Priyanshu Jha

AI/Backend Engineer | Building AI agents, voice AI & backend systems | Python • FastAPI • LLMs • RAG • PostgreSQL | Open Source

[Priyanshu Jha @Codewithjha]

GPT-6 Astra:

98% FrontierMath Tier 4
100% ExploitBench
92.7% ScreenSpot-Pro
1.05M token context

And it can use your computer.

Browse. Code. Research. Test software. Execute workflows.

AI is moving from answering to doing.

The agent era is getting real.

#openai #astra #llm

[Priyanshu Jha @Codewithjha]

@tiya_decodes At this point Claude subscription is the cheaper employee 😂

[Priyanshu Jha @Codewithjha]

@SaraDiscovers Yes worth it but all depends what price are you getting that

[Priyanshu Jha @Codewithjha]

@hi8esh_ Actually dude!!

[Priyanshu Jha @Codewithjha]

Hot take:

AI engineers are learning too many frameworks and not enough engineering.

LangChain, LangGraph, MCP, vector DBs are useful.

But bad APIs, databases, evaluation and observability will still break your system.

What are we underestimating?

[Priyanshu Jha @Codewithjha]

@FaizalSukaimi Looks good, will check this.

Though I am too working on a chatbot which is primarily focused for AWADH region, maybe you don't know where is awadh. So lucknow and nearby places is awadh so the chatbot will answer in awadhi, hindi, urdu and english and that region speaks that.

[Priyanshu Jha @Codewithjha]

@FaizalSukaimi Umm, interesting. Is it open source?

[Priyanshu Jha @Codewithjha]

@DavidOndrej1 You're thinking very small man, go for bigger things. You have Astra man!! 🙂

[Priyanshu Jha @Codewithjha]

@tiya_decodes Let's connect 😃

[Priyanshu Jha @Codewithjha]

@neolaj Damm, growth is insane.

Would love to connect with you.

[Priyanshu Jha @Codewithjha]

@DineshSwami001 I guess everyone knows this, if someone is on x and following tech news then he/she must knows what is the capability of a particular model.

[Priyanshu Jha @Codewithjha]

@CharuMitraDubey Count me🤚

[Priyanshu Jha @Codewithjha]

@animatrix01 Which model are you using?

[Priyanshu Jha @Codewithjha]

Nobody talks about this part of building software:

- You spend 20 minutes writing the feature.

- Then spend 2 hours figuring out why it doesn't work.

- Then discover the problem was one missing environment variable.

Debugging isn't a side quest.

It's most of the job.

[Priyanshu Jha @Codewithjha]

@yaserabbass I’m building around AI/LLMs, backend and AI agents right now, while also working on a few projects.
What are you building in SaaS + AI? Would love to check it out.

[Priyanshu Jha @Codewithjha]

I want to build my X circle with more people who actually build things.

Looking for:

🤖 AI/LLM/Agents
⚙️ Backend & Python
🗄️ Data/Databases
🧠 ML/GenAI
🚀 Founders & SaaS builders
🛠️ Open-source contributor
🏗️ community builders

What are you building right now?Drop it below👇

[Priyanshu Jha @Codewithjha]

@hi8esh_ Agreed. Every feature adds maintenance, complexity and cognitive load. The best products are often the ones that remove more than they add.

[Priyanshu Jha @Codewithjha]

@XFreeze Damm, crazy man

[Priyanshu Jha @Codewithjha]

@joshforweb This is actually an interesting concept. How are you handling the bidding/ranking logic behind the scenes?

[Priyanshu Jha @Codewithjha]

@gpusteve Are you guys hiring for any backend or AI interns?