@davert
Account based inUkraine
About this account
- Account based in
- Ukraine
- Connected via
- Web
Account-level information from X, not a live location or the device used for a specific post.
Prince of Persia on day, developer at night. Author of @CodeceptJS and @Codeception testing frameworks. Currently working on @testomatio to make tests visible!
Kyiv. Ukraine
Joined November 2008
- Tweets18.3K
- Following925
- Followers3.1K
- Likes45.8K
Pick cheapest models and run exploratory tests for hours with no interruption
⬇️
Explorbot uses three model slots:
model — reads the page
agenticModel — plans and judges
visionModel — interprets screenshots
Same model in all three, or three different ones. Routing is part of the runtime, so you can balance reasoning quality, latency and cost.
github.com/testomatio/explor…
davert 🇺🇦 retweeted
"AI testing" can mean a generator, a recorder, a copilot or an autonomous agent.
Explorbot is the last one: it plans an exploratory session, drives the staging UI and returns structured tests with video and console evidence.
Humans review the run, not every click. Early access:github.com/testomatio/explor…
Transalting from gptish to english:
Cursor will store openai models responses and use them to train next Grok models.
In this schema Cursor is not an IDE but a laundary. So decision seems reaaonable.
P.S. It's totally my personal interpretation.
We’re ending our partnership with Cursor following its acquisition by SpaceX. Under our proposal, Cursor’s direct access to our models would end on November 12.
We know that the people most affected by this decision are the developers who rely on OpenAI models in Cursor. We care about their experience in this transition and we’re ready to go above and beyond to support them.
openai.com/index/our-decisio…
The loop is finally closed
Requirements ➡️ Tests ➡️ Runs ➡️ Defects ➡️ Analytics
Everything to establish full scope QA process is on Testomat.io
Catch the bugs that block your tests before they block your release 🐞
New in Testomat.io: a Defects page in every project, a Defects report in Analytics, and Defects + Users tabs on milestones.
Plus @ mentions in comments, paused manual runs, and trend charts.
which is explorbot
testomat.ai/explorbot/
which can test an app as user, while you are building it!
Everyone is talking about gpt-5.6-sol but I must say
gpt-5.6-luna from @OpenAI is awesome for agentic web testing!
For testing web you don't need smartest model, just the smart one. Paired with gpt-5.4-nano for actual execution they do great job testing apps!
See them running
You can also launch agentic testing with Explorbot
It's opensource and BYOK
testomat.ai/explorbot/
I found that @AnthropicAI has unclosed gap in tgeir API platform
There is no cheap, 20-30b model in their stack costing up to 0.5$M tokens. The smallest Haiku is too slow and expensive and can't be used for simple parsing and analysis tasks
There is gpt nano, gemma lite...
Runs on your infrastrcture
Consumes minimal tokens using simple models like gpt-oss-20b and gemma-4-31b
I'm eager to share as I think in age of vide coding we eagerly need vibe-testing solution as well!
testomat.ai/explorbot/
Hey @GroqInc is there any chance that llama 4 stays?
It was the cheapest, fastest, image recognition model. There is no alternative to it! The new Qwen you propose is 6x times more expensive.
It is sad to see instead of bringing new models you remove old good working ones
Our testcases are in #markdown format. The AI native format for agents. But you can write them in a notion style editor with a cool visualization
Check it out. Best of two worlds: classical tms step editor but for agentic era
New in Testomat.io: block-based editor for classical test cases. Each step gets its own expected result field, images attach inside the step.
Still markdown under the hood, so AI features and git workflows keep working. Toggle back to raw markdown anytime.
#QA #TestManagement #TestAutomation
Automated tests are not only scripted ones.
Now you can plug automated exploratory tests into pipelines and have AI discover new scenarios accross the website
More on it here:
testomat.ai/explorbot/
it's free, opensource, BYOK
Ai driven Exploratory testing is already essential part of our qa pipeline as unit and end 2 end tests. The thing is exploratory testing always tries new things and provide us with fresh feedback on UI
Its free and opensource, you can spin your own explorbot on your CI!
What if exploratory testing could run on its own? 🤔
That’s exactly what our CTO Michael Bodnarchuk (@davert ) explored while building Explorbot — an open-source AI-powered testing agent that navigates web apps like a real user.
Instead of following rigid predefined scripts, Explorbot:
▫️ Reads the UI and discovers flows on its own
▫️ Tests CRUD actions, filters, forms, and edge cases
▫️ Tries unexpected inputs, SQL/JS injections, weird characters, and hidden paths
▫️ Converts successful flows into real Playwright or CodeceptJS tests
One of the most interesting parts?
Explorbot generates around 30–50 meaningful tests per hour while running continuously in parallel sessions 🚀
And it’s not about replacing QA engineers or regression suites. It helps cover the areas teams usually don’t have time to revisit:
✔️ New features
✔️ Low-coverage areas
✔️ Forgotten admin pages
✔️ Edge-case behaviour
Even better — the project is completely free and open source, so teams can run it locally, in CI, or on their own infrastructure.
Really exciting direction for automated exploratory testing 👀
Worth reading the full article if you want to see how it works behind the scenes: testomat.io/blog/meet-automa…
#QA #TestAutomation #Playwright #CodeceptJS #TestingTools #Testomat
