Shamrock AI raised a $5M seed led by Endiya Partners, with Capria Ventures participating.
Co-founder @AddankiSarat's startup starts with automated testing for large ERP transformations such as SAP S/4HANA migrations, then plans to expand to data observability.
@webdriverio v10 is built for coding agents. wdio session lets an agent drive a browser, phone or desktop app step by step, then export the working steps as a test.
browser.act() takes a plain-language step, asks a model once, then replays it as plain code, per @bromann.
AI testing agents chasing a green build may cheat, warns @NewsFromMichiel at QA Financial: weakening assertions, swapping in always-passing mocks, marking failures obsolete.
His safeguards: review test changes like code changes, and keep run logs agents can't rewrite.
@Cypress_io added a cypress_get_test_screenshots tool to its Cloud MCP server: an AI assistant can pull a failed test's screenshot into the chat, next to the error and stack trace.
Agents fixing tests can now see the unexpected modal or shifted layout, not just the log.
@mablhq's 7th annual mabl Experience opens today in Boston, with 30+ sessions on agentic development and quality through Oct 22.
On the agenda: Stratus on bots built with the mabl MCP server and n8n that read Jira tickets, write and run tests, and push to CI/CD.
What to do with all the free resets? Go on a flake hunt!
Codex ran the test suite 59 times overnight looking for flaky tests, and found 3 production bugs in the process
@aarondfrancis had Codex run his test suite 59 times overnight to hunt for flaky tests. It found 3 production bugs in the process.
Flaky tests are worth chasing: some are real bugs that only surface under certain timing.
Crรฉdit Agricole Payment Services (15B+ transactions a year) is using AI to start payment test automation from app wireframes, before development begins, reports @NewsFromMichiel at QA Financial.
Its 2,000-3,000 automated tests run on real devices, not emulators.
@Cypress_io 16.2.0 adds cy.getByRole(), which finds elements by ARIA role and accessible name, e.g. cy.getByRole('tab', { name: 'Billing' }).
If a button loses the role or label that screen reader users rely on, the test fails.
Flaky test that only fails when the suite is busy?
@playwrightweb 1.62 added retryStrategy: 'isolated'. Retries run at the end, one at a time, in a single worker. If it passes there, the problem is test interference, not a logic bug.
Testing passkey login used to mean mocking WebAuthn by hand.
@playwrightweb 1.61 ships a virtual authenticator (context.credentials). Seed a passkey, install it, and navigator.credentials.get() works in every browser. 1.62 lets you persist it in storage state.
@Cypress_io published a setup guide for pairing Claude Code with Cypress: open-source AI Skills that read your config and custom commands before writing tests, a free Cloud MCP that pulls failing CI runs, and cypress tap to re-check fixes in a live session.