IETF publishes HTTP QUERY method as RFC 10008, first new HTTP verb since 2010

HTTP gains a new standard verb with RFC 10008

The IETF's June 2026 publication of RFC 10008 adds the QUERY method to HTTP, the first new standard verb since 2010. QUERY is intended for safe requests that still need a body — the gap developers have long papered over with POST, which is neither safe nor idempotent, or with GET, which forbids a meaningful body. By carrying filters in the body while remaining cacheable, safe, and idempotent, QUERY is positioned to clean up API designs that today rely on overloaded POST endpoints.

For working developers, the immediate change is at the API design layer: server frameworks, proxies, and API gateways will need to recognize a new method, and existing tooling that assumes GET-or-POST for queries will have to be revisited.

Spec-driven development moves from theory to practice

A wave of recent commentary frames the bottleneck in AI-assisted coding as no longer code generation but code verification: AI-generated output can drift from intent, and teams are turning to spec-first workflows to catch that drift early. The argument is that once a machine writes the implementation, the value of the team shifts toward writing and policing the specification the machine is supposed to satisfy. For practitioners, that means treating intent documents — not prompts — as the primary artifact, and putting review energy into verification harnesses rather than line-by-line critique.

The piece lands alongside other developer-coverage items this week that discuss verification, refactoring, and team autonomy in the same breath.

Incremental Rust refactoring as a Python performance lever

A practitioner talk walks through replacing Python hot paths with Rust via PyO3 without a high-risk rewrite. The pattern is incremental FFI refactoring: identify a function-level bottleneck, wrap it in a Rust binding, and verify with the existing integration test suite so infrastructure costs fall without paying microservice overhead. For teams running Python services with measurable per-request latency or compute bills, the implication is that PyO3-based replacements can be slotted in behind stable interfaces and benchmarked directly against production traffic. For data engineering work, a tool that helps profile and rank Python call sites by cost would be a natural follow-up.

JetBrains traces Rider and ReSharper startup regression to Microsoft Defender

JetBrains reports that Microsoft Defender was scanning its out-of-process ReSharper component for longer than expected, producing slower startup times on Windows after the OOP architecture launched. The fix landed in collaboration with Microsoft, and JetBrains built a tool along the way to let users exclude the process. For Windows-based .NET developers, the actionable item is auditing Defender process exclusions for any IDE process that loads heavy managed components, since similar scans can surface anywhere managed runtimes are involved. The post also underlines how performance regressions on Windows often live one layer below the IDE.

CLI releases cluster around AI coding agents

GitHub Copilot CLI picked up an update adding a `copilot instruction list` command in the same 24-hour window the Supabase CLI moved to v2.118.0-beta.18. A separate third-party project, oh-my-pi, shipped fixes to keep the GitHub-owned Copilot CLI OAuth app on a stable sign-in path and patched its `@oh-my-pi/pi-agent-core` package. For developers running multiple coding-agent CLIs side by side, the practical consequence is a tighter refresh cadence: token, scope, and instruction-set behavior is moving, and shell aliases or CI scripts that assume a specific CLI surface should be re-tested after each release.

Embedded RTOS meetup and openclaw's package recovery path

The Zephyr Project, an open-source real-time operating system collaboration hosted by the Linux Foundation, will hold an in-person meetup at the JetBrains office in Amsterdam on September 15. Zephyr sits in the embedded space, where small footprint and scalability are the design constraints. Separately, the openclaw package manager shipped a 2026.9.4 release whose headline feature is recovering from failed compatible updates by retaining the previous package and restoring it — a behavior worth knowing about for any team that has been burned by partial-upgrade states.

GitHub reports five August incidents, signaling reliability drift

GitHub's monthly availability report for August 2026 records five incidents that caused degraded performance across its services during the month. The post does not enumerate each incident in detail within the published excerpt, but the count itself matters for any team whose CI, package registry, or code search rides on GitHub infrastructure: availability is not a one-off, and incident-aware runbooks remain worth writing. For platform engineers, a dashboard that surfaces GitHub status events alongside internal CI queues is a natural follow-up.