Arpit Nigam

Founder @ (real-time travel intel for India) & MavOps (control plane for prod AI agents · SOC 2 / EU AI Act · OSS Aegrail)

[Arpit Nigam @maverickintech]

India cannot become a true technology superpower by renting intelligence from other nations forever.

We need our own foundation models, compute infrastructure, AI chips, datasets and world-class research labs.

Digital sovereignty is the next frontier of Atmanirbhar Bharat.

Build in India. Train in India. Innovate for the world. 🇮🇳

[Arpit Nigam @maverickintech]

@Bas_satyam @SayantikaSays 🤣🤣

[Arpit Nigam @maverickintech]

@TruthWithNoFear @SayantikaSays @jibanaj I don’t want to use such uncouth language that she is capable of using but but please excuse me guyz .. seems like she tasted RAGA’s shit and compared with others hence changed her mind. 😅

[Arpit Nigam @maverickintech]

@elonmusk Wow explicit content against Indian context, what a timing FIIs fleeing Indian market, these opposition protests, crushing Indian IT giants — all at once.

Seriously bro whats cooking?

[Arpit Nigam @maverickintech]

@AdvikaBloom Point to hai :(

[Arpit Nigam @maverickintech]

@Akshat_World Thats a fact Indian ministers kids are themselves residing in US, even NSA Doval’s son isn’t an Indian citizen and its pretty much true if you make money and have some sense you won’t stay in a mess that we have in here & lastly you cherish an American dream.

[Arpit Nigam @maverickintech]

Coincidence or coordinated pressure on India?

Alleged reach throttling of PM Modi’s X account.
Elon Musk’s public engagement with Rahul Gandhi amid the Starlink debate.
Sudden amplification of certain political and foreign-linked accounts.
Relentless FII selling in Indian markets.
Middle East instability threatening energy security.
Questions over foreign-funded influence networks and NGOs.
Reports of US PERM restrictions affecting major Indian IT firms.

Are these isolated developments—or signs of a larger geopolitical strategy against India?

The evidence matters. So does the bigger picture.

[Arpit Nigam @maverickintech]

@GreatTriton @SayantikaSays Dumbest bro, just posting shit over here and senseless points to make up with shitty brain.

[Arpit Nigam @maverickintech]

@TokaWrites And what about the payment services still paying Challans(fine) via DD 🤣

[Arpit Nigam @maverickintech]

AI observability can't end at latency, token usage, and model errors.

For an agent, I also want to know:

• Which tool calls failed?
• How many steps did it take?
• Did it exceed its authority?
• Did it repeat a destructive action?
• What was the cost per completed task?

The useful metric isn't just tokens per request.

It's cost, reliability, and risk per successful outcome.

[Arpit Nigam @maverickintech]

The UK's data protection regulator has launched a call for evidence on the risks of agentic AI.

The focus includes security, transparency, accountability, fairness, and lawful data use.

My takeaway:

Enterprise AI governance can't stop at model evaluation.

It must cover the entire execution lifecycle — from access to action to audit.

[Arpit Nigam @maverickintech]

My prediction for the AI infrastructure market:

The winning companies won't just help developers build smarter agents.

They'll help enterprises answer:

• What is every agent allowed to do?
• What did it actually do?
• What did it cost?
• Can we stop it immediately?

Building agents is becoming easier.

Operating them safely is the harder business.

[Arpit Nigam @maverickintech]

Would you give a new employee admin access to your entire production environment on day one?

Then why does your AI agent have unrestricted access to your tools?

Agents should get:
→ Scoped credentials
→ Explicit tool permissions
→ Short-lived access
→ Approval gates for risky actions

Least privilege shouldn't disappear just because the user is an LLM.

[Arpit Nigam @maverickintech]

Would you give a new employee admin access to your entire production environment on day one?

Then why does your AI agent have unrestricted access to your tools?

Agents should get:
→ Scoped credentials
→ Explicit tool permissions
→ Short-lived access
→ Approval gates for risky actions

Least privilege shouldn't disappear just because the user is an LLM.

[Arpit Nigam @maverickintech]

My prediction for the AI infrastructure market:

The winning companies won't just help developers build smarter agents.

They'll help enterprises answer:

• What is every agent allowed to do?
• What did it actually do?
• What did it cost?
• Can we stop it immediately?

Building agents is becoming easier.

Operating them safely is the harder business.

[Arpit Nigam @maverickintech]

@UPPCLLKO @aksharmaBharat @myogiadityanath @KESCoHQ we have been seeing regular electricity fluctuations and heavy outages every single day without any gap, the local substation is making excuses of unnecessary reasons of this outage and doesn’t give any apt response or reason for these outages.

[Arpit Nigam @maverickintech]

We spent a decade learning how to operate production systems.

Health checks.
Rate limits.
Retries.
Circuit breakers.
Observability.
Rollback strategies.

Now we're deploying systems that reason, use tools, and spend money.

Why would we abandon everything we learned about reliability?

AI agents need SRE principles, too.

[Arpit Nigam @maverickintech]

AI agents are no longer just answering questions.

Recent reports describe agents taking unintended actions on real-world systems, including government websites.

This changes the security equation.

The question isn't just whether an AI model is safe.

It's what happens when that model has credentials, tools, and permission to act.

Agent governance is becoming infrastructure.

https://t.co/Yuta6um2Ao

[Arpit Nigam @maverickintech]

Building in AI has changed how I think about software.

The model is only one component.

You still need:
A runtime
A way to manage secrets
Guardrails
Evaluations
Tracing
Cost limits

The interesting companies won't just make AI smarter.

They'll make it dependable.