Pinned Tweet
Movies teach us lessons, help us uncover truths about ourselves, act as a window into other time periods and places, and give us empathy. What other medium delivers these gifts in 120 minutes or less? Yes! I am a Cinephile🤞
5
Every automation builder learns this the hard way:
Stop using an AI agent for what an IF filter does in 2ms for $0.
Deterministic logic for routing.
AI only for ambiguity.
The most reliable agent systems are 80% static code and 20% LLM.
#AIAgents #Automation
31
In the West, EVs mean $60k cars. In India, 85% of EV sales are 2 & 3-wheelers.
Yet while 2-wheelers make up 78% of charging sessions, cars devour 67% of grid energy.
Can our grid handle mass electric cars, or is India forever 2-wheeler first?
#EV #CleanTech #India
22
India built a $10B quick-commerce market on 10-minute delivery. The trade-off? Over 2 lakh local kiranas shut down.
Convenience won today. But what happens to consumer prices when neighborhood competition dies and apps double platform fees?
#QuickCommerce #Startups #India
9
India’s new E-Commerce Rules take aim at fake discounts: platforms must now display the lowest price from the last 30 days alongside any sale price.
No more inflating base prices to slap '70% OFF'.
Will festive sales finally become honest?
#Ecommerce #Tech #India
10
Everyone hypes 3nm AI chips, but India's Semicon 2.0 is doubling down on 'boring' 28nm+ legacy silicon, equipment & chemicals.
Unsexy? Yes. But 28nm runs 80% of cars, defense & grids.
Smart foundation first, or are we subsidizing yesterday's tech?
#SemiconIndia #Tech #AI
7
Vector search treats data as isolated islands.
Fine for simple lookups. But when an agent needs multi-hop reasoning across docs, cosine distance fails.
Production RAG doesn't just need embeddings. It needs GraphRAG and relationship graphs.
#AIAgents #Engineering
14
40% of queries in agent workflows are semantically identical.
Exact string caching misses them. Semantic caching via vector distance drops latency from 2,000ms to 20ms and cuts token bills in half.
Before upgrading models, cache intent.
#AIAgents #Engineering
8
Building custom REST API wrappers for AI agents is dead.
Before Model Context Protocol (MCP), connecting tools across models was an M×N nightmare of glue code.
MCP standardized the tool layer. Stop writing custom API wrappers.
#AIAgents #SoftwareEngineering
1
2
44
Everyone’s flexing neon and synth-pop, meanwhile I’m out here flexing the real ’80s power move: green phosphor screens, a massive 10MB drive, and floppy disks for days. 💾📟
#RetroTech #80sAesthetic #DeskSetup
2
89
Testing AI agents with playground prompts isn't QA. It's a vibe check.
A single prompt tweak silently breaks 20% of tool-calling edge cases.
If your CI/CD doesn't run automated trajectory evals before merging, you're flying blind.
#AIAgents #SoftwareEngineering
24
POV: debugging at 2AM while Walter White watches you cook. 💀
One screen: code.
Other screen: “I am the one who knocks.”
Peak tech life. 😂💻
19
Companies spend millions on AI agents, yet only 7% have data ready for them.
If your CRM fields are dirty and database schemas are messy, a smarter LLM won't save you. It just fails with higher confidence.
The real AI moat isn't intelligence. It's data hygiene.
1
23


















