Sharan Babu
@sharanbabu2001
MS in CS @GeorgeMasonU | LLM enthusiast | prev: @Amazon
Delighted to share that my project Gemini won the 3rd prize in @TigerGraphDB Hackathon (Track: Most Innovative) with a cash prize of 25,000$ Project Link: github.com/Sharan-Babu/Ge…
Having voice is definitely super helpful!
🚨Wave 11🚨 @windsurf_ai won’t stop shipping Voice has been great for me- I can just talk to Cascade and explain all my thoughts without having to worry about writing a detailed prompt Also, I can now chat with Windsurf Browser and ask questions about previous conversations
JSON as input prompt format for text-to-video is interesting and seems to be working well.
We're excited to announce that @stripe -sync-engine is now available as a standalone npm package:
Much needed!
Smart Turn v2: open source, native audio turn detection in 14 languages. New checkpoint of the open source, open data, open training code, semantic VAD model on @huggingface, @FAL, and @pipecat_ai. - 3x faster inference (12ms on an L40) - 14 languages (13 more than v1, which…
Nice!
Supabase Storage is getting better for everyone: - We've increased the maximum file size to 500 GB, up from 50 GB - Reduced egress costs for requests cached by our API Gateway - Free plans and paid plans get more GBs of cached, and uncached egress
Veo3 api now available! Fast model coming soon...
Check out the Veo 3 video+audio generation model, now in the Gemini API.
Prompt injection is a huge problem...
every single model release gets this from reddit now
Ya, I think skills on @diabrowser have great potential! eagerly waiting for developer beta access
Okay so while I'm digging @PerplexityComet I have to say, I still love @diabrowser too and use it daily. I'm in a strange world now where I find myself using two browsers and loving both of them for different reasons. I think I might just be a two-browser kind of guy going…
Great interview! Wanted to quickly add on a recent exp. that might help with the should ppl be encouraged to pursue CS point. Was vibe coding a mobile app and repeatedly running into an error. Then I asked the AI if a modal within a modal is allowed in that framework in the…
“There's an unspoken covenant that as a founder, you go down with the ship. For better or worse, it's changed a bit over the last year and I think it's disappointing, to be honest.” Enough said. This show is everything and more on: - What really happened behind the scenes -…
the @supabase mcp server is so well built! from their apply migrations tool to being able to create and manage queues and vector indices through extensions, from within your AI IDE like @cursor_ai. Very cool and drastically decreases time to develop new features.
All your bookmarks in 1 spot and easily accessible and searchable. #justloftit Checkout @justloftit for more details! DM if you want early access
Context <--> LLM performance relation. Great watch
Introducing our latest technical report: Context Rot - How Increasing Input Tokens Impacts LLM Performance Our results reveal that models do not use their context uniformly. full report in replies
Watching @Amplitude_HQ stats about your new app performance with #F1TheMovie movie bgm just hits right...
What if your bookmarks didn’t just sit there? What if they resurfaced when you needed them, whispered insights back to you, and helped you connect the dots? That’s what we’re building with Loft — a smarter home for everything you save. Coming soon. @justloftit
A late-night idea. A 15-day build sprint with @guptapallavi_ @sharanbabu2001 and @mohamedrasheq Now—we're almost ready. Launching soon 🚀 Follow: @justloftit #BuildInPublic #Startups #ComingSoon #AI #TechTwitter
Looking forward to!
"We believe in crafting simplicity." Looking forward to this weekend for our big release. Learn more at loftit.ai