After much deliberation and pondering, I have finally started a newsletter. If you are a startup founder or a developer building a new product or trying to scale your application, these emails will help you avoid the pitfalls I came across.
😄
https://t.co/ampamCZney
I tested models from llama3.2 to qwen, benchmarked their summarising skills on real articles, and found the sweet spot where local LLMs become genuinely useful. The biggest quality jump? Not from switching models—from better prompt engineering.
https://t.co/W6reWxnHdC
I no longer read Medium articles, I ask Claude or ChatGPT to summarise it. Coz it's so frustrating to read articles nowadays. They follow this weird similarish tone and feels like it's written by the same author.
A recent interaction I had
Doing @HustleBandBLR's 2nd outdoor jam at MakeMyGarden, this time with a 'festive twist'. Let's usher in the festival season with music.
Everyone's welcome. Especially bathroom singers.
Only thing banned is staying silent :)
RSVP fast. 20 slots left: https://t.co/sNjLOPE0OH
Join Sibin Bhaskaran at #FifthElephant2025 for
“Real-world Lessons from Multi-Agent Architectures”
Latency, accuracy, LLM testing, and scaling agentic AI from PoC to prod.
🗓️ July 19
📍 BIC, Bengaluru
🔗 https://t.co/UsCqJ4RBdY
@SahajSoftware@fifthel@hasgeek
Just saw a tweet claiming Baahubali released 10 years ago. Didn’t believe it, so Googled, and yes, it’s been a decade already. Amazing (and a bit scary) how time flies.
Catch @sthreepi from @SahajSoftware at #TheFifthElephant '25!
"Democratizing Large Model Training on Smaller GPUs with FSDP"
🗓️July 19
📍Bengaluru
🔗 https://t.co/UsCqJ4RBdY
Learn how to train/fine-tune #LLMS using #FSDP without blowing your GPU budget.
@hasgeek@fifthel