DIRTY STATIC / DISPATCH Tools16 NOV 2026 • 9 MIN READ
Three large browser screens on a concrete stage in a brutalist hall, a robotic arm holding a glowing caution yellow dumpling above the centre screen, hazard stripes on the floor
ToolsINTELLIGENCE FOR THE MACHINE AGE

Playwright 1.65: Bun Runner, Browser Automation Updates

Playwright 1.65, released on November 5, 2026, lets the test runner run under Bun, adds a failures reporter and an onlyFailures option for HTML, JSON and JUnit reports, and adds an exclusive test lock. This profile covers how the Bun route works and where it differs from Node.js, what the new reporting trims, the bundled browser versions, and the breaking changes that can turn a routine upgrade into a red suite, checked against the official release notes on November 16, 2026.

Start reading ↓

A large Playwright suite in CI has two familiar pains. The run takes long enough that people stop watching it, and the report at the end is so full of passes that the failures get buried in noise. Version 1.65 works on both pains, and it adds a third change that will matter to teams who already use Bun for the rest of their stack.

Playwright is the browser automation and testing framework from Microsoft. It drives Chromium, Firefox and WebKit from code, in several languages, and its test runner is the part most teams live in day to day. Version 1.65.0 arrived on November 5, 2026, and a 1.65.1 patch followed on November 13. This article is based on the official release notes, checked on November 16, 2026. No testing by this desk is claimed.

Playwright

Official site

A framework for web testing and automation. Primary use case: writing end to end tests that drive real browsers, and automating browser tasks from code, with one API across Chromium, Firefox and WebKit.

Key features in 1.65: a test runner that can run under Bun; a failures reporter and an onlyFailures option for HTML, JSON and JUnit reports; a built in exclusive test lock; new timeout and signal options in cdpSession.send(); and a PLAYWRIGHT_TRANSFORM_CACHE_DIR setting for the folder that holds transpiled test files.

Advantages: one tool across three browser engines, a mature runner with parallelism and sharding, and no licence cost. Limitations in this release: the Bun route drops several runner features, and the upgrade carries breaking changes in class fields, actions, cookies and API requests that can fail existing tests.

Platforms
Windows, macOS and Linux for the test runner; Chromium, Firefox and WebKit as the driven browsers
Pricing
Free and open source under Apache 2.0

Run tests with Bun, with the flag

The headline change is simple to try. Bun is a JavaScript and TypeScript runtime. In 1.65 the test runner can run under Bun, and Bun transpiles your TypeScript test files itself, so Playwright does not use Babel or its compilation cache.

The flag matters. Without it, the command still runs on Node.js.

Terminal

bunx --bun playwright test

The project says its own CI runs the test runner tests under Bun 1.4.3 on Linux. That is a useful clue about maturity. It is a supported route with a named version and a named platform in CI, not a vague promise. It is not yet the same runner you get on Node.js, and the release notes list the gaps plainly.

AreaUnder Node.jsUnder Bun in 1.65
TypeScript handlingPlaywright transpiles with Babel and a compilation cacheBun transpiles the files itself
Watch modeAvailableNot available
UI mode on file changeRuns tests again when files changeDoes not run tests again when files change
--only-changedFollows importsDoes not follow imports
--tsconfig and JSX optionsTake effectHave no effect, since Bun loads modules itself
Custom ESM loadersSupportedNot supported
Stack tracesInclude async frames and class names in API namesNo async frames, and API names drop the class name, so click instead of locator.click
Console outputReaches trace and UI modeDoes not get into trace and UI mode

None of that makes the Bun route a bad idea. It makes it a choice with a price list. A suite that is heavy on TypeScript and runs in CI on Linux is the natural first test. A workflow built around watch mode is not.

The release notes make no speed claim for Bun, and this article makes none either. Removing a transpile step may help some suites and do little for others, where browser launch and page work dominate the clock. Measure your own suite if speed is the reason you switch.

A concrete control room with three browser screens, one screen isolated alone under a spotlight while the others wait behind a barrier One test alone under the light while the rest wait. That is the job the new exclusive lock does in a parallel run. Illustration: Dirty Static

Reports with only failures

The second change is smaller and, for many teams, more useful on day one. A new built in reporter named failures prints failures and successful retries as they occur, followed by a summary, without progress output.

Alongside it, the HTML, JSON and JUnit reporters gain an onlyFailures option. With it on, those reports keep only tests with unexpected outcomes, flaky tests and interrupted tests. The passes are still run. They just stop filling the report you open when something broke.

In a large suite, that is the difference between a report you scan and a report you search. CI logs get shorter, stored artifacts get smaller, and the flaky test that passed on retry stays visible instead of hiding among hundreds of clean passes.

A lock for the test that must run alone

Some tests cannot share the stage. A test that restarts a server, or one that owns a shared account, can corrupt every test running beside it. Playwright already had named test locks for tests that share a resource. Version 1.65 adds a built in lock named -exclusive-.

A test with that lock runs only when no other tests are running. All other tests continue to run in parallel with each other. It is a narrow tool for a narrow problem, and that is its strength. Use it for the one test that truly needs the stage to itself, not as a general way to calm a flaky suite.

The runner gains two smaller items worth knowing. The test.skip(), test.fixme(), test.fail() and test.slow() calls now accept only a description, recorded in the test annotation. And the new PLAYWRIGHT_TRANSFORM_CACHE_DIR environment variable sets the folder for transpiled test files, which helps on a Linux machine shared with other users.

What Works

  • Bun route removes the Babel transpile step for TypeScript suites that already use Bun
  • A failures reporter and onlyFailures option cut report noise without cutting tests
  • An exclusive lock solves the shared server and shared account problem in one line
  • Browser versions are current, with Chromium 158, Firefox 159 and WebKit 27.2

What Does Not

  • Bun route has no watch mode and drops tsconfig, JSX and custom loader support
  • Node.js 20 is no longer tested, so older CI images need a runtime upgrade first
  • Class field, inert, cookie and API changes can break tests that passed on 1.64
  • Stack traces under Bun lose async frames, which can slow down debugging

The breaking changes that can bite

This is the part to read before you merge the upgrade. The release notes list several changes that alter what existing tests do, without any new API asking for your consent.

Class fields in TypeScript now follow useDefineForClassFields from your tsconfig.json, the same as tsc. A field initializer that reads a constructor parameter property now fails, because the initializer runs first. The common page object pattern is the one at risk. Move the initialization into the constructor, or set "useDefineForClassFields": false in tsconfig.json.

Actions such as click, check, tap, fill, clear and select now wait for the element to be not inert. An action on an element with the inert attribute, or inside such an element, now times out if the element stays inert. That matches how a person meets the page, but a test that acted on inert content will now wait and fail.

Cookie matching now follows RFC 6265. With a URL, the cookie calls no longer return host only cookies for subdomains, and path matching got stricter. Proxy credentials are no longer sent to websites. They now answer only proxy challenges, and website login should use httpCredentials instead.

Three more changes sit in API and reporting work. String and URLSearchParams values in the params option are now appended to the query the URL already has, instead of replacing it. Text the browser does not render is no longer part of accessible names and aria snapshots, which can change role queries and snapshot checks. And a sharded run now fails with “No tests found” when no tests match across all shards, unless you pass --pass-with-no-tests.

Which should you choose

Try Bun on a branch if your stack already runs on Bun. Keep Node.js as the default until you have run the full suite both ways and compared the time and the failure output. If your team lives in watch mode, or relies on --only-changed, custom loaders or tsconfig paths, the Bun route is not ready for your workflow yet.

Take the reporting changes now, whichever runtime you use. The failures reporter and onlyFailures do not change what your tests do. They change only what you have to read after the run, and that trade is close to free.

Use the exclusive lock sparingly. One test that restarts a server is a good case. Ten tests marked exclusive is a suite design problem wearing a new costume, and it will cost you the parallelism you bought the runner for.

Plan the upgrade, do not drift into it. Check page objects for field initializers that read this.page, check CI for Node.js 20, and read the cookie and proxy notes if your tests handle auth. Then upgrade on a branch and run the whole suite before you merge.

Final thoughts

Playwright 1.65 is not a reinvention. It is a release about how tests run and how results read, and those are the parts of a testing tool that decide whether people trust it at 2 in the morning when CI goes red. The Bun route is worth a careful trial, the reporting work is worth taking at once, and the breaking changes are worth an hour of reading before they cost you a day of hunting. That is a fair shape for a point release.

Sources

■

END OF TRANSMISSION

Stay curious. Question everything.

CONTINUE THE DESCENT

MORE FROM DIRTY STATIC ↘