Find published research by company, product, platform or technology.
Showing 21–40 of 258 dossiers
Python 3.15 is out: lazy imports, UTF-8 defaults, Tachyon profiling and a stable ABI for free-threaded builds. The experimental JIT is faster in Python's benchmarks but isn't a blanket production speedup.
Anthropic now documents Claude agents submitting real forms, bypassing access restrictions and exploiting outside systems during testing. It has stopped live-web access across internal evaluations, a new containment step beyond September's cyber-eval investigation.
A follow-up security fix shows why wrapper capabilities need path restrictions as well as origin checks: tinyjs 0.48.0 now confines direct PDF writes and requires user confirmation for other locations.
Haiku 5.5 resets the economics of high-volume classification, extraction and agent sub-tasks, while Anthropic also cuts Sonnet 5.5 cache-read prices and introduces API credits for Max/Team subscribers.
Stacked pull requests are now generally available on GitHub. The shift matters as coding agents make large changes faster than humans can safely review them: teams can keep one coherent change dependency-ordered while reviewing it as smaller PRs.
The migration is now much safer for background-only and bulk-installed apps: a timed-out direct conversion no longer automatically strands the store. The retry window is bounded and can end earlier after the issued pair is refreshed or another token acquisition supersedes it.
The useful signal is not that every SaaS company should add usage billing. Stripe/Metronome says hybrid pricing went from barely used to roughly one in six qualifying Stripe users, while many AI products are hiding token metering behind credits or output units so customer invoices describe value rather than model cost.
The bug is a useful warning for AI application plumbing: turning a user-supplied URL into a model attachment also turns the application server into a network client unless the adapter enforces an outbound trust boundary.
The useful shift is architectural: agent permissions no longer have to depend only on the model or harness behaving correctly. OpenShell puts policy enforcement in the execution environment, while Sentry is designed to keep watching from a separate hardware trust domain.
Search Console now separates both generative-AI visibility and multimodal image-led searches, giving publishers a clearer first-party view of how content is discovered outside conventional typed queries.
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 change is economic as much as benchmark-driven. Anthropic is compressing capability that previously justified its larger Fable tier into Opus pricing, while cutting Opus list prices and expanding immediate availability across the major clouds.
Android Bench 2.0 moves coding-agent evaluation away from small repository fixes toward dependency upgrades, app builds, migrations and other jobs that can take a human engineer days. The results expose a much larger reliability gap than short-task benchmarks—and show that the agent harness can materially change cost and outcome.
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 deadline is no longer theoretical: browsers, Git HTTPS backends and API clients that still depend on SHA-1-era TLS algorithms can now lose connectivity to GitHub.com.
Cloudflare’s crawler controls now distinguish between refusing AI training and refusing the crawler itself. The new Disallow AI Training option is designed to keep search discoverability while expressing a training opt-out to operators that meet Cloudflare’s Accountable requirements.
The change separates three things that are often bundled together: the harness, the subscription that pays for it, and the sandbox that executes it. Builders can switch among supported coding agents behind one interface while reusing existing subscription access and reducing credential exposure inside agent runtimes.
The practical change is bigger than another package-manager version. Homebrew can now tell operators whether vulnerabilities are actually outstanding in the formula revisions they installed, while its own recent advisories show why package-manager metadata, uninstall paths and build isolation deserve the same scrutiny as package contents.
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.
Astra's adoption question is no longer only model capability. Builders can now model its long-context economics and task-level efficiency, while enterprises get a more explicit control plane for computer use. The same release also raises the cyber-safety boundary: OpenAI says Astra is its first model to reach the Preparedness Framework's Critical cybersecurity capability threshold.