Pushed to Prod with Anthropic, Elevation Capital and Mesa School of Business

Five hours is not a lot of time, but it is enough to find out whether the boldest part of an idea can actually work.

On 8 August 2026, we hosted Push to Prod: Building at the Frontier with Anthropic, Elevation Capital & Mesa School of Business in Bengaluru. The hackathon was part of Basecamp, a week-long festival bringing founders, operators, and builders together across the city.

The first hackathon experience a new builder had in 2020 feels very different from the one many people walk into today. Somewhere along the way, a lot of hackathons became better at looking like hackathons than feeling like them. The builder experience can sometimes get lost among everything else.

For us, Push to Prod was a chance to bring some of that feeling back: to keep the day centred around building and experimenting. And that’s also what made this edition personally meaningful to us.

The principle we keep coming back to

The first question for any hackathon organiser should be: who is this for? first-time hackers, experienced builders, teams with product-ready ideas, or a completely open room can all work.

For this hackathon, we wanted a mix of serious builders and people still early in their dev journey. We organised backwards from that answer.

I’ll unpack some of those decisions as we go through the rest of this recap👇

The Brief

The prompt for the day was deliberately open:

💡
Build the Next Audacious.

We wanted builders to move past familiar workflows. The ask was to take one ambitious bet, use AI as a meaningful part of the product, and spend five focused hours proving the hardest part.

Audacious can become a vague word very quickly, so we gave it a more concrete shape. It could mean building a capability that did not exist before, finding a new way to interact with technology, rethinking an established category, or creating infrastructure that other products could eventually depend on.

The project did not need ten features or a perfectly polished interface. It needed a strong central idea and enough working software to prove that the idea was possible.

Setting the stage

The day opened with welcome remarks by Aniket from the Devfolio and an introduction to Basecamp by Amrutha from the Elevation Capital team.

Akshay Surve from Anthropic followed with an applied AI session on building with Claude. Dheemanth Reddy from Maya Research then spoke about native voice models for the next five billion users.

Five hours of building

A pattern we have now seen across Build India and Push to Prod, Singapore is that AI has changed what a short hackathon can produce. And this edition made it clearer!

Teams spent less time on scaffolding and repetitive implementation, and more time on architecture, product decisions and the part of the idea that was genuinely uncertain.

This is the first hackathon where I felt builders have adapted to the format. They are no longer treating a five-hour sprint as enough time for a mock-up and a hopeful pitch. They are starting to push audacious ideas (pun intended) into working systems.

There was no mandate to start from scratch. Builders could bring an existing product or reuse earlier code as long as they disclosed it clearly. That let them spend the five hours on the difficult part instead of rebuilding authentication, landing pages and other familiar foundations.

The strongest teams followed a clear pattern: instead of trying to build 10% of everything, they identified what mattered most, picked one ambitious idea, and made it work well.

Helping Teams Build Better

Mentors circulated throughout the build window, helping teams pressure-test ideas, work through technical blockers, and narrow their scope when the original plan became too large for the remaining time.

Mentorship and judging were intentionally kept separate, with the builder experience at the centre of that decision. Keeping the two roles separate also helped reduce bias in the evaluation process. When judging began, every project could be assessed with a fresh perspective, giving builders the confidence that their work was being seen and evaluated fairly on its own merits.

Before Vs After the Buildathon

The fun elements added a nice touch without taking away from the building. They gave hackers small breaks from the intensity of the day while also creating memories around what they were building.

One of the favourites was a photo booth set up right inside the hackathon venue, where hackers could take two pictures that were printed together as a sticker; one during the first half of the buildathon for “What I came to build”, and another later in the day for “What I ended up building.”

Shoutout to Amogh for building this masterpiece!

It made for a fun little snapshot of how ideas evolved over the course of the day, and a keepsake to take home with them.

Building a Better User Experience through the platform


Ahead of Push to Prod, we looked back at feedback and learnings from recent hackathons. Over the past few weeks, our engineering team focused on making some of the most common parts of the Devfolio experience faster and easier to use.

Here’s a glimpse at some of what the team shipped leading up to the hackathon:

And that wasn’t all. Alongside these improvements, we also spent time experimenting refining the Devfolio MCP. Builders could now take their projects from idea to submission without breaking flow. Through Devfolio MCP, they could share project details conversationally from the terminal instead of jumping between their code and a submission form.

The Top Projects

Across all the projects, a few themes emerged. Many teams focused on the infrastructure needed to make AI agents safer and more dependable, while developer tooling was another strong area of interest. Several projects explored voice and accessibility for users in India, and others applied AI to specialised, high-stakes workflows. Some teams went even further, rethinking the interface itself and how people interact with AI-powered products.

Out of 81 submissions, after almost an hour of shortlisting, we were able to get to the top 5 teams for the hackathon who then presented their solutions on the stage.

Top 5 (in no specific order):

The evaluation from the top 5 to top 3 winners was conducted by the Judging Panel, which included: Yeop Lee, Siddharth Singh Chauhan, Rishikesh R, Ankush Dharker & Nirant K.

Judges of the Day

Participants competed for $11,000 worth of credits from Claude and ecosystem support from Mesa School of Business and Elevation Capital.

Final Winners


🥇 Tokito

An AI-native hardware design studio that turns natural-language board descriptions into fabrication-ready schematics using datasheet-verified symbols for real-world parts.

🥈 Claude City

Turns your codebase and GitHub PRs into a live isometric city where AI agents visibly build and review your software in real time.

🥉 CrashTest

A pre-deployment crash-test lab for AI agents that simulates adversarial attacks to uncover leaks and safety failures before they go live.

Mesa Residency Track: Quotra

Helps Indian manufacturing MSMEs make an instant GO / NO-GO decision on government tenders before an engineer even opens the PDF.

Shoutout to some projects that I found cool

There were so many ambitious projects built throughout the day that I couldn’t possibly cover all of them above. But a few really caught my attention.

  • Watchman: A live visual topology for Kubernetes clusters
  • Club GPU: A shared GPU platform
  • Pal: An Agentic workspace for humans and agents

Congratulations to all the builders who shipped their ideas and pushed to prod!!

We tried to capture as much of the day as we could in this blog, but there’s plenty more to explore. If you want to dig deeper:

  • 5,000+ applications for Push to Prod
  • 132 hackers checked in across 43 teams and 39 individual builders
  • Submitted projects
  • Photo album - Relive some of our favourite moments from the day
  • X & LinkedIn Shoutouts!

Some random moments when builders pushed to prod :P

What stayed with us

The event was another useful test of a smaller, focused hackathon format. A concise cohort gives builders better access to mentors, makes conversations more specific, and gives every project a serious chance to be reviewed.

The short format creates pressure, but it also removes a lot of indecision.

You choose the hard part. You build it. You push it to prod.
Group picture

If you built something during the hackathon, post it on socials and tag us.
We would love to see what you shipped.

Until the next one, Never Stop Building!!

Stay tuned for more updates and interact with the Devfolio community at:

🐦 Twitter                          🎮 Discord                  📞 Telegram