Microsoft's own estimates show its emissions rose 25% in 2025, driven mainly by data center expansion.

AIVercel's COO Jeanne DeWitt Grosser described how the company built an AI agent that runs the top of its sales funnel, starting from a roughly 125-line prompt written by its best SDR. The team moved the agent from supervised drafting to autonomous operation by August, then split the prompt into 14 deterministic rules and a model-handled judgment layer. Grosser said the system runs inbound for about $1,000 per year in inference and infrastructure.

AIClaude Code v2.1.290 adds serverToolUses to plugin turn.step results and agentId to tool.check hook events, so hooks can distinguish subagent permission checks. The release also adds a Deny button to the Claude apps gateway sign-in approval page, plus claude attach and claude logs accepting partial session names.

AIEthan Mollick reports that he moved much of his complex Cowork work to the new Claude Projects, which persistently chat with a dedicated cloud VM, finding them much better in most ways but poorly documented. Felix Rieseberg, who works on Cowork, explains that the new version runs model inference and the VM in the cloud, with each session in its own sandbox that is destroyed when the session ends. Files are accessed only from folders the user explicitly adds, with the desktop app handling those requests.

AINVIDIA's Olympus is a 10-wide out-of-order server core running at 3.3 GHz that prioritizes per-clock performance over high clock speeds. It uses a simultaneous multi-threading (SMT) implementation, unlike Arm's Cortex X925, and has out-of-order structures larger than X925's. In SPEC CPU2026, its branch prediction accuracy is slightly behind AMD's Zen 5 and slightly ahead of Intel's Lion Cove.

AIAnthropic's Cowork will run tasks in the cloud instead of on users' local machines starting tomorrow. Tasks can still access files and tools on the user's computer, limited to folders explicitly added to the session.

AIDex Horthy advises writing all decisions and context into documents in the artifacts, such as design or research files, so sessions can resume after compaction or be handed to another person. He suggests loading them in a new session with a skill like `/rpi:iterate-design-discussion`, or simply @-mentioning the relevant artifacts. His core principle is that nothing important should live only in the context window.

AIPyTorch has consolidated all media decoding and encoding for images, video, and audio into TorchCodec, which now runs on CPU and CUDA. TorchVision and TorchAudio are narrowed to focus on their transforms, with models, datasets, and pipelines no longer under active development. All three libraries are now ABI stable and no longer need rebuilding for each PyTorch release.

AISemiAnalysis measured usage meters on Anthropic and OpenAI subscription plans to estimate each plan's API-equivalent value. At mid-tier models, it found Anthropic offers roughly 5x the value of OpenAI, after OpenAI halved its $200 plan limits and introduced a $500 tier. The analysis also argues that subscriptions take a large share of inference compute while providing a small share of revenue, so their limits materially affect lab margins.

AIPika announced that developers can build with Ideogram 4.5 through the Pika API Club. The post links to the Ideogram 4.5 model page on Pika's developer platform.

AIThe llama.cpp v0.6.0 release adds Clef support for text and vision, along with high-quality support for Qwen3.8-Flash-Next. It also brings a major Metal performance improvement and a new llama_batch_ext API, and the project website at llama.app has been refreshed.

AIReflection AI says its Beam model, with a 500B form factor, combines strong agentic performance and efficient reasoning for enterprises, governments, and developers. Beam is in final red-teaming and will be released this month under an Apache 2.0 license, with quantized FP8 and NVFP4 versions for efficient deployment. Early access sign-ups are open on the company's platform.

AILowe's enterprise AI transformation leader outlines six guidelines for governing AI systems, arguing that people must set principles, decision rights, and escalation thresholds rather than only building the technology. The author, who coauthored The Enterprise Brain, cites a 2025 MIT Media Lab Project NANDA report estimating that about 5 percent of integrated generative-AI pilots generated substantial value.

AICognition introduced Agent Memory Repo, which stores memory records as files versioned with Git. The post says this simple primitive enables uses such as a message board shared across agent sessions.

AIOpenAI is extending its content provenance approach to text, starting with watermarking eligible text from ChatGPT and Codex in the EU over the coming weeks. The company says this is in response to EU AI Act requirements and acknowledges the significant limitations of current text watermarking technology. API customers can turn on text watermarking for select models worldwide starting today.

AIOpenAI has optimized inference for GPT-6 Astra and GPT-6.1 Sol, making them about 50% faster by default across subscription plans and Sign in with ChatGPT partners. The change requires no action from users and should be noticeable within two hours of rollout.

AIThe article examines whether multi-agent swarms could become a new scaling law, comparing them with inference scaling from o1. OpenAI researcher Noam Brown said its models are now sometimes trained with other agents, while the cited Anthropic data suggests gains beyond 10 agents are smaller and mainly speed-related. The article also raises the risks of groupthink and misaligned agents, and it notes that a Microsoft Research and UC Berkeley paper found teams sometimes solved tasks solo agents could not.

AIGoogle Antigravity has integrated the AlphaGenome Atlas Skill into its scientific workbench, enabling AI agents to help researchers prioritize genetic variants, generate structural plots, and build testable hypotheses. The company showcases researchers Natasha and Kyle using the tool in a demonstration video.

AILiving Models paired Google's Gemma 4 with BOTANIC-1, a plant-DNA model trained on 320 species, to identify causal genetic variants. In a melon yield test, the pipeline ranked the target mutation first out of 2,494 possibilities in under four minutes. The approach aims to speed up breeding of climate-resilient crops that would otherwise take years of field trials.

AICursor's SDK now supports steering and system prompts for local TypeScript agents. Background subagent results are returned locally in TypeScript and Python.