Rajeev Kumar

Product Manager, technology enthusiast, passionate about creating cool products

[Marc Andreessen 🇺🇸 @pmarca]

Ultra amazing —>

[alon @alonzuman]

Introducing Pops: AI games with real friends

[Brett Caughran @FundamentEdge]

A big pivot from Ken Griffin on AI:

“Number one is, in the last few months, there has been a step change in the productivity of the AI toolkit. It is profoundly more powerful than it was just nine months ago.

And for us at Citadel, that has allowed us to unleash a much broader array of use cases for AI. And it has been really interesting to watch, to be blunt, work that we would usually do with people with masters and PhDs in finance over the course of weeks or months being done by AI agents over the course of hours or days.

These are not these are not mid-tier white collar jobs. These are like extraordinarily high skilled jobs being, I'm going to pick a word, automated by agentic AI. And I gotta tell you, I went home one Friday actually fairly depressed by this because you could just see how this was going to have such a dramatic impact on society.

When you witness it in your own four walls, when you see work that used to be man years of work being done in days or weeks, it's like, wow, like that's the first time I've seen real impact in our four walls.”

This echoes my own experience with agents and the conversations I am having with students, friends & clients. The toolkit has dramatically transformed and it feels like in finance, for the first time, AI is real.

[Satya @solidview]

Gen Z is asking the right questions and that is very good for Bengaluru.

@RishvanjasR must watch 4 minute video.

See it. Share it. Sort it.

[Ghar Ke Kalesh @gharkekalesh]

On the streets of Mumbai, Minister Girish Mahajan was speaking about women's reservation... A woman, frustrated by the traffic jam Confronted him infront of Everyone

[Ramanuj Mukherjee @law_ninja]

I have hired lawyers for the last 15 years. I have helped law firms hire good talent too. For most of those years, I knew exactly what I was looking for.

Give me a candidate who can research a legal issue thoroughly. Who can draft a clean contract and understand the second order effects of changing a clause.

Who knows the Companies Act cold. Who can write a memo that a partner does not have to rewrite from scratch but just fine-tune.

That is what I screened for. That is what I trained students for.

That is what every law firm in India screened for.

Day Zero interviews at NLUs test your knowledge of the Contract Act and try to understand what you did in your internships.

Internship PPOs go to the intern who submits good research memos. Pre-interview screening rounds often include a drafting test.

Research. Drafting. Legal knowledge.

These three things decided who got hired at every good law firm in India for 30 years.

AI made all three of them free.

A 2nd year law student with ChatGPT can produce a research memo today that would have taken a strong 1st or 2nd year associate two full days in 2022. And AI would do it in better style.

Law students using AI can draft a shareholders agreement that reads like it came from a 3rd year associate.

They can answer a question on Section 230 of the Companies Act and related recent NCLAT judgments more accurately than a lot of lawyers with 5 years of experience, provided they have an AI agent open in the moment.

I see this every day at Lawsikho. We train thousands of lawyers. The output quality of an average student using AI is now hard to distinguish from a good student working without it.

This breaks hiring. But not in the way people think.

The common fear is that bad candidates will fake their way in. Use AI to look good. Get the job. Then fail.

That is not what happens. They use AI on the job too. And for most junior work like research, drafting, due diligence and contract review, it works fine. A mediocre lawyer with ChatGPT produces acceptable output for most of what a first year associate does day to day.

The real problem is bigger.

When every candidate's output looks the same, you lose the ability to find the ones who are genuinely exceptional. The ones who can spot the mistakes AI makes. Those who can think further than what the AI already thought. The ones who can predict what kind of trouble a perfectly formatted contract or a seemingly polished petition will land the firm into.

You find out you hired the wrong person 7 months down the line. That is 7 months wasted.

Also, how do you hire the ones who will become your best senior associates in 5 years? The ones who will bring in clients in 10? The ones who will make partner?

You are not hiring bad people. You are hiring randomly. Because the signals you relied on for 30 years no longer separate anyone.

A owner of a boutique law firm told me recently. "I used to rank my interns within the first week. Now they all look the same to me. I give PPOs based on who seems like a good person."

Who seems like a good person is not a hiring strategy.

So what do you actually screen for now?

We spent the last few months rethinking this. Not just for law firms we help hire talent, but for our own hiring at Lawsikho and Skill Arbitrage. Three changes made most of the difference.

The first change. We stopped sending take-home tests. We started sending teardowns.

Every take-home assignment you send a candidate today is an open-book exam where the book writes the answers for them. You cannot outrun this.

So we flipped it. Instead of asking them to write a memo, we send them a memo that AI already wrote. A 4-page legal opinion on a real Companies Act issue. We tell them upfront. We wrote this with ChatGPT in 10 minutes. It may be brilliant. It may be garbage. We are not telling you which.

You have 8 hours. Send back every single thing that is wrong, weak, missing, or dishonest in this opinion. Explain why.

This test does what no other pre-screen does. Asking AI to critique its own output produces bland, hedging feedback. Real critique requires actually knowing Indian law well enough to see what the AI left out. A 2nd year student who catches 3 genuine errors is more valuable to us than a 5th year student who catches zero. Every AI memo has different flaws, so the test cannot be rehearsed.

The second change. We stopped testing skills AI made free. We started testing skills AI made rare.

Everyone is focused on the skills AI now does for you. Research. Drafting. Memo structure. Case law recall.

These were the skills law firms hired for. They are now in infinite supply to any candidate with a laptop. Their value as a hiring signal is zero.

The skills that became more valuable because AI commoditized everything else are the opposite. Problem formulation. Taste. Translation. Judgment under ambiguity.

We test these three ways.

We send candidates a 10-minute voice note of a real client speaking. Rambling. Emotional. Mixing personal and business context. Switching between Hindi and English mid-sentence. We do not ask them to draft a response. We ask them a single question. What are the 3 questions you would ask this client before doing anything legal, and why those three?

We ask this live.

ChatGPT can answer any legal question you ask it. It cannot tell you which question will bring our the reality of the client's situation. That skill is the new moat.

We give them a 15-page M&A term sheet that was mostly AI-drafted, with three subtle errors hidden inside. A cross-reference to clause 7.3 when the indemnity cap is actually in 7.4. A representation that contradicts a warranty two pages later. An arbitration seat that does not match the governing law. These are exactly the mistakes AI produces and exactly the mistakes AI cannot catch. Spotting them requires taste built from reading real contracts.

We want to see how they interview a client. We do a roleplay for this. That one roleplay tells us more than we could understand from any HR questions that one can easily rehearse.

The third change. We stopped trying to detect AI. We started requiring it.

The entire conversation about "how do we detect AI in applications" is a losing war. Detectors are unreliable. They flag real writing. They miss polished AI. Every hour a firm spends on this is wasted.

The move that changed everything for us is the opposite. Require AI. Watch them use it.

A 30-minute screen. A real legal problem. The candidate has choice of their AI open. We are watching their screen over Zoom. Solve it live.

The answer is not what we evaluate. We evaluate how they use AI.

Do they accept the first output? Junior. Do they cross-check a citation? Competent. Do they notice when the model has gone off track and re-prompt? Good lawyer. Do they ignore AI entirely for a specific step because they know it will get that part wrong? That is the professional instinct we want.

The way someone uses AI in 2026 is now the most honest signal we have of how they think. Not whether they use it. Everyone will use it. Whether they know when to trust it and when not to.

The old hiring tested: can you produce good legal work?

The new hiring needs to test: can you think, verify, translate, and judge?

Every law firm in India is still hiring the old way. Day Zero interviews still ask what Section 34 of the Arbitration Act says. Internship evaluations still grade the quality of research memos that the intern may or may not have written. Screening still includes a drafting test on a hypothetical M&A deal that ChatGPT could pass in its sleep.

These are all now testing one skill. How well can this person use AI. That is a useful skill. But it is not what separates a good lawyer from a great one. It never was.

The firms that figure this out will quietly build the best teams in the industry. Not because they found a way to ban AI from hiring. Because they stopped screening for the things AI gives everyone for free and started screening for the things it cannot touch.

[Ajay @ajaybuilds]

https://t.co/1E7nFtO2Yx

[Vijay Shekhar Sharma @vijayshekhar]

This is going, where we all dreaded AI might go.
Not just predicting will you click an ad but what will you react, seeing a post or ad or anything else.
Powerful and next-level move by Meta!

[TFTC @TFTC21]

Balaji: "Keynesianism is like communism but for wimps... Soviet communism had to send guys with guns house to house to steal 10% of all property. American Keynesianism just hits a button, inflates the money supply, and nobody feels it."

[Rajeev Kumar @rajeevkm]

@nicochristie Interested to try out

[Rajeev Kumar @rajeevkm]

@nickkokonas @threadreaderapp unroll

[Mahima Jalan ماهيما جالان @MahimaJalan2]

After finishing my Master’s in Journalism, I decided to start my own agency and create jobs instead of working for someone else.

But soon after, I got a serious reality check.

My GST application was rejected, not due to any issue in paperwork, but because the officers expected an in-person visit, just so I could "settle" things with them under the table.

Getting international payments is a headache. RBI rules make it really tough. Even if a client pays you, the payment can get flagged, your account might get blocked, and you have to fight just to get your own money.

And when you try to complain? No one gets back to you.

Yet the same ministers go on public platforms to say entrepreneurs aren't building real value.
It hurts.
Not because building a business is hard, I signed up for that.
But because when you're trying to do something right, you’re met with red tape, bribes, and zero support from the very system that claims to champion entrepreneurship.

[Murtaza Amin @murtuzi]

Dear @PiyushGoyal @PiyushGoyalOffc

I run a 100 people software company from #Burhanpur MP, I bring M$+ / yr to Bpur economy & I am the largest white collar employer in #Burhanpur.

Since you've been talking about how Indian #startups are not innovating enough, I'd like to ask

[Rajeev Kumar @rajeevkm]

@ItsAustinAI Free

[Dan Go @CoachDanGo]

You can have 1000 problems in life until you have a health problem.

Then you only have one.

[What about 2029? @DesiDude46]

I applied for a loan yesterday.
Hdfc bank started the process.
Starting from today morning, i have got 8 calls from 4 different banks.
All had details of application with hdfc.

I further enquired as how did they get such accurate details, all revealed that they got

[Vijay Patel @vijaygajera]

This is very big.

1. Here is the list of so-called Independent journalists funded by USAID-funded Earth Journalism.

This list has the names of 86 Indians who write articles on their pay roles.

Do you know where their articles are published?

To know read this thread now👇

[Balaji @balajis]

It might turn out that the GPT wrapper has more of a moat than GPT.

[Balaji @balajis]

AI came out of academic research and big tech, so it started with high prestige.

Crypto came from anonymous message boards and internet degens, so it started with low prestige.

But AI makes everything easy to fake and crypto makes it hard again.

So prestige may converge.

[The Rabbit Hole @TheRabbitHole]

Lee Hsien Loong, the former prime minister of Singapore, gives an outside perspective on Wokeness in the United States.

🔊