Working Student - Business Operations

AWS toolkit gives coding agents specialized cloud development skills. * By John K. Waters * August 27, 2026 Key takeaways. * AWS released an Agent Toolkit that gives coding agents specialized instructions for working with AWS services and current development practices. * The toolkit supports agents including Kiro, Claude Code, Codex, and Cursor, and can be installed through the AWS CLI. * The release reflects a broader shift in developer documentation toward information packaged for AI agents as well as human developers. Amazon Web Services has released an Agent Toolkit for AWS that gives AI coding agents access to curated instructions for working with AWS services, the latest effort to adapt developer documentation and tooling for software agents as well as human developers. The toolkit, announced Aug. 26, works with coding agents including Kiro, Claude Code, Codex, and Cursor. Developers can install it through the AWS Command Line Interface, which detects supported agents and installs a default collection of AWS skills. AWS describes the skills as specialized, current knowledge intended to help agents select appropriate AWS services, use modern APIs, and follow the company's security practices. The setup process also configures a connection to the AWS MCP Server for selected agents. Developers can search the available skills and add, update, or remove them through AWS CLI commands. The toolkit can also be configured non-interactively for environments such as continuous integration systems and development containers. AWS says version 2.35.0 or newer of the AWS CLI is required. The approach addresses a limitation of general-purpose coding agents: their underlying models may contain broad knowledge of AWS, but that knowledge may not reflect the latest APIs, services, or recommended development practices. AWS is instead providing a maintained layer of domain-specific instructions that can be installed into multiple coding agents. The release follows similar efforts elsewhere in the developer toolchain to provide agents with specialized instructions rather than relying entirely on information learned during model training. Gradle, for example, is developing official skills intended to teach coding agents how to interact with its build system. AWS's toolkit suggests that software development documentation may increasingly serve two audiences: developers who read it themselves and AI agents that use structured instructions to perform development tasks on their behalf. John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].

When AI code overloads CI, Universal Cache keeps builds fast and costs in control. Gradle Inc. is excited to share the launch of Develocity Universal Cache, a critical new capability designed to supercharge Continuous Integration (CI) pipeline efficiency, speed, and reliability. But first, the problem Universal Cache is addressing... AI is introducing new pressures on CI efficiency. The software delivery ecosystem is currently at an inflection point, driven largely by the sheer volume of code generated by AI. GenAI is magnifying challenges for engineering organizations, leading to an explosion in code output, test volume, and overall build complexity. As a result, CI pipelines are becoming overloaded, inefficient, and expensive. Several factors are contributing to slow CI systems: * Build volume explosion: AI has introduced more code more quickly (which also means more tests), all of which must be built and executed, putting tremendous strain on CI resources. * Ephemeral build overhead: The increasing reliance on ephemeral build environments means that every build starts on a blank slate. This forces massive, repeated overhead, where dependencies, JDKs (toolchains), and build intermediaries are constantly re-downloaded and recalculated. * Escalating costs: Historically, performance issues were mitigated by scaling CI infrastructure, but this approach is often no longer financially bearable. And now, consumption-based services, such as external dependency repositories or cloud artifact storage, drive financial costs linearly. * The vicious feedback loop: This strain extends to dependent systems, leading to a loss of Quality of Service (QoS). Higher load across dependencies can lead to slower feedback cycles, which yields increasing test flakiness, which in turn necessitates even more reruns - all of this further strains your already overloaded systems and creates a vicious feedback loop. A major source of these CI inefficiencies is redundant dependency downloads, wasting massive amounts of time. In one sample organization, the total wall-clock download time over a single week was over five days of idle CI time spent just on downloading artifacts. This constant strain damages developer productivity and organizational performance. The solution: Universal Cache. The clear path back to CI efficiency is comprehensive caching. Caching relieves pipeline pressure by reusing previous work (calculations, generations, and downloads), thereby reducing the load on CI and its dependent systems. Develocity Universal Cache is the unified, highly-available, and globally distributed caching layer that spans your entire organization, accelerating software delivery. Universal Cache efficiently stores build inputs, setup tasks/goals, and build outputs. By reusing data from jobs already run on CI, this solution dramatically reduces build times and removes one of the biggest sources of instability in the delivery pipeline. Artifact Cache: caching inputs before the build. Until now, development teams have focused primarily on caching the outputs of the build (using a build cache). But to truly optimize CI resources, Gradle Inc. must start caching before the build even starts - in other words, start focusing on inputs. Artifact Cache is designed to efficiently store and restore critical build inputs such as dependencies, toolchains (like JDKs), and build system (eg. Maven, Gradle, npm) artifacts. Downloading all of these files in one connection - versus many separate downloads over the course of the build - further boosts efficiency. As touched on above, in a typical ephemeral CI execution (without optimization), dependency download time can completely dominate the overall build time. For one example build, downloading dependencies took ~44 minutes, contributing to a total build time of ~50 minutes. When this same build was repeated using the Artifact Cache, the results were transformational: the time spent on network requests plummeted to just over 2 minutes, resulting in an overall build time reduction from nearly 50 minutes to 5.5 minutes - an acceleration of over 20x. Setup Cache: caching environments at build time. Setup Cache provides a fast, reliable, pre-configured build environment, ensuring every CI run is identical, fast, and stable from the moment it begins. Setup Cache accelerates the initialization phase of builds. Build tools like Gradle perform compute-intensive steps, including compiling scripts, building the task graph, and calculating file hashes before the actual build execution begins. By restoring the initialization state on the CI agent before the build starts, Setup Cache prevents builds from spending time doing work already done by a previous build, making the prepared configuration result available to your entire team. This optimization typically results in a 50% reduction of the Gradle configuration phase time, ensuring all builds start faster and more consistently for everyone. Build Cache: caching outputs for future builds. Build Cache remains as critical as ever for accelerating feedback cycles - even more so with the AI-driven increase in code volume. Gradle Inc. introduced the concept of a build cache within Gradle Build Tool in 2017, as part of its mission to automate and accelerate the build process for developers. With Develocity, Gradle Inc. took this idea of build caching even further with the remote (or distributed) cache. Develocity Build Cache doesn't just allow developers to cache and reuse outputs on their local machines, it also automatically caches tasks/goals previously executed on CI for reuse across different machines. So the larger the team and the higher the build volume, the more time saved by avoiding redundant work. The Spring organization saved almost 27 days of wall-clock time in CI builds over a single seven-day period by using Develocity Build Cache. Observability and caching must go hand-in-hand. For caching to be successful, observability must come first. Without deep insights, tracking reliability and measuring savings (whether realized or potential) is impossible. That's why Develocity's observability platform - Develocity 360 - is the foundation for any caching strategy. Build Scan(R) provides a persistent, shareable record of every build, containing detailed information on tasks/goals, tests, and dependencies. This data feeds into dashboards, allowing teams to rapidly identify bottlenecks and apply targeted caching for sustained speed and reliability gains. This also enables engineering leaders to measure the reliability improvement and validate the investment with hard data. The organizational benefits of Universal Cache. For engineering leadership and platform teams, Universal Cache addresses the critical challenges of cost, operational complexity, and governance. Reduced cost. Universal Cache cuts infrastructure expenses by eliminating redundant work across both local development environments and shared CI pipelines, preventing the need to continuously start up and scale ephemeral CI environments on expensive cloud infrastructure. Since caching reduces execution volume and traffic, it also minimizes reliance on costly external repositories with consumption-based pricing. Operational simplicity. Universal Cache is designed to operate as a single caching layer that requires just one single deployment and delivers full toolchain coverage, eliminating the complexity of maintaining multiple separate caching systems. Central infrastructure teams can deploy the service on Develocity Edge, placing it geographically close to CI agents and developer workstations - for example, in Availability Zones - to ensure optimal low latency and performance. Lightweight replication across Edge instances guarantees high availability and disaster recovery. Artifact Cache is smart, meaning it helps ensure secure governance of your software. It intelligently caches only stable artifacts. If unstable snapshots are required, it correctly makes the network request to the repository, ensuring the system avoids using potentially outdated or incorrect code. Artifact Cache also seamlessly integrates with Develocity Provenance Governor to support continuous Governance, Risk, and Compliance (GRC) automation throughout your toolchain. Leveraging detailed provenance data, attestation publishing, and policy enforcement, Develocity can block potentially vulnerable code, artifacts, dependencies and more to ensure a stable and secure system. The path forward. The systemic strain on software delivery pipelines caused by the exploding volume of AI-generated code, coupled with the friction of ephemeral environments and reliance on external consumption-based repositories, demands a strategic infrastructural solution rather than a patchwork approach. Develocity Universal Cache answers this demand. By integrating all three caching layers - Artifact Cache, Setup Cache, and Build Cache - it provides an observable, highly-available, and geo-located caching layer that accelerates builds and improves reliability. With the observability foundation provided by Develocity 360, teams can not only identify bottlenecks and deploy targeting caching, they can measure the true savings of their caching strategy and calculate ROI. Engineering teams that adopt Universal Cache will achieve the rapid throughput required to keep pace with modern software delivery, while organizations without a unified caching strategy may find that the downsides of AI code begin to outweigh the developer productivity benefits.