Puneet Khurana

@EquityValueIn @amitmantri Humbly , disagree with a lot of it. AI is not magic bullet. But a lot of what is ‘believed’ can’t be done with AI is in large portion about what ‘we don’t know how to’ get it done via AI. It’s not 100 but it’s not 20 either as many investors believe. Closer to 80 if you ask me.

Was reading the latest Investor Note by 2point2 Capital's @amitmantri , few things that really resonated with the way I think about differentiated insights in the world of AI One thing that AI has done is democratized research for anyone who is willing to read and learn , but the issue is what most people don't understand yet is, more data does not mean more insight In fact , it can mean quite the opposite , I have seen AI in some ways make investors lazy on core fundamental investing processes which include reading long annual reports , credit ratings , going on ground , creating mechanism for channel checks , understanding the cycle of the business Now you see most people just trying to engineer prompts to shorten the research cycle , and focus on data gathering rather than diffrentiated insights we need to realize that comes at the cost of the depth that actually derives the conviction that helps us to actually own businesses through an earnings cycle , shortening the research cycle will lead to no depth in research , especially when times are tough , you can build the best predictable model with huge DCF's with Claude predicting earnings to the last decimal , but what will you do when the stock falls 20% in a day , and falls 5% the next day ? None of the data points that seemed to provide the illusion of complete research will help you survive that , but just core depth of research , on ground insights and conviction will , none of which can be extracted from AI Use AI that enables you to improve functionality of research , but trying to use it to build a case for conviction and making AI as a large part of the investment thesis wouldn't be a good idea in my opinion On the contrary , I feel the Lazy investor that uses AI tools optimally to help rather than fully ideate has a bigger edge , as this market is not short of fancy data points or models it is short of conviction and the ability to back your thesis when times are tough Use the powerful AI tools to help your individual ideation process and increase efficiencies , not to be dependent for things that are the most powerful in the human mind , the ability to think , do not surrender that !

Nothing is too early or too late. This changes almost every couple of weeks. Gemini improved so much then crashed. Claude was up there for long. Last week or 2, I am noticing GPT 5.4 thinking giving much better response than opus 4.6. Next model, next few days again it will change.

Maybe said it too early. This weekend, for some strange reason, quality of outputs have dropped drastically with token usage increasing exponentially. Don't say it man. Just 'drishti'/'nazar'.

@DustinDavis Me too.. A bugging problem.. Moving to GPT 5.4 now for Openclaw setup

I'm seeing this daily with Anthropic on my Claude Max subscription in OpenClaw: "The AI service is temporarily overloaded. Please try again in a moment." Anyone else?

@NYDoorman @openclaw There is nothing you can do. Its Anthropic and its restrictions on third party apps. I am going to move to GPT...

I feel like laughing on people who things that Openclaw’s advantage is just file management and connectors and declare RIP whenever some new features are dropped. Either it’s just engagement farming or they are serious idiots who don’t understand the reason you need Openclaw. Both reasons can’t be excused.

My orchestrator agent with ‘Claude’ snitching on my data agent in ‘GPT’

@a_basumallick 🙏🏻

If you have not read this - read it now. Brilliant. "The market is a machine that transfers money from people who need to look smart today to people who can stand looking wrong for a while. Not wrong in the sense of reckless. Wrong in the sense of unpopular. Unexciting. Quiet."

@a_basumallick Online karlo, to hum bhi gyan le le thoda :)

We are planning the next VP Kolkata session on Sunday, 8-Mar-26 at Hotel Lindsay (opposite New Market main entrance). Time: 9:30 AM. For those who have not attended before, we discuss investing, pay for our own breakfast and more or less leave by 12-12:30 pm. There is no pre-registration required.

@suru27 Will do that too. 🙏🏻

@PuneetK009 Good one Puneet bhai. Still, lazy Puneet can't beat data Puneet :). High time you must write something out of your data crunching insights

Haven't written in a long time. Ironically, AI freed up some time. Here's a first X article on why doing nothing might be the hardest skill in investing.

@tusharbohra lol. Good point :)

@PuneetK009 On an interesting note, as soon as you were doing nothing (had free time).... you produced a gem of an article! :-)

@fiscalSapien Thanks. Glad you found it helpful

@PuneetK009 Excellent writing...this should be read every week.

@Feluda71 Thanks. I am glad it was helpful.

@PuneetK009 Excellent writeup..specially today..reading couple of times :) Thanks a lot

The biggest mistake people are doing currently while making agentic systems is thinking more and more agents. I thought the same. This is a valid thinking for human organisations. More the specialist the better it is. But it’s wrong when it comes to agents. The loss of context and information between agents is very real and diminishes output quality. Thinking like human organisation to design a AI org is single biggest mistake you will make if attempting one.

So many sleepless nights setting up the Agent organisation. Desperately need a 2-3 days zero screen time .

@niteen_india IT aata hai nahi aapko, aisi hi gyan de rahe ho . Farzi CFA ho aap 😉

It’s only a matter of time before this hits Indian IT companies. Period. Many seem to be in denial mode in my below post. Some are questioning my background, calling me a CFA with no understanding of IT. For clarity: I have spent 25 years working inside IT companies, both services and product firms. I started with coding on legacy systems, moved in different roles. I have seen cycles of boom, slowdown, restructuring, offshoring waves, automation shifts, and cost resets. I have seen the best of this industry, and the worst. What is building now is not a routine. It’s a structural shift. Companies will survive, they always do. But survival will come with aggressive cost rationalization, reprioritization of talent, and a reshaping of demand. Headcount models will change. Skill premiums will change. Hiring patterns will change. Employees should prepare, not panic, but prepare. The storm is large. Ignoring it won’t make it smaller.

It’s somewhere between 20 pc to 90 pc :) On a serious note, it’s about how the AI is guided. I don’t doubt that it can do a lot of heavy lifting but the key issue is that it gives a lot of false sense of accuracy to people who don’t understand the domain . People who understand the domain iterate and improve over it and takes it to its capabilities but are also aware where it is goofing up. In research , it’s not give and forget mode.

@PuneetK009 What is the likelihood of it being good ?

If you think your AI agent is giving you amazing research output and 'institutional quality' research. Bet on it. Put some money and see how good or bad it is.