Aniket Chauhan

CS Undergrad | Exploring Full Stack Web Development | Exploring AI | Learning daily, building for tomorrow

[Aniket Chauhan @aniketdotcom]

@AbhinavXJ 's post shows why prompt injection matters.🙃

A malicious prompt can make an AI ignore instructions, expose sensitive data, or take unintended actions.

As AI agents gain access to tools, APIs and private data, guardrails are essential.

#AI #LLM #PromptInjection

[abhinav @AbhinavXJ]

😭😭😭😭😭

[Aniket Chauhan @aniketdotcom]

Chain-of-thought made LLMs better at breaking complex problems into smaller steps, planning, and reasoning toward an answer.

But there was a catch: reasoning with no access to external data can still produce confident hallucinations.
Hence comes ReAct into the picture

#LLM #AI

[Aniket Chauhan @aniketdotcom]

Round 2 🚀

Qualified for Round 2 of the Adobe University Hackathon 2026!

Excited to see what the next round brings.

#AdobeHackathon #Hackathon #BuildInPublic

[Aniket Chauhan @aniketdotcom]

@ripa_codes Great progress 🔥, btw which playlist are you following for fastapi ^⁠_⁠^

[Aniket Chauhan @aniketdotcom]

@ripa_codes 50ish, but i looked at the solution for 49 of them tbh

[Aniket Chauhan @aniketdotcom]

Ever wondered why GPT models doesn't wait to finish before showing its answer?

Because waiting feels broken.🔃

Streaming sends tokens as they're generated, making AI feel responsive and conversational instead of making you stare at a blank screen.

#LLM #GenAI #UX

[Aniket Chauhan @aniketdotcom]

The previous post ended at embeddings. That's only half the story.
LLMs generate tokens one at a time. Recomputing every previous token for each new one would be painfully slow.
That's why they use KV Cache—reusing past attention computations instead of recalculating them.
#LLM

[Aniket Chauhan @aniketdotcom]

Most people think LLMs read English...
They don't.
Your prompt is tokenized, every token becomes a numerical embedding, and that's what the model processes before predicting the next token.
The chat interface hides a lot of interesting engineering.
#LLM #GenAI #LearnInPublic

[Aniket Chauhan @aniketdotcom]

@ripa_codes congratulations man , The dedication is surreal !

[Aniket Chauhan @aniketdotcom]

✅ Demo Link: https://t.co/r5Rq66XUV5
🔥 Repo Link: https://t.co/Tc26dIl9Gl

[Aniket Chauhan @aniketdotcom]

Finished building a mini project while learning LangGraph.

Built an AI Article Generator using an iterative workflow: Generate → Evaluate → Optimize.

Deployed it so anyone can try it.
Building projects makes these concepts click far better than reading docs.

#LangGraph #LLM

[Aniket Chauhan @aniketdotcom]

@__aman_verma looks clean ✨

[Aniket Chauhan @aniketdotcom]

Finally wrapped up the core LangGraph workflows...

Sequential → Parallel → Conditional → Iterative ✅
Iterative stood out the most. Having an agent generate, evaluate, improve, and repeat feels much closer to real problem solving.

Time to build.

#LangGraph #AI #Python

[Aniket Chauhan @aniketdotcom]

@Amoghjaiswal_ Yea it amazes me too how attention lets the model weigh every single token's relevance to others, turning each embeddings into rich contextual understanding for much better generations

[Aniket Chauhan @aniketdotcom]

Most people think LLMs read English...
They don't.
Your prompt is tokenized, every token becomes a numerical embedding, and that's what the model processes before predicting the next token.
The chat interface hides a lot of interesting engineering.
#LLM #GenAI #LearnInPublic

[Aniket Chauhan @aniketdotcom]

The word "bank" can mean a financial institution or the side of a river.

A token embedding can't tell the difference.
Transformers refine embeddings using surrounding context until each token has the right meaning.
That's where much of an LLM's intelligence comes from.
#LLM #AI

[Aniket Chauhan @aniketdotcom]

@ripa_codes Good going bro! btw What are you doing to practice for oops 😶‍🌫️

[Aniket Chauhan @aniketdotcom]

@dubeytech_ @striver_79 @takeUforward_ Congrats 💯

[Aniket Chauhan @aniketdotcom]

@utkarshj014 looks clean

[Aniket Chauhan @aniketdotcom]

@dubeytech_ @GrowthX_Club @NousResearch Looking forward to what you're cooking up ;-