Find published research by company, product, platform or technology.
Showing 221–240 of 372 dossiers
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.
The useful finding is not that BRIN is bad: it is that a table can still report strong column correlation while update churn has destroyed the physical range locality BRIN actually depends on. DeepSQL published the full Docker/SQL harness so teams can reproduce the failure mode on their own workloads.
Adobe Commerce and Magento merchants should treat CVE-2026-71362 as an urgent patch: independent security telemetry reports exploitation attempts even though Adobe’s bulletin still says it has not observed exploitation in the wild.
Panda CSS 2.0 keeps the familiar styling API but changes the compiler, minimum Node version and how teams can distribute design systems.
Woodpecker's agent labels were self-reported and unsuitable for authorization. Version 3.19 adds server-held filters and patches a clone-step environment-variable leak; administrators should verify their worker policies.
Effect 4 changes runtime architecture and maintenance guarantees, not just APIs. Its reported 5x smaller bundles and 86% lower fiber memory are vendor benchmarks requiring workload-specific validation.
GLiNER2.5-Decide attacks the same bounded-decision layer as Jev and CLM from a much smaller encoder architecture. Its strongest benchmark claims are vendor-produced, but CPU deployment and constrained joint decoding make it a materially different option for software-facing AI decisions.
The useful shift is not another AI wrapper around CI. sem-ai exposes CI/CD as structured, self-describing operations that Claude Code, Codex and other MCP-aware agents can call directly, including failure diagnosis and pre-push testing in CI.
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 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 useful lesson is architectural rather than vendor-specific: coding agents inherit execution paths from ordinary developer tooling. If an agent shells out to Git without sanitising repository-local configuration, a hidden `.git/config` can become a host-level command channel that bypasses the controls users think govern the model.
The interesting change is security economics rather than another hosting feature. A control that previously sat behind a $150/month add-on is now free across plans, changing the cost boundary for private dashboards, internal tools and pre-launch production domains.
The change is separate from post-quantum TLS. DNSSEC signatures authenticate DNS records, and ML-DSA-44 makes them dramatically larger — 2,420 bytes per signature — while dual-signing with older algorithms creates a downgrade path unless resolvers enforce the post-quantum chain deliberately.
The useful shift is automation at the CDN-to-origin boundary: operators no longer need to manually force post-quantum key exchange, while Cloudflare says its measured HelloRetryRequest rate fell from about 52% to 3.7% across the scanned cohort.
The counting-rule change is no longer theoretical. Early post-cutover data suggests public views can materially outpace Engaged views, with the size of the gap varying by channel size, category and discovery surface.
The change moves maintenance work earlier in the contribution funnel: instead of filing a report and waiting for a maintainer to reproduce it, package users are being asked to arrive with an executable patch candidate. It is a real workflow experiment, but Otwell's prediction that this becomes the norm should remain a founder/maintainer view rather than an industry fact.
Product teams can launch a root-cause investigation from an Insights report, an alert or Mixpanel Agent instead of manually trying breakdown after breakdown. The result is operationally useful, but it remains an automated statistical diagnosis rather than proof of causation.
For some queries, Google is no longer waiting for a user to open a longer AI Overview. The system can expand it automatically, creating a larger zero-click answer surface and moving traditional results further down the page.
Microsoft Advertising is taking Max CPC out of new standalone automated campaigns from October 1. Existing capped campaigns and portfolio strategies retain the control for now, but advertisers creating new campaigns will need to rely more heavily on conversion targets, budgets and portfolio bidding.
Google’s September Search changes now form a broader migration story: legacy campaign-level Broad Match and standalone Automatically Created Assets settings will be converted into AI Max, while language targeting stops affecting Search delivery and related API mutations begin failing.