@mehulmpt Building https://t.co/E0z4u8J7VH. Started as a local AI clipper and auto-captioning Mac app, but like every side project of mine, it didn't stop there. Now it's an AI editor. Works with OpenAI, Anthropic, OpenRouter, or your own local LLM. Coming soon.
5/
Two massive openings for engineers right now:
→ Use code to unlock problem spaces that were never worth coding for before.
→ Build the infra that holds up when 10x traffic shows up.
Question is whether you can do either.
4/
Second thing: all that new code is hammering existing systems.
More API calls. More scrapers. More agents. More of everything.
Building for humans is one problem. Building for agent traffic is a completely different one. And way harder.
India's first
OpenCode Buildathon.
100 builders. $100K cash & credits. 8 hours.
@opencode is the worlds biggest open source coding agent, period. 6.5M monthly active users. 143,000+ github stars. The company has 20 years of shipping open source developer tools. Backed YC, the founders of paypal, linkedin, yelp, and youtube.
this weekend, we’re bringing OpenCode team to organize the craziest buildathon in bengaluru.
what to expect?
every frontier model available (claude, gpt, gemini) plus the best open models (kimi, qwen, llama). And multiple tracks (viral consumer, hack for good, & multi agent as a service). prizes across every track.
where is this happening?
@Razorpay HQ. the tech sanctuary that has produced a generation of operators, founders, and the top 1% tech talent. you’re building in a space that already carries builder history. come to build your next big thing at Razorpay :))
who should apply?
experienced engineers. founders who ship. IC's with strong github graphs. people who already live in a terminal and want a weekend to build something real with the best open coding agent out there.
applications close in 48 hours 👇🏻
I went deep into my own Go agent framework last weekend and found something the major Python frameworks still haven't nailed.
What happens whn your AI agent hits its iteration limit?
It's the difference between an agent saying "I failed" after 15 minutes
and saying "here's what I found in the first 3 sources, I was about to check 4 more - want me to keep going?"
The secnd one is what users actually want. And nobody's shipping it.
The reframe is simple: budget exhaustion is a checkpoint, not a failure.
Same model. Same conversation. Just tell the LLM honestly that it ran out of room and ask for the best partial answer it can give.
Then let the caller decide whether to spend more.