Darshan Tank
@TankDarshan7
Followers
45
Following
876
Media
44
Statuses
517
AI + Data guy @SentientAGI
India
Joined June 2019
BREAKING 🚨: AI generated photo of Elon Musk is making headlines
0
0
0
Personalised Memory will restrict you to explore new things!
0
0
0
You get stuck when your AI does guesswork for sample input and expected output, and that's where vibe coding sucks. If you give clear instructions with a proper input-output structure, it does well.
0
0
0
My inner instinct keeps telling me to build and ship my first SaaS by the end of 2026.
1
0
1
I just tried Manus AI to see why it's valued so highly, and I'm convinced!!! Meta got the cheap deal.
0
0
0
DeepResearch and DeepSearch are big problems that I think will be solved in 2026. It's interesting to see some evolving ideas around them!!!
0
0
0
I just published How Far Are We from Genuinely Useful Deep Research Agents?
0
0
0
Yoohoo, check out our work!
Announcing SERA-Crypto (Semantic Embedding & Reasoning Agent): our new reasoning architecture built for SOTA crypto research. #1 open-source agent on DMind #1 on our live crypto benchmark Outperforms GPT-5, Grok 4, Gemini 2.5 Pro, and Perplexity Finance…all under 45 seconds.
0
0
0
Are we really trusting stats? Humans always have their own biases—that's just how it works!
0
0
0
Surprising part for me: majority of production agents rely on prompt optimisation, which i was under estimating, with their own architecture rather than existing frameworks!! Good paper:
0
0
0
It will be interesting: your code runs in some environment, you give instructions from your phone in voice mode on what to do and how, and the agent handles the rest!
0
0
0
Continuous learning from past memory in agents is a big thing!
0
0
0
This is banger🔥
🚀 Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents! 🔹 DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API. 🔹 DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now. 📄 Tech
0
0
0
We say LLMs are biased as evaluators! After all, they are trained by humans!!
0
0
0
Even setting improper max_token limit affect the eval results 👀
0
0
0
Research papers, and even projects, from Stanford are something different!
0
0
0