Ankit Kumar ☁️

@debasishbsws @brahma_4u @YouTubeIndia @YouTube See you here after a long time 😂

@brahma_4u @YouTubeIndia @YouTube Thank good mere pass YT premium hay

Dear @YouTubeIndia @YouTube Kindly do something about this guy, he ruins every video.

@manishtyagi__ @youdotcom Congratulations!

Thrilled to see @youdotcom hit a huge milestone! I've been fortunate to partner with the team for nearly 2 years, growing the developer community and seeing our APIs power 1B+ queries monthly for businesses building the next generation of AI. Big congrats to the team, and huge credit to @RichardSocher and @BMarcusMcCann for leading with vision and pushing the boundaries of what’s possible in AI.

@me_sagar_utekar Congratulations 🎉

New Objective: Stop Breaches. Power Business I am beyond excited to share that today I joined CrowdStrike as Site Reliability Engineer, leading SRE Operations from India. Joining the Cyber Security leader feels surreal. I am ready to dive in, contribute to our unified mission.

Where and whom to reach out in India 🇮🇳 to contribute to CNCF

@nitin_gadkari @thirumaofficial It’s deeply disappointing that while people are dying in road accidents like the one at Mahi Bridge, our leaders remain silent. Common citizens deserve your attention and empathy, not just birthday wishes to other politicians.

Birthday greetings to the Lok Sabha MP from Chidambaram, Tamil Nadu Shri @thirumaofficial ji.💐 May you be blessed with good health and long life.

What started as a dicussion on discord voice chat From a Discord chat to an 11th-floor office in Noida’s IT hub. What started as an open-source project with 50+ contributors turned into something none of us fully anticipated — @PixelwandHQ In the early days, we were juggling between college and writing code from our apartment. We were figuring out how to get our first client — and when none came, we decided to build our own product. That first product wasn’t perfect. But it taught us everything — and surprisingly, it brought us our first big client. Not everything was smooth. We lost some key team members — and with them, came our lowest moment. But we didn’t break. We learned, grew wiser, and came back stronger. Today, as we step into our new office at Noida, it feels like a small validation that we’re on the right path. When I first entered this space, my teammate by my side, I felt a calm clarity: This is the place I want to build in. To some, this might look like just an office. But for me — a boy who couldn’t afford tuition once, who almost gave up on education after 12th — this space is proof that dreams born in scarcity can find a place in abundance. When my parents visited the new office, I saw their eyes — part disbelief, part pride. That moment will stay with me forever. At Pixelwand, our ethos is simple: Learn by building. Deliver more than expected. Never compromise on quality. This office isn’t a destination — it’s just the beginning. Let’s build more. Together.

Hiring at @PixelwandHQ Roles: React-native intern B2B Sales intern Growth Marketing Why you should join? The View

@misalpavv @okta Congratulations 🎉🎉🎉

Not going for placements in college turned out to be one of my wisest decisions, and I stand by it completely. 🙃

@ghumare64 Congratulations Rohit!

🚨Career Update: I've joined Motia to break misconceptions in the world about AI Agents.

@nitishfy Congratulations 🎉🎇🎉

Career Update...!🎉 Joined the team as a full-time Software Engineer! :)

@nitishfy Take care!

Heard like 3 blasts in Gurdaspur, terrifying situation here.

@julianharris @michbarsinai Totally agree with you both— even great AI output needs validation. MBT helps here by modeling expected behavior and testing against that, not just trusting what one LLM (or another) says. It’s one way to close the trust gap.

Sure but aren't you comparing one LLM's output with another? In the case of a specialised AI tool that promises "a really good and up to date test suite" it can only promise that to the extent that any LLM can promise anything: it can't guarantee that. However with dedicated AI engineering work the hallucination gap can be reduced, of course.

AI is writing code faster than ever. But what happens when it hallucinates? This is how I almost shipped a major bug — and how I fixed it with something smarter: Model-Based Testing🧵

@julianharris Totally Agreed as prompt engineering helps a lot, but like you said, LLMs often cap out at ~80%. Tools like Provengo help catch what slips through that last 20%, especially when quality or safety really matters.

Seems like a solvable prompt engineering problem to me but I'm probably being naive as I do appreciate that for any solution that raw LLMs offer, for anything nontrivial it's rarely more than 60-90% of the quality needed, and the hard graft of AI engineering is required to add 9s to that.

@julianharris That's a great point — we do require tests for every change. But AI-generated code often passes basic tests while hiding subtle bugs. Model-Based Testing with Provengo models system behavior and auto-generates smarter tests — catching edge cases we might miss.

@ankiit111 Very interesting. But why not just have a rule that requires tests to be created and run for each change?

Full story here 👇 🔗 dev.to/ankit_kumar_cf… #AI #DevTools #Testing #MBT #SoftwareEngineering

After a few months using AI + MBT: ✅ Debugging time dropped ~60% ✅ Team confidence improved ✅ Faster, safer releases