
Tweets/ RT on AI, #ProductManagement, Tech & Startups. CXO. Finance. đź’™ Cricket. Not here to give/ take offence. All commentary personal.
Delhi
Joined June 2009
- Tweets11.4K
- Following2.1K
- Followers604
- Likes1.8K
In my recent 50 presentations that I would have made with AI, the final touch and the last edit version has always been handmade. No AI can still give me the confidence of having reviewed and read every slide and giving it a handcrafted finish.
34
Sreeram Narayan retweeted
'Working Here Is Joyless & Demoralising': Microsoft Employees' Open Letter Accuses Company Of Putting AI Ahead Of People
freepressjournal.in/tech/wor…
Cos. preparing to register as consent managers from 13 November still have no settled revenue model, with the industry split over whether data fiduciaries, technology companies, consumers or the government should pay
20
Falling Markets | one of the good interviews recently i have seen on where the stock markets are heading , no doubt its coming from @dhirendra_vr youtu.be/z8dLQv22_VA?si=iA78…
Good and simple analysis @fayedsouza
Every PM case study in interviews I review now looks great.
Clean structure. Sharp visuals. A roadmap. Built with AI in an evening. It no longer filters anyone out.
Inside teams too - What separates PMs now sits upstream of the output:
→ Did discovery start with real data and customers, or a prompt?
→ Which ideas did you kill, and on what evidence?
→ Where are you concentrating time and money for a 100X return?
Most PMs can build anything now. Few can decide what deserves to be built.
That's the Super PM. The top 10%.
Where are you?
1
43
Sreeram Narayan retweeted
This is how it's done
We instituted an official AI Writing Policy at @Clay.
Massive thanks to @sophiebits on our engineering team who wrote this.
Originally this was just for eng, but other teams found it so helpful we expanded it company-wide.
Here are the four guiding principles:
1. You must stand behind every idea and sentence
It is your responsibility to make sure that the entire document is representative of your own thoughts before you share it.
2. Writing is thinking
Spending time on the writing process teaches you more about your topic. If you circumvent this process, you will walk away with a poorer understanding of the subject matter.
3. More time should be spent writing a document than consuming it
If you generate a document from a short prompt then ask your readers to go through the longer output, you are disrespecting their time. They can talk to ChatGPT themselves if they want to.
4. Longer is not better
AI makes it easy to generate long docs, and it loves padding them with sentences that say nothing. If you're producing docs from a short prompt, consider just sharing the prompt.
You can read (and borrow) the full policy below. Hopefully we all communicate a bit more clearly now :)
Sreeram Narayan retweeted
BREAKING: Zuckerberg just described the near-future of computing, and it has nothing to do with screens.
His call: within 5 years, every person on Earth has a personal AI that knows their life, understands their goals, and does the work for them.
Not a chatbot. Not a search box.
An agent that is deeply, intimately embedded in how you live.
When the person building the social layer for 3 billion people says this is where we're going, it's not a prediction.
It's a roadmap.
Is the personal AI agent the most valuable product anyone can build right now?
Sreeram Narayan retweeted
DPS RK Puram Mafia keeps giving.
> Rohan Nayak
> spawn in Delhi
> DPS RK Puram → IIT Kharagpur
> build Cube26, PayTM acquires it
> quit Paytm to start Pocket FM
> build an audio entertainment platform
> neither investors nor users buy it
> 10K downloads, fail 11 times
> run out of money
> discover serialized fiction
> 8-12 minute stories, built around cliffhangers
> DAU explodes from 25 min → 150+ min
> product-market fit, crack US market
> revenue jumps at $200M ARR
> $103M Series D at ~$750M valuation
> go all-in on AI content production
> 25K → 2.5M+ hours of content/year
> adds $250M net-new ARR in one year
> 550K creators, 250M+ listeners
> $500M ARR
from 10,000 downloads to $500M ARR - all because one guy got bored in metros.
Rohan Nayak for you, ladies and gentlemen.
Introducing Sherpa: the most advanced fiction writing AI
We accelerated from $250M in ARR to $500M because Sherpa helped increase content production by 1200% in 1 year
Sherpa was trained on 5.5B hours of playtime with minute by minute dynamic retention data.
550K+ creators have produced 2.6M hours of content annualised using it
Pocket FM is like Netflix for audio-only dramas, with our own pool of one-person studios.
10% of eligible writers on Pocket FM make >$200K
One blockbuster produced >$100M in revenue
3 writers have become millionaires in <2 yrs
We built Sherpa to enable anyone to make >$1M by writing world-class fiction stories:
1. The Idea: Drop a 1-2 sentence concept. Sherpa interrogates it like a veteran editor on tension, stakes, and psychology
2. World & Characters: It builds out the complete lore, tone, and character psychologies
3. Sub-Plot planning: Breaks the premise into arcs, arcs into episodes, and episodes into scenes
4. Scene-by-Scene Generation: Outlines and drafts entire episodes, with you able to steer, rewrite, or override anytime
5. Editorial Review: Stress-tests every draft for pacing, engagement drop-offs, prose, and coherence before it locks
6. One-Tap Production: Pick a voice, convert to audio drama, and publish directly to Pocket FM’s millions of listeners
7. Global Scale & Monetization: Revenue-share on performance, with automatic localization so you earn across international markets
Test Sherpa for free here: pocketfm.com/sherpa
_____________________________________________
Generic LLMs fail at serialized fiction because they lack a long-horizon narrative reward function.
Sherpa solves this through three core technical leaps:
1. Narrative World Model (State Tracking & Retrieval): Context windows degrade over long runs. Sherpa constructs an evolving semantic knowledge graph tracking character states, secrets, and plot dependencies. High-speed retrieval surfaces exact context on demand, maintaining zero continuity decay across hundreds of episodes
2. Hierarchical Story Planner: When writing a 500-episode story like Naruto, you need to plan 100s of sub plots. Rather than generating linearly, Sherpa decomposes narrative across discrete levels: season -> arc -> sequence -> episode -> scene.
Rather than generating everything upfront, like a generic LLM, Sherpa uses progressive planning and dynamic replanning. As the story evolves, it identifies what changed, traces the downstream impact, and replans only the affected parts.
3. Prose Engine (Trained on series' retention data): LLMs write robotically, but serial fiction needs emotion, tension, pacing, and dialogue that sounds like real people.
Sherpa's Prose Engine was designed specifically for storytelling. It was built on 1B+ tokens of Pocket's own stories, trained by learning from what listeners engage with, where they drop off, and what keeps them hooked.
Feedback is taken from specialized evaluator models that measure every scene against a 40-item checklist. (Evaluator models were benchmarked against human reviewers and matched them 80–90% of the time.)
_______________________________________________
Owning distribution and creation puts us in a very unique spot.
More shows -> More data -> Sherpa becomes better -> more creator success -> more creators -> more shows
Pocket FM has already seen one $100M IP. I believe Sherpa will soon lead to dozens of single-person studios creating billion-dollar shows.
Most people are scared of AI but I think it'll unlock more human creativity, help creators earn more, and bring the next great IPs to life. This will create millions of jobs and new income streams.
One Rate, 3 Realities: What UPI's MDR Actually Means for Retail
UPI's zero-MDR era ends October 15. A 0.4% fee now applies to P2M transactions above ₹2,000, capped at ₹300 so anything above ₹75,000 effectively pays less than 0.4%.
The number everyone quotes roughly 4% of merchant transactions cross ₹2,000 is true, but it's an average, and averages hide category-level reality:
→  Mobile & Electronics retailers report 97–98% of their transactions already sit above ₹2,000 (AIMRA, ORA)
→  High Street Fashion & Lifestyle retailers have flagged real concern, though no precise industry figure exists yet (RAI, Clothing Manufacturers Association of India)
→  Kirana stays fully exempt below ₹1 lakh/month, regardless of ticket size
Same rate. Three very different realities depending on what you sell.
On the fee itself: estimates put the resulting pool at ₹5,000–10,000 crore/year by FY28. A reported (not yet officially confirmed) split allocates roughly 40% to issuing banks, 30% to acquiring banks, and 30% to TPAPs. The platforms that have run UPI at zero margin for six years. The stated purpose, per RBI and the Finance Ministry, is funding infrastructure, security, and UPI's continued expansion.
1
47
Dear @HDFC_Bank @HDFCBank_Cares Whoever designed your Virtual RM support - needs to get a downgrade. It makes it impossible to reach the RM when needed, all you get from them is missed cal logs like these. Hope someone is measureing and improving CSAT on this metric - Contact Ratio and Customer survey. I have a so called imperia account status but the experience is quite like imperial times.
1
89
Why MDR exists, where it goes, and the real risk worth watching ? My views
UPI has run on zero MDR since 2020. Someone has always paid for that just not visibly. Banks and PSPs have absorbed the infrastructure cost: uptime, fraud detection, settlement, the compute behind processing 23+ billion transactions a month.
The government's incentive scheme was supposed to cover the gap, but industry estimates on what it actually costs to run UPI have consistently run well above what's been budgeted for it.
Now -> 0.4% MDR, capped at ₹300, on P2M above ₹2,000, is the mechanism built to close some of that gap.
Money shudnt go to the government but it must flow to banks, PSPs, and NPCI, the same entities carrying the cost today.
The open question is what it does to merchant behavior at the top end. High-ticket, thin-margin categories like fuel, jewellery, travel, insurance, education, hospitals are exactly where a flat percentage fee bites hardest in absolute terms, even with the cap.
These are also the categories where UPI displaced cash and cards in the first place because it was free and instant.
If a fuel pump or jewellery counter finds 0.4% meaningfully changes their math, the historical pattern is that businesses either pass it to the customer or quietly nudge them back toward cash. Neither undoes UPI as a rail but both would be a real regression in exactly the categories digital payments took the longest to win over.
1
47
Sunday reflections #AI
Two things happened in AI this week
Dario Amodei published an essay titled "We Must Pace the Frontier."
darioamodei.com/post/we-must…
Anthropic, his own company, is voluntarily slowing down how fast it pushes AI capability. Not stopping. Slowing. And he's inviting outside safety inspectors in with badges and laptops who get to publish what they find without his company editing it first.
Same week, same company published its threat intelligence report.
anthropic.com/threat-intelli…
No hypotheticals this time. Real cases. A Russian state group using AI to rebuild its own malware on a loop every time it got caught. A single hacktivist in France building a political doxxing platform that used to need a small team and a budget. A Turkish firm selling an "AI political operations" package that was used to manipulate a Malaysian election along race and religious lines.
Put those two documents side by side and you get the actual state of AI in September 2026
One says: we need to slow down before this outruns our ability to control it. The other says: it's already outrunning us, in small doses, everywhere, right now.
Every company shipping AI is now in the business of imagining misuse before someone else demonstrates it for you.
I don't think the honest response to this is fear. I think it's the same discipline good product people already claim to have and rarely apply seriously: design for the edge case, not the happy path.
Slow down enough to actually see what you're building. Move fast enough that it still matters when you ship it.
That tension isn't going away. It's the whole job now.
1
188
This lands right around Anthropic's IPO buildup (~$2T valuation) and ongoing "who's the safety-first lab" positioning battle against OpenAI.
A CEO publicly saying "we're the responsible ones, please regulate my rivals too" is also a competitive move as it pressures OpenAI (already under fire for the Hugging Face incident) while costing Anthropic little.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
1
74
The obsession with being a “category creator” is overrated.
Apple is perhaps the best reminder.
Apple didn’t invent the MP3 player.
It built the iPod.
It didn’t invent smartphones.
It built the iPhone.
It didn’t invent tablets.
It built the iPad.
It didn’t invent smartwatches.
It built the Apple Watch.
It didn’t invent wireless earbuds.
It built AirPods.
And it certainly didn’t invent foldables.
There’s a pattern here.
Being first to market means you start the race first. Being first in the customer’s mind means you can redefine where the finish line is.
Early movers spend years proving that a technology can work. Great product companies wait until they understand how it should work.
They remove friction. Simplify the proposition. Nail the experience. Build distribution. Connect the ecosystem. And then market the product so well that consumers occasionally forget someone else invented the category in the first place.
That’s Apple’s real superpower.
Not category creation.
Category interpretation.
1
1
107




















