The Future of Software Testing

The Future of Software Testing is a manifesto and a short discussion series about how AI changes software quality. The idea-submission terms, last updated 2 October 2026, say the series is sponsored by Katalon, Inc. and its affiliates. The footer says "FOST · powered by Katalon" and "© 2026 Katalon, Inc.", with mail to communications@katalon.com. Checked 2 November 2026. This note did not subscribe, did not post a comment, and did not send an idea.

It is not a test runner. e2e is a piece of software you install. This site is essays and a public position.

The marketing page is Astro 7.3.1, served from CloudFront. The copy received here was last modified 22 October 2026 (last-modified and the sitemap lastmod for /, /manifesto/, and /idea-submission-terms/). The sitemap lists five URLs. Two article URLs, the manifesto, the idea terms, and the home page.

Who signs it

The homepage names four contributors and gives these titles and year claims. The years were not checked against another source.

Each card links to that person's LinkedIn. The company links are Katalon on LinkedIn and the YouTube channel @KatalonTruePlatform.

The manifesto

Manifesto for testing in the age of AI is marked "Published September 2026" and signed by all four. The line they repeat is "AI accelerates testing. People lead quality." The page says more code, more tests, and faster delivery do not by themselves make better software. Five commitments follow. This is a paraphrase of the headings, not the essay:

1. Earn confidence. A green check and a large test count are not evidence. Look at what was tested, assumed, and left out, and whether that evidence matches the business risk.
2. Keep people accountable. A model can generate code, tests, and analysis. It cannot own what ships. Calling quality everyone's job does not make everyone an experienced tester.
3. Change the craft, not freeze the old mechanics. The page expects some repetitive work, and some roles built on it, to go away. It also says manual exploration, deterministic automation, and autonomous testing should sit side by side, chosen by risk rather than by fashion.
4. Give the model context, then check the model. The list they treat as the new testing infrastructure is requirements, source, test cases, results, production behavior, architecture, boundaries, constraints, and domain knowledge. They name hallucination, inconsistency, bias, omission, and false confidence as failure modes of the model itself.
5. Judge testing by the product, not by testing activity. Besides functional correctness they name usable, accessible, secure, resilient, performant, and worthy of trust.

The closing call says the future of testing should not be set by AI vendors, defended by people protecting the past, or left to machines.

What is published

The homepage describes three formats: the manifesto, thought-leadership discussions, and a how-to series. The how-to block says "Coming soon". The subscribe line on an article page says new discussions, articles, and playbooks. No playbook was on the sitemap.

Two discussions are marked published. Two are upcoming, with no date and no URL.

Article 01 is labeled "the series opener", by Alex Martins, "5 min read". The document title is "Who owns software quality in the age of AI?", which is not the H1. The piece restates the manifesto line, then asks who validates model output, who accepts the risk, and who is accountable for what reaches customers. A figure caption says AI magnifies the process it enters: strong practices scale, and weak coverage scales too. The closer still says "this month" the series will start exploring those questions. That sentence is from a 30 September page. By this check, article 02 had already been published and 03 and 04 were still upcoming.

Article 02 is "9 min read". Its document title is "Is manual testing dead? What AI really changes for testers". The header does not name a single author. The body is a conversation among the four. The opening provocation, which the article says nobody fully accepted, is that manual testing, automation, and everything else in testing are dead because AI will take over. The argument they keep:

• The work moving first is mechanical: scripted checks, repeated regression, routine generation, maintenance, and result sorting. Mush Honda's line in the piece is that the repeatable layer goes away and judgment stays.
• Cristiano Caetano's bottleneck story: manual execution was automated, automation then created brittle suites and noisy results, and AI is now speeding those up too. The decision that remains is whether to release.
• Vu Lam's line is that running a test someone else already wrote is the easy job, and the easy job is also the easy job for a model.
• Alex Martins's observation in the piece is that many large enterprises still have a lot of manual testing, often because those people know the domain.
• Three capabilities they say get more valuable: judgment (what the evidence means), domain expertise (a plausible output can still be wrong), and accountability (Mush: "You can't blame the AI"). They also say a model scales the process you already have, including a weak one.
• The Monday section says to onboard the model as a teammate that still needs context, not as a takeover. Vu: "AI is a member of the team, not a takeover of the team."
• The ending, in Vu's words as printed: the easy part of manual testing is dead; manual testing in another form is not, because the model has limits. What they say is disappearing is pre-scripted, repetitive, low-judgment execution.

The homepage comment box under the manifesto excerpt said "0 responses" and "Protected by reCAPTCHA". It was not used.

What they ask before you write in

The idea terms apply to the "Share your ideas" form, to mail sent to communications@katalon.com, and to any other channel they open for the series. A short reading, not a substitute for the page:

• You keep ownership of the idea. You also grant Katalon a perpetual, irrevocable, non-exclusive, worldwide, royalty-free licence to use it for any purpose, including the series, and to sublicense that right. Where the law allows, you waive moral rights.
• They may name you, skip your name, or anonymize. You can ask not to be identified. They say they will make a reasonable effort after they receive that request.
• Ideas are not confidential. There is no payment. You must be at least 18. Do not send employer secrets, someone else's personal data, or anything you cannot afford to see published.
• They do not have to read, use, or answer. A later Katalon piece that resembles your idea is not, under these terms, a claim.
• Personal data on the form is pointed at the Katalon privacy policy. Site use is pointed at Katalon terms.
• The terms in force when you submit govern that submission. Governing law is the State of Georgia, USA, with courts in Fulton County, and disputes are individual rather than a class action.
• The English text controls.

The subscribe blurb says you can unsubscribe, and that subscribing agrees to those same Katalon privacy and terms pages. Article pages also show a cookie preference center: accept, customize, or opt out, with performance cookies described as a way to count visits and see how people move through the site. This check did not accept cookies and did not open the cookie policy.

See also: e2e · Testing Tools