Find published research by company, product, platform or technology.
Showing 1–20 of 79 dossiers
AgentControl now spans more production stacks: applications can resolve different prompts and models by context, track token/cost behavior, require approvals, use Bedrock without proxying inference through LaunchDarkly, and inspect multi-step agent runs as one conversation.
CLM-8B targets the same narrow decision layer as Jev, but with open weights, local deployment and a contrastive architecture that separates state and action representations. The headline speed and coding results are researcher-produced and need careful interpretation.
The important change is economic rather than another flagship benchmark win. OpenAI is making capable agent and coding workloads materially cheaper, with Luna approaching older Sol-class results at a tiny fraction of the task cost and GPT-6 prompt caching discounting reused input by up to 90%.
The interesting part of Fastly’s AI launch is consolidation: model gateway economics, LLM security and agent-to-API authorization now sit in the same request path as the CDN/WAF infrastructure many applications already use.
A third-party GEO dataset recorded an 86.4% relative collapse in Reddit’s visible ChatGPT Search citation share while Google AI citation changes were much smaller. The result is a useful warning against building an AI-discovery strategy around one source platform, not proof of an OpenAI penalty or Reddit removal.
Edge Scripts no longer have to sit behind bunny.net’s automatic cache. New pre-cache hooks can execute before cache lookup, while a separate Cache API lets scripts read, write and delete regional cache entries under application control.
Cloud CDN can now honor CDN-Cache-Control separately from browser-facing Cache-Control. That gives builders a standards-based way to set shared-cache behavior at the edge while preserving different client-side caching rules.
The material change is that model routing is no longer a single opaque optimization target. Developers can now tell Copilot whether to bias Auto toward lower cost, a middle ground or higher quality while GitHub still chooses a model prompt by prompt.
The pilot attacks a persistent evaluation trade-off: labs do not want to reveal frontier-model internals, while evaluators do not want benchmark prompts leaking back to the model provider. DeepMind says a Singapore AI Safety Institute pilot kept both sides’ sensitive assets hidden during execution.
Ahrefs is standardising Brand Radar on an estimated AI-demand metric because major AI platforms do not publish prompt volume. The new number can improve relative weighting between prompts and platforms, but it remains a modelled proxy rather than a count of how many people actually asked an AI system a question.
The important failure is not another prompt injection. Plugin4Shell breaks the mechanism intended to guarantee that an AI-agent plugin is still the exact code a marketplace reviewed.
The useful change is containment rather than another browser-agent feature. Teams can let an agent operate a real browser while constraining its HTTP and HTTPS reach to the site and dependencies the task actually needs, reducing the blast radius of prompt injection, bad tool decisions or untrusted page content.
The important change is at the gateway boundary, not just inference placement. OpenRouter says prompts can now stay in-region from decryption through provider execution and supported server tools, while teams can enforce the rule per workspace, team or API key.
SwarmLLM does not route whole prompts to separate machines; it pipelines one model across browser tabs. A MacBook and iPhone can jointly hold Qwen 3.8 27B even when neither device can hold the full 15GB quantized model alone, with no inference server in the loop.
Google is changing Gemini Notebook’s packaging from feature-style quotas toward a compute budget. That gives users more flexibility but makes the effective cost of one request less predictable and ties premium upgrades more directly to computational intensity.
Private Safety Processing is OpenAI’s attempt to reconcile stronger multi-turn safety monitoring with Zero Data Retention. Early customers are testing it now, with rollout and a technical white paper planned for September; important implementation details remain unpublished.
Meta has made the privacy-versus-price trade explicit in its Model API: developers can choose standard pricing or a contributor model ID with steeply discounted inference in exchange for training-data permission. The choice matters for proprietary code, customer data and AI SaaS workloads.
The change turns cache poisoning from mostly a workflow-design warning into an enforceable permission boundary. Teams can let untrusted jobs restore caches without writing them, prevent reusable workflows from escalating cache access and isolate jobs that only need to publish cache entries.
Next.js 16.3 separates two kinds of improvement: default Turbopack memory/build changes that existing apps can gain from an upgrade, and opt-in Cache Components/Instant Navigations that change how route shells, prefetching and blocking data are designed. Teams should evaluate those migrations independently.
Muse packages persistent autonomous execution, credentials, payments, app access and memory into a mainstream consumer product. A September macOS hotfix now provides an early real-world lesson: agent containment has to protect not only the cloud runtime but also the local control path into the agent.