Ankit Jain

A podcast where I had a wonderful time discussing a spectrum of topics on Product Management: - some non-obvious Aha moments in Product career - early career level-up map & hacks - experiment hacks - 0-to-1 journeys & scale-ups - B2B vs B2C flavours in PM youtube.com/watch?v=TTrP8F…

Wiring analytics tracking into your app is a powerful forcing function to do a thorough review of your app's modular structure, message passing brittleness and dead-end try-catch-error blocks.

Product is never deep & wide enough moat when you have a determined and resourceful competitor (Snapchat vs Facebook). AI agent loops & tokens war chest powered brute force make copying products a weekend chore. The irony is that human-to-human relationships (Enterprise Sales, Account Management, Support) may prove to be the only sustainable & defensible moat. Relationships that involve six months of human-to-human emails, meetings and careful negotiations to secure a deal - not an AI agent sending out emails.

Drafting a good Positioning is only half the job. Making it stick & preventing fragmentation/dilution/drift is the remaining half. Great set of practical advice from @aprildunford ! Source: youtube.com/watch?v=D9pf_i…

Using AI agents to work on several problems at once is very draining for the human mind. This exhaustion happens because there is a limit to how much information a person can hold in their head at one time. High-intensity work with these tools can leave even experts feeling completely "wiped out" by mid-morning. ----------- Food for thought: 1. If AI is meant to make humans more productive, why are the people using it the most working harder than ever before? 2. How can a person find their "new limits" to avoid burning out when their digital tools can work 24 hours a day? 3. What happens to the quality of a person's thinking when they "pop the stack" of their mental capacity by managing too many tasks at once? 4. In what ways does the "novelty" of these new tools encourage people to work at a pace that is unsustainable?

Rocket science can become "rocket science" for rocket scientists if it is not practiced often enough. #ArtemisII

Claude Mythos blog post leak Claude Code source code leak --------------- Reminds of "The Law of Leaky Abstractions" essay by @spolsky

Is piping `bash` commands an agentic workflow? #AI

@sir4K_zen I am wondering whether organisations will guard their iterated, optimised "process" captured into .md files as a confidential/proprietary differentiating edge or make it Open Source (like, GStack by @garrytan or Compound Engineering by @every or Paperclip by @dotta ).

@ankitjain Agree, decision design is the differentiator

Agentic AI is converting Software-as-a-Service to Process-as-a-Product (I prefer this more to the other moniker being popularised, "Service-as-a-Software"). Now you can ship your "workflow" and "process diagram" for others to reuse.

@sidbharth @emergentlabs I am curious how they got a sniff about this? Whistle blower(s)?

SCOOP: at least three major indian startup publications are working on investigative reports debunking the high ARR claims made by @emergentlabs

@nikunj @mpauldaniels All the "smart money" knows about this, but still goes along with the narrative. No GRR and NRR metrics spoken about, even at the monthly or quarterly level.

No, that would be something.. people sign up for $20/mo to get credits. They might spend some extra money for credits. Let’s take someone who put $100 extra in credits that month. So they spent $120 that month. That entire $120 is multiplied by 12 for annualized revenue run rate even though there’s no reason for that person to spend $100 again next month. Not to mention that $20 is probably zero because of free trial.

Alright let’s put this ARR debate to rest once and for all.. If ANY vibe coding startup with scale (>$5M claimed ARR) publicly discloses how they calculate it (esp sub vs. inference) and how much of it they subsidize, I’ll donate $2,500 to a charity of their choice today 🫡

A great magic show is 30% trick complexity, 70% presentation finesse. Both are needed. But one is needed more.

Like "Texting" and "Emojis" entered our general communication lexicon, "Compacting" has a likelihood too. 🙃

When you have a car that can move fast, it is tempting always to drive it at its top speed. After all, you paid for it. You've got to use it. But that's when we put ourselves and others at risk. Drive safely. Feel the joy of having that raw power at your beck and control. AI coding is similar in many ways. #AI #Coding

Providing human customer support will start counting as a differentiator and a marketing / customer acqusition cost.

Trying to solve a hard bug and ratcheting up the "intelligence" deployed: GPT-5 mini ---> Gemini 3 (Flash) ---> Sonnet 4.6 ---> Opus 4.6 Feels like escalating from junior dev to senior dev to team lead to principal engg!

The Claude team is s inadvertently speed-running the typical platform/ecosystem story of: 1. we help you build custom/niche use cases ---> 2. we help you distribute & monetize ---> 3. we learn from what succeeds ---> 4. we bake-in the success stories as native capabilities/offerings in the platform The developer ecosystem is iterating fast on top of Claude to build & evangelize new use cases. Claude is then subsuming all the winning ideas. Maybe the team already had it on their roadmap, maybe not. #Anthropic #Claude

A whole new class of dev tools will emerge to save ourselves from the catastrophic impact of a rogue AI agent executing commands that it shouldn't and ideally can't. May introduce a best practice in the industry of having a layer of redundancies in whatever the AI Agent touches - like, using "soft deletes" instead of actual data deletion to ensure information can be recovered if the Agent makes a mistake. #AI