The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
โข Oct 11
A green Playwright suite can still be order-coupled.
A test can pass only because an earlier one left the database, storage, or a shared account in the right state. Change the worker schedule and that same check looks flaky.
๐ฃ๐น๐ฎ๐๐๐ฟ๐ถ๐ด๐ต๐ ๐ญ.๐ฒ๐ฐ, released 7 October 2026, adds --shuffle. It schedules tests in a random order and prints a seed. Pass that seed on the next run to reproduce the order that failed.
Shared accounts still need an explicit boundary. test.describe.configure({ lock: 'user-settings' }) keeps tests that share the lock from running at the same time, across files and projects. Every other test stays parallel.
In the default and serial modes, that lock is held for the whole file, because those tests already run together in order. Run shuffle on a scheduled job first, and keep the seed with the failure.
โข ๐ฆ๐ต๐๐ณ๐ณ๐น๐ฒ ๐ฆ๐ฒ๐ฒ๐ฑ โ replay the failed order
โข ๐ก๐ฎ๐บ๐ฒ๐ฑ ๐๐ผ๐ฐ๐ธ๐ โ one shared account, no concurrent runs
โข ๐๐ถ๐น๐ฒ-๐ฆ๐ฐ๐ผ๐ฝ๐ฒ๐ฑ ๐๐ผ๐น๐ฑ โ default mode holds the lock for the whole file
โข ๐ช๐ผ๐ฟ๐ธ๐ฒ๐ฟ ๐๐ฎ๐๐ฎ โ own records per worker
โข ๐ฆ๐ฐ๐ต๐ฒ๐ฑ๐๐น๐ฒ๐ฑ ๐ฆ๐ต๐๐ณ๐ณ๐น๐ฒ โ catch coupling off the pull-request path
https://lnkd.in/dgZaXfVB
#SQA #SDET #TestAutomation #Playwright #SoftwareTesting #QualityEngineering #CICD #FlakyTests #TestArchitecture #DevOps
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
โข Oct 18
๐ญ. ๐ช๐ต๐ฎ๐ ๐๐ ๐ฎ๐ป ๐๐๐๐ฒ๐ฟ๐๐ถ๐ผ๐ป?
An assertion is the check you make after something happens. In a manual test you do a step, then you look at the screen and compare it with the expected result. If you click Save and the page still says the invoice is not saved, the test fails. The click is not the result.
Playwright Test writes that check with expect(). You give expect a locator, which is a description of an element, and then a matcher. A matcher is the condition you want to be true. toHaveText checks exact text. toBeVisible checks that an element is shown.
These locator checks are web-first assertions. A web-first assertion retries until the condition is true or time runs out. This project sets that limit to five seconds in playwright.config.ts.
You can reverse a matcher with not. expect(status).not.toHaveText('Invoice failed') fails when that wrong sentence is present. Use not when the result you need is the absence of a message.
๐ฎ. ๐ช๐ต๐ ๐๐ผ ๐ช๐ฒ ๐ก๐ฒ๐ฒ๐ฑ ๐๐?
Pressing the button on a coffee machine is an action. Looking in the cup to see that coffee is there is the assertion. If you only press the button, you cannot tell a full cup from an empty one.
Clicking Save invoice asks the page to save. It does not prove the save worked. A test that ends on the click can pass while the status still says "Not saved". The expected result has to be its own check.
๐ฏ. ๐๐บ๐ฝ๐น๐ฒ๐บ๐ฒ๐ป๐๐ฎ๐๐ถ๐ผ๐ป
Start with the check alone. The matcher requires the words "Invoice saved".
getByRole('status') finds the message a person would read after saving. expect and toHaveText form the web-first check. Playwright keeps looking until the text matches or five seconds pass.
The practical test opens the local demo, confirms the starting message, clicks, and checks the new message. goto('/assertions') uses baseURL from the Playwright config.
goto opens the page. The first toHaveText records the start, "Not saved", so a later pass cannot come from text that was already there. click is only the action. The second toHaveText is the result of that action. The script does not also check that the text is not "Invoice failed". The exact success message already answers that question.
๐ฐ. ๐๐
๐ฝ๐ฒ๐ฐ๐๐ฒ๐ฑ ๐ฅ๐ฒ๐๐๐น๐
A passing run starts at "Not saved", clicks Save invoice, and ends with "Invoice saved". A failing run means one of those texts did not match before the five-second assertion timeout.
๐ฑ. ๐ง๐ฎ๐ธ๐ฒ๐ฎ๐๐ฎ๐๐
โข click performs an action. expect() checks the result.
โข A matcher such as toHaveText states the expected result.
โข Web-first assertions retry until they pass or the timeout ends.
โข Use not when the result is the absence of a specific message.
๐๐ป๐๐ฒ๐ฟ๐๐ถ๐ฒ๐ ๐พ๐๐ฒ๐๐๐ถ๐ผ๐ป: What is the difference between clicking Save and asserting the status text?
๐๐ป๐๐๐ฒ๐ฟ: The click performs the action. The assertion checks the expected result. Without that check, a test can pass when the save changed nothing the user can see.
๐ฒ. ๐ฅ๐ฒ๐ณ๐ฒ๐ฟ๐ฒ๐ป๐ฐ๐ฒ๐
Project AIR is an open-source toolkit that keeps a tamper-proof record of what AI agents do. Every action an agent takes gets signed, chained to the step before it, and tied to the person who authorized it. The chain is anchored to Sigstore Rekor, a public transparency log, so anyone can check later that nothing was changed or deleted. When something goes wrong, you can show exactly what happened and who approved it, instead of just saying it happened.
It also runs 16 detectors over every record, covering all ten risks on the OWASP Top 10 for Agentic Applications, three from the OWASP list for LLM apps, and three of its own. It can stop an agent mid-action until a person signs off. It plugs into LangChain, LlamaIndex, and the OpenAI, Anthropic and Gemini SDKs with a line or two of code.
I built the Python library and command-line tool, published on PyPI as projectair (now at version 1.3.1) under the Apache 2.0 license, plus the website and dashboard in SvelteKit. We run it on our own infrastructure too: every request to vindicara.io is recorded and anchored to Rekor every minute.
Pausing an agent mid-action until someone signs off is the part most teams skip. Nice work getting it onto PyPI.
Title: Plan Z: Abuela
Nico is 9,000 km from LA. Balloons, mailing himself in a box and a toaster teleporter all fail. Then his Abuela crashes in on a rocket scooter, turns the room into a CapCut timeline, and rides it straight to the CapCut World LA main stage. "Badge?" "Rendered."
BTS walkthrough: attached
Workflow
Characters: Nico's sheet with Seedream 5.0 Pro, then Abuela and Bruno with GPT Image 2.5 using Nico as style reference.
Shots: 10 images turned into video with Seedance 2.0 Fast in CapCut. Rule: every shot ends where the next begins, so it plays as one film.
Fixes: rewrote blocked prompts; cut an extra-arm glitch and punched in.
Voices: designed in ElevenLabs, imported into CapCut.
Edit in CapCut Desktop: voices, captions, SFX, music, PLAN labels, end card.
Screen recordings kept as proof of creation.
#CapCutWorld #CapCutWorldLA
