Who contributed to Vercel on May 21, 2026?

8 developers shipped this update, including unstubbable, wbinnssmith, lfades, nsidnev, falcoagustin, pranaygp, mitul-s, and github-actions[bot].

What were the notable Vercel updates?

use the action cache for passing tests (#93954), Add experimental.appShells feature flag (#93997), and Prerender HTTP access fallbacks with Cache Components semantics.

@vercel

Next.js + the Vercel platform - frontend infrastructure for the web

github ↗

Pick a date

Topics: JavaScript Full archive →

The Wire · Showcase

NEXT.JS TEST CACHING LANDS, APP SHELLS PLUMBING SHIPS

By RepoJournal · Filed 06:01 UTC on May 21, 2026 · About Vercel

8 people shipped this

unstubbable @unstubbable 2 cited

pranaygp @pranaygp 2 cited

wbinnssmith @wbinnssmith 1 cited

lfades @lfades 1 cited

nsidnev @nsidnev 1 cited

falcoagustin @falcoagustin 1 cited

mitul-s @mitul-s 1 cited

github-actions[bot] @github-actions[bot] 1 cited

Next.js cut test reruns by memoizing passing tests across workflow attempts, while experimental App Shell support starts wiring up the client-side prefetch pipeline.

The action cache for passing tests [1] is live in next.js, which means re-triggered CI runs skip tests that already passed on the first attempt. Sync writes to a flat file track the winners, so timeouts or cancellations don't waste compute on redundant passes. In parallel, the experimental.appShells feature flag [2] landed as pure plumbing: config field, zod schema, and `process.env.__NEXT_APP_SHELLS` define are in place, but nothing reads it yet. Subsequent PRs will wire the server and client behavior behind it. Both the Cache Components prerender semantics [3] and the corresponding test coverage [4] shipped to lock in fallback boundary behavior when `notFound()`, `forbidden()`, and `unauthorized()` recovery routes through Cache Components. The devlow-bench runner [5] now reports p50/p90/p99 percentiles instead of means, with Mann-Whitney p-values for cleaner signal, plus per-attempt buffering to isolate flaky page-load runs from the final stats. Across the AI desk, Vercel's SDK locked in serviceTier reads from both the Gemini response header [6] and the usageMetadata body [7], so streaming and non-streaming calls both surface the applied tier. CLI got a `--project` flag [8] for all major commands (build, deploy, pull, dev), enabling non-interactive CI and agent-driven workflows without prompts. Python functions now support runtime cache configuration [9], and blob commands speak OIDC [10] alongside the legacy token path. Workflow repo fixed trace viewer zoom clipping [12], reports corrupted event logs as a distinct error code [11], and forward-ported a stale wait replay race fix [13]. SDK bumped to 1.21.6 [14] with vcrRepository permission fields added to microfrontends and projects APIs.

Action items

→ Test and merge Next.js test caching into your CI, then redeploy workflow to see reruns skip vercel/next.js [plan]
→ Upgrade @vercel/blob to use OIDC auth if you're already running `vercel env pull` vercel/vercel [monitor]
→ Update SDK to 1.21.6 if you consume vcrRepository permission fields vercel/sdk [monitor]

References

[1] use the action cache for passing tests (#93954) vercel/next.js
[2] Add experimental.appShells feature flag (#93997) vercel/next.js
[3] Prerender HTTP access fallbacks with Cache Components semantics ↗ vercel/next.js
[4] [test] Prerendering HTTP access fallback pages with Cache Components ↗ vercel/next.js
[5] devlow-bench: percentile-based comparison and run retries ↗ vercel/next.js
[6] fix(google): read serviceTier from x-gemini-service-tier response header (#14937) vercel/ai
[7] fix(google): read serviceTier from usageMetadata in stream + generate (#15488) vercel/ai
[8] [cli] Add `--project` flag to `build`, `deploy`, `pull`, and `dev` ↗ vercel/vercel
[9] [python] configure runtime cache for Python functions ↗ vercel/vercel
[10] [cli] Support OIDC auth for blob commands ↗ vercel/vercel
[11] Report corrupted event logs distinctly ↗ vercel/workflow
[12] trace viewer bug fix + file cleanup ↗ vercel/workflow
[13] [codex] Forward port stale wait replay fix ↗ vercel/workflow
[14] chore: 🐝 Update SDK - Generate VERCEL 1.21.6 ↗ vercel/sdk

Quick answers

What shipped in Vercel on May 21, 2026?: Next.js cut test reruns by memoizing passing tests across workflow attempts, while experimental App Shell support starts wiring up the client-side prefetch pipeline. In total, 50 commits, 55 pull requests, and 12 releases landed.
Who contributed to Vercel on May 21, 2026?: 8 developers shipped this update, including unstubbable, wbinnssmith, lfades, nsidnev, falcoagustin, pranaygp, mitul-s, and github-actions[bot].
What were the notable Vercel updates?: use the action cache for passing tests (#93954), Add experimental.appShells feature flag (#93997), and Prerender HTTP access fallbacks with Cache Components semantics.

NEXT.JS OPTIMIZES APP ROUTING ENTRY POINTS WHILE WORKFLOW CUTS STREAM LATENCY 20 PERCENT

Next.js extracted invariant app-page handlers into shared runtime modules while Workflow's stream write batching moves into core, eliminating per-token round trips on idle producers.

+11

javascript 126 shipped 2-min read

@vercel 1 day ago

TWO CRITICAL SECURITY PATCHES LAND IN WORKFLOW AND EVE OVERNIGHT

Next 16.2.11 and PostCSS 8.5.12 patches ship across workflow to seal CVE vulnerabilities, while Eve locks down workflow failure visibility.

javascript 153 shipped 1-min read

@vercel 2 days ago

WORKFLOW 5.0 BETA ARRIVES WITH HOOK TOKEN RETENTION, AI SDK FIXES GEMINI AND BEDROCK HALLUCINATIONS

Workflow shipped its most ambitious beta yet with experimental token retention, while the AI SDK shipped five critical fixes across Gemini, Bedrock, and dynamic tool validation in a single push.

javascript 132 shipped 2-min read

@vercel 3 days ago

NEXT.JS PATCHES FIVE SECURITY HOLES OVERNIGHT WHILE VERCEL CLI GAINS AI GATEWAY BUDGET CONTROLS

Five security advisories dropped across Next.js in the last 24 hours, ranging from origin validation on internal redirects to fetch cache key collisions that could leak cached responses between requests.

+12

javascript 145 shipped 2-min read

Elsewhere on the wire

AI Agents about 10 hours ago

CLAUDE OPUS 5 LANDS ACROSS THE STACK

The newest Anthropic model is now live in langchain, Cline, and llama-index, with native support for extended reasoning and 1M context windows.

ai-agents 28 shipped 1-min read

Local LLMs about 10 hours ago

OLLAMA LANDS LAGUNA SUPPORT AND CRUSHES MEMORY LEAKS WHILE SGLANG HITS V0.5.16 WITH CONFIDENCE-DRIVEN SPECULATIVE DECODING

Ollama shipped three critical performance and reliability fixes for Metal residency and concurrent access patterns, while SGL-Lang released 0.5.16 with a new speculative algorithm hitting 383.7 tok/s on DeepSeek-V4.