Shrinath V

2.2K posts

Shrinath V banner
Shrinath V

Shrinath V

@shrinathv

Product coach | Startup Mentor | Trainer. Love travel and stories

Bangalore, India Katılım Nisan 2009
994 Takip Edilen1K Takipçiler
Shrinath V
Shrinath V@shrinathv·
@HedgieMarkets The smarter way would be to set the right priorities of usage - see which parts are deterministic, set common guidelines, which parts of your codebase should AI work on. Those who do that will see measurable value.
English
0
0
0
22
Shrinath V
Shrinath V@shrinathv·
I wouldn't write this story off yet. Most companies rushed to get everyone AI coding tools without constraints. They set up token usage standards, absurd if you think of it, which incentivised people to use it sloppily. Most of the initial stages were just about usage without constraint or purpose.
English
1
0
0
736
Hedgie
Hedgie@HedgieMarkets·
🦔Microsoft canceled its internal Claude Code licenses this week after token-based billing made the cost untenable, even for a company with effectively infinite cloud resources. Uber's CTO sent an internal memo warning the company burned through its entire 2026 AI budget in just four months. American AI software prices have jumped 20% to 37%, and GitHub (owned by Microsoft) is dropping flat-rate plans for usage-based billing across its products. My Take The AI subsidy era is ending in real time. The same company that put $13 billion into OpenAI and built the Azure infrastructure powering most of Anthropic's compute just looked at the bill from a competitor's coding tool and decided it was not worth paying. That is not a productivity failure on Anthropic's end. Token-based pricing is forcing every enterprise customer to confront the actual cost of running these models at scale, and the number turns out to be far higher than the flat-rate experiments suggested. This ties directly to my Gemini Flash post yesterday. Anthropic, OpenAI, and Google all raised effective prices in the last six months. Enterprises that built workflows assuming AI costs would keep falling are now watching annual budgets evaporate in months. Two outcomes look likely from here. Either enterprises scale back AI usage to fit budgets, which slows the revenue ramp the labs need to justify their valuations ahead of IPOs, or the labs cut prices and absorb the losses, which makes the unit economics worse at exactly the wrong moment. Both paths land in the same place, the numbers stop working, and somebody has to take the writedown. Hedgie🤗
Hedgie tweet media
English
1K
4K
19.6K
7.7M
Shrinath V
Shrinath V@shrinathv·
@Calls_a_Spade @howard No, bad engineering means a lot of complex code that takes forever to fix. Good engineering is about smart system design and operating principles.
English
0
0
0
35
Harsha Nagaraja
Harsha Nagaraja@Calls_a_Spade·
@howard I still have no idea how AI can make a bad engineer worse. Irrespective of the bad taste, output of AI will be an order of magnitude better than without AI.
English
8
0
5
4.8K
Howard Lerman
Howard Lerman@howard·
Everyone is obsessed with AI making a 10x engineer a 1000x engineer. The recent reductions at CloudFlare and Click have me me realize the plot is equally about the inverse: AI amplifies the *negative* impacts of poor performers. If a person with poor taste, who makes mediocore judgement calls, and doesn't properly build things customers love is able to produce 10x more work - does a company want that? Hell no! Productivity isn't just about as many people as possible tokenmaxxing. AI is a double edged sword, especially when it's used to produce net new work. If you give a bad artist a pen that can draw 100x as fast, you're going to pile up with a lot of junky artwork very quickly. And since it happens so quickly leaders are now able to see quickly who is Picasso and who is not and adjust accordingly.
Zeb Evans@DJ_CURFEW

Today we reduced headcount by 22%. The business is the strongest it's ever been. So I think it's important to be direct about what I'm seeing and why. First, I made this decision and I own it. I did it because the way to operate at the highest level of productivity is changing, and to win the future, ClickUp needs to change with it. Second, this wasn't about cutting costs. Most savings from this change will flow directly back into the people who stay. We'll be introducing million-dollar salary bands. If you create outsized impact using AI, you'll be paid outside of traditional bands. Most importantly, I have the deepest gratitude for those affected. We're doing this from a position of strength specifically so we can take care of people properly. Everyone affected receives a package aimed at honoring their contributions and easing the transition. I only see two options: wait for this to play out gradually in the market or be honest about what I'm seeing and act proactively. THE 100X ORGANIZATION The primary change is that we're restructuring around what I call 100x org. The goal is 100x output. The roles required to build at the highest level are fundamentally different than they were a year ago. Incremental improvements to existing systems won't get us there. We need new ones. That means creating enough disruption to rebuild rather than iterate on what's already broken. The common narrative is that AI makes everyone more productive. It doesn't. Many of the workflows of today, if left unchanged, create bottlenecks in AI systems. These roles will evolve. But waiting for that to happen naturally means falling behind now. The 100x org is actually heavily dependent on people - infinitely more than today. This is only possible with 10x people that have embraced and adopted new ways of working. THE BUILDERS, AGENT MANAGERS, AND FRONT-LINERS — THE BUILDERS: 10X ENGINEERS I don't think most companies have internalized what's actually happening with AI in engineering. The common narrative is that AI makes all engineers more productive. That may be true in isolation, but at an organization level - that is the farthest thing from reality. Here's what we've validated recently at ClickUp: the great engineers, the ones who can orchestrate, architect, and review, are becoming 100x engineers. They're not writing code. They're directing agents that write code. The skill is judgment. AI makes the best engineers wildly more productive, and everyone else using AI slows these engineers down. Think about it - the bottlenecks are (1) orchestration - telling AI what to do, and (2) reviewing - what AI did. Everything is leapfrogged and no longer needed. So who do you want orchestrating and reviewing code? And how do you want your best engineers to spend their time? If your best engineers are spending time reviewing other people's code, then this is inherently an inefficient bottleneck. These engineers can review their agent's code much faster than reviewing human code. The new world is about enabling your 10x engineers to become 100x. The wrong strategy is to push every engineer to use infinite tokens. Companies doing this are celebrating 500% more pull requests. But customer outcomes don't match the volume of code being generated. I call this the great reckoning of AI coding, and every company will face this soon if not already. More code is just another bottleneck to the best engineers, and ultimately to your company's impact as well. — THE BUILDERS: 10X PRODUCT MANAGERS Product management and design roles are merging. Designers that have customer focus, become more like product managers. And product managers that have intuition for UX become more like designers. The bottleneck of user research is gone. It takes us just one mention of an agent to kickoff research and analyze results. The bottleneck of product <> design iteration is also gone. The product builder iterates on their own, along with agents and skills that ensure alignment with quality and strategy. Also controversial today - I believe that the wrong strategy is to have your PMs shipping code - that just introduces another bottleneck that the best engineers will waste their time on. To be clear, PMs should be coding but they should do this in a playground to iterate, validate, and scope. That code should not go to production. Everything outside of managing systems, orchestrating AI, and reviewing output becomes a bottleneck. That's why the other roles that are critical along with these are the systems managers (to reduce bottlenecks) along with a bottleneck you can't replace - customer meeting time. — THE SYSTEM MANAGERS Ironically, the people that automate their jobs with AI will always have a job. They become owners of the AI systems - agent managers. We have many examples of these people at ClickUp. The underlying systems in which we operate are absolutely critical to get right. I think most companies are delusional to think they can iterate on existing systems and compete in this new world. You must create enough disruption so that old systems are deprecated entirely. If there's any definition for 'AI native' that's what it is. — THE FRONT-LINERS In a world that will become saturated with AI communication, the human touch will matter more than anything to customers. This is a bottleneck that you shouldn't replace - even when agents are high enough quality to do video meetings. One-on-one meeting time with customers is something that shouldn't be automated. The systems around the meetings should be - so that front-liners spend nearly 100% of their time with customers. REWARDING 100X IMPACT In a world where companies are able to do so much more with less, where does that excess money go? In our case, much of the savings in this new operating model will flow directly back to those that enabled it. We must reward people that create productivity accordingly. This aligns incentives on both sides. Plus, in a world where your best people create 100x impact, you can't afford to lose them. You should aim to retain these employees for decades. The context they have and their ability to efficiently orchestrate and review will be nearly impossible to replace. Compensation bands of today should be thrown out the door. We're introducing $1 million cash/year salary bands with a path available to nearly everyone in the company if they produce 100x impact by creating or managing AI systems. THE FUTURE Nearly every company will make changes like these. The ones that do it proactively will define what comes next. The future is not fewer people. It's different work, new roles, and better rewards for those who embrace it. We're already seeing entirely new roles emerge, like Agent Managers, that didn't exist a year ago. ClickUp is positioning to lead this shift, not just internally, but for our customers too. I've never been more certain about where we're headed.

English
95
305
1.7K
324.7K
Shrinath V
Shrinath V@shrinathv·
When I started using serious coding agents — Claude Code, Cursor, the early Codex experiments — the velocity was addictive. A vague prompt produced working modules in minutes. It felt like magic. Then I moved to systems that had to survive real production pressure. Strange things emerged. Context drifted. One model quietly altered assumptions another had established. A reviewer rejected patterns introduced two steps earlier. The code often ran, but the underlying decisions didn’t hold. This pulled me back to hard-won software engineering principles. Stable interfaces, explicit contracts, clear handoffs, and multiple layers of review exist for a reason. I’m writing about workflows this week — how value is shifting from cheap tools to the systems that coordinate them. This is the first concrete example. I built the Virtual Product Factory as a governed workflow. Every major stage produces concrete, named files as the single source of truth. One agent generates. @red-team audits risks early. Others review, guard, and test. They all work against the same shared artifacts. This approach recently helped me upgrade a Django backend, add a clean API layer, and completely rearchitect the frontend in React + Next.js for Founding Fuel — with support from just one backend engineer. Work that would previously have required a full team. I follow a test-driven rhythm. At checkpoints the system proposes changes. The agent proposes. I promote or reject. Accepted changes go into CHANGELOG.md with full reasoning. Nothing mutates without explicit human sign-off. Decisions stay traceable in the files and git history. Guardrails live in baseline files that no skill can override. The audit trail is the changelog. Unlike self-learning loops, this setup is deliberately human-led. Errors compound fast over large codebases. Human oversight is still required when systems grow. The repo is built to be forked and adapted. I’ll post links to the rest of the series (writing, learning, strategic workflows) in the comments as they go live. This is what governed workflows look like in practice. The models remain imperfect. The leverage comes from the workflow that coordinates them. AI does not remove these requirements. It makes them more expensive to ignore.
English
0
0
1
52
Shrinath V
Shrinath V@shrinathv·
The models remain imperfect. The leverage comes from the workflow that coordinates them. AI does not remove these requirements. It makes them more expensive to ignore. More on the series (writing, learning, strategic workflows) in the comments as they go live.
English
0
0
0
7
Shrinath V
Shrinath V@shrinathv·
I built the Virtual Product Factory as a governed workflow. Every major stage produces concrete, named files as the single source of truth. One agent generates. Another audits risks early. Others review, guard, and test — all working against the same shared artifacts. This recently helped me upgrade a Django backend, add a clean API layer, and rearchitect the frontend in React + Next.js for Founding Fuel — with support from just one backend engineer.
English
0
0
0
17
Shrinath V
Shrinath V@shrinathv·
This pulled me back to hard-won software engineering principles. Stable interfaces. Explicit contracts. Clear handoffs. Multiple layers of review. These exist for a reason.
English
1
0
0
8
Shrinath V
Shrinath V@shrinathv·
I want to use AI, but I don't want it to look like I used AI. At the same time, I want to display fluency with it where it counts. I fear this path leads straight to redundancy or premature obsolescence. A senior product manager put it plainly: she drafts PRDs with ChatGPT then rewrites them heavily before sharing. "If they think I use AI to do my job, what happens in the next round of cuts?" The tool accelerating her work has become the very evidence that could make her replaceable. This is the bind most professionals now live inside. We've abandoned the old way of working. The new one remains undefined. On one side: disdain for clumsy use. Drop a memo with that even rhythm and padded transitions and peers instantly tag you as someone who outsourced their thinking. The flood of slop has sharpened everyone's detector. On the other: companies pushing hard. Usage metrics, internal leaderboards, performance conversations that quietly ask - are you actually leveraging the tools? Fall behind and you look obsolete. The deeper fear cuts underneath both: if I get too good at this, am I writing my own replacement script? When everyone has the same models, how do I still stand out? We use the models aggressively in private - structuring, drafting, researching - then spend extra cycles erasing the traces so the final output carries our fingerprint. The same capability brings pride in some contexts and suspicion in others. Until the dust settles, our challenge is clear: how much workflow to own publicly, how to inject personal judgment and texture that no model can replicate, and how to signal mastery without triggering the replacement anxiety.
English
0
0
1
42
Shrinath V
Shrinath V@shrinathv·
I reduced my WhatsApp agent's token usage by 90% by teaching it to forget. We design agents to remember. We rarely design them to forget. Most people get this wrong. The default instinct is to fix memory: add vector stores, persist threads, build long-term recall. Make the agent remember where it left off. That solves one failure mode and creates another. I saw it while wiring my own WhatsApp agent. The longer the chat ran, the worse the agent performed. Old threads leaked into new ones. Topic switches disappeared into noise. The context window filled with stale instructions, half-finished thoughts, and irrelevant history. Token costs climbed. Clarity collapsed. Real assistant work is not one long task. It is task-hopping. I might dictate notes for an article, then ask for meeting-minute summaries before a call, then quick research on the person and company, then record an expense five minutes later. The flow is free-form, like speaking to a human assistant across the day. Agents are not human. When they carry every prior exchange forward, the workspace pollutes. An article idea colors the meeting prep. Meeting research lingers next to expense capture. Old instructions stay active long after they should expire. That is when "memory" becomes drag. So I built the opposite discipline. Every ten minutes the agent wrote a crisp state summary -goals, decisions, active threads - then wiped its working memory clean. When I returned, I fed it only that summary. The conversation restarted sharp and focused. Memory and forgetting are not opposites. They are the two blades of the same control system. Keep structured, high-signal memory for what compounds: core goals, resolved decisions, key facts. Enforce scheduled, ruthless forgetting for everything else. At natural breakpoints—end of task, time interval, topic shift—compress, archive, and purge. Working context stays narrow and fresh. This is not a frugal hack. It is better architecture. Long, muddy histories reward recency bias and punish precision. Deliberate forgetting forces compression and priority. It mirrors how sharp thinking works: retain signal, discard noise, refresh often. The agents that scale remember exactly what they should. And forget the rest on purpose.
English
0
0
2
38
Shrinath V
Shrinath V@shrinathv·
Drove back from Chikmagalur yesterday and saw dozens of autos plying on the Bengaluru-Hassan highway. Where are they all headed? Do they regularly run intercity on this route? Seems unique to this corridor.
English
0
0
0
50
Shrinath V
Shrinath V@shrinathv·
Call me cheap, but I have no interest in spending outrageous amounts on tokens & APIs when building AI systems. A friend asked me recently: "You must be paying a lot to maintain your system, right?" Wrong. I pay far less than what people estimate. The prevailing thought seems to be that serious builders must burn serious money. I differ. We should be building frugally while still using frontier-grade intelligence. The fear I always have is lock-in and uncontrolled scaling. If I poured all my data, workflows, and context into a single provider - Claude, ChatGPT, or Gemini - I would not be able to shift. Once prices rise (which they will), the system that makes me more effective will be an expensive dependency. I have seen this pattern earlier - what starts as convenient tooling quietly turns into a strategic vulnerability. When I set out to build my own systems, the requirement was simple: they must save me money and time over the long run. I made a calculated bet on using the frontier models. At the moment, they offer fantastic value for what we get. Many people are surprised I have not gone beyond the standard $20 plan for two providers (Claude and ChatGPT). I have a Jio Sim that gives me (for one year) Google Pro for free. My entire setup still stays under Rs. 5,000 a month - far cheaper than hiring even a junior developer. This frugality is not an afterthought. It is a core design principle to be both effective and disciplined. One simple hack I use and recommend: once the problem is clear, ask Claude or Cursor to write proper scripts instead of doing everything inside LLM context. This forces a clean split - deterministic work stays in code, LLM calls happen only where judgment is truly needed. I maintain a markdown index for every project that the system can scan quickly. I also look for trusted open-source libraries I can self-host. Faster Whisper, for instance, run locally. Since most of my files are stored in markdown, I don’t require huge storage data. Rather than buy a Mac mini, I repurposed an older Windows laptop to work as a server. My design system prevents drift and ensures brand consonance with a pre-set pipeline. I estimate that my total cost of building and running my personal intelligence system would be below Rs. 15000/-. The frugal mindset around token spend forces better architecture overall: tighter context, clearer prompts, and more intentional workflows. I have designed the two production-grade systems frugally, first building a virtual software team with clear handoffs - PM to architect to dev team. Each agent talks to the next through a fixed set of documents. This made early corrections easy, prevented drift, and added a self-learning loop so the systems improve over time. Once we move beyond the tokenmaxxing stage, we’ll all find that the real constraint in AI is not how much you can afford to spend. It is how intentionally you choose where to spend - and where not to.
English
0
0
0
37
Shrinath V
Shrinath V@shrinathv·
We often talk about how fast AI is replacing expertise. The story we hear is this: domain knowledge is becoming commoditized. What we rarely discuss is how some industries stubbornly resist that narrative. A friend called me today. He was excited about a startup idea in shipping. There was a genuine market gap and he wanted guidance on pricing. I had to ask the obvious question: who on his team actually comes from the industry? Would they be able to add their name to the startup? Without that, chances of success are low. I have seen this pattern repeat across renewable energy, oil & gas, mobility, and logistics in India. These are tightly knit worlds. Beyond the Silicon Valley playbook, an insider's stamp is essential. You can build something. Selling it, winning trust, and navigating real operational constraints is another matter. Without a qualified domain expert, companies engage you in endless POCs, sales cycles stretch, deals evaporate, and you never quite learn the real decision-making criteria. The buyer also sees you as a risky bet. They feel they would have to teach you their industry from scratch. Often, nothing you say or do can break that conviction. And yet, the pressure on these experts is real. AI tools can already surface patterns, generate options, and simulate scenarios at speeds no single human can match. The old reliance on deep but slow intuition built over decades suddenly looks vulnerable. Many domain experts quietly wonder: will my hard-earned context be reduced to prompt material for someone less experienced? I'd argue that the tools reveal a different possibility. The same AI that appears to threaten expertise actually demands sharper context judgment than ever before. It can generate plausible answers at volume. It cannot yet read the operational risk in a supplier's casual comment during a site visit. It cannot sit across from an intransigent procurement head and say with authority: trust me, I know this domain. You will pay more if you ignore what I have to say. The real question is not whether domain experts remain relevant. They do, especially in complex B2B plays. The question is in how they must operate. The experts who thrive will stop treating AI as a junior assistant or a threat. They will use it as a high-bandwidth collaborator. They will learn to set better context, challenge outputs ruthlessly, and govern the loop between insight and execution. They will move from being the sole source of answers to becoming the best asker of questions and the final arbiter of what actually works in reality. And be the voice and face of the technology that inspires confidence. Domain experts will not become obsolete in the age of AI. They will become more powerful - but only if they are willing to change how they work.
English
0
0
0
26
Shrinath V
Shrinath V@shrinathv·
We often talk of the hard skills around AI. Prompting. Agentic workflows. Fine-tuning. We treat it like a checklist of techniques to master. We rarely discuss the emotional layer underneath. I saw how much that layer mattered when I started looking at AI seriously last year. I had walked away from coding twenty years earlier. The decision was deliberate. My leverage came from reading business models, shaping product strategy, and working with people under pressure. Those skills shaped how I analyze problems. Returning to code looked like regression. I had watched strong engineers build clean solutions that still missed real context. The fear was sharp. Would learning AI pull me back into that pattern? Would it cloud the instincts I had spent two decades refining? Would I lose the distance that let me see the bigger picture? Yet the tools kept revealing capabilities that felt almost unfair. In client work I repeatedly hit the same wall: I could diagnose the gap between insight and execution, but I had to hand off to technical teams. The results often took weeks or months and landed differently from what I expected. These systems promised a different speed. What if I could move from idea to prototype myself? What if I could govern outcomes more precisely instead of handing off and hoping? The fear and possibility were both real. One day I stopped waiting for the perfect frame. I chose to play. No goal of "becoming technical." Just small, low-stakes experiments for their own sake. GitHub and the command line first. Then cloning repos, running local models, wiring simple scripts to workflows I cared about. One focused week at a time. Each small win chipped away at the old story. Fear dissolved. Curiosity moved in. I did not have to master AI. I learned alongside it without needing all the answers first. The work did not dilute my strategic instincts. It sharpened them. My background in product strategy and systems thinking let me set better context, challenge outputs, and build systems that solved actual problems. The boundaries between product work and technical work started blurring. Over the last few months I have built two production-grade systems with minimal technical assistance and set up a functional assistant for myself. Working with AI depends on how much you're willing to expand your curiosity and how much you're willing to tackle your fear.
English
0
0
0
30
Shrinath V retweetledi
Santosh Desai
Santosh Desai@desaisantosh·
Why did it take so long to invent the walker for the elderly? To put wheels on suitcases? To invent the menstrual pad for a problem that affected half the population in the world? Why necessity is not always the mother of invention. Today in the TOI
Santosh Desai tweet media
English
5
30
127
13.4K
Shrinath V retweetledi
Cliff Pickover
Cliff Pickover@pickover·
FREE Math Book. "Calculus Made Easy," originally published in 1910 by Thompson, is a beloved classic that demystifies calculus with the playful motto: "What one fool can do, another can." He wrote the book to make the subject accessible and fun for beginners, using plain English, everyday analogies, and a light touch: famously declaring that the mysterious "d" in differentials is just "a little bit of x." It has inspired generations (including Richard Feynman and Martin Gardner) and remains in print over a century later precisely because it proves calculus doesn't have to be intimidating. The book focuses on intuition and key concepts rather than intricate formulas, using a common sense approach with simple language and examples. It explains fundamental ideas like differentiation and integration for all to understand. Link: gutenberg.org/ebooks/33283
Cliff Pickover tweet media
English
11
75
415
42.1K
Shrinath V
Shrinath V@shrinathv·
Trust is the new usability. As AI takes more actions on your behalf, the core UX problem shifts. It's no longer "can the user figure out how to do this?" — it's "does the user trust what just happened?" Explainability and control are the new affordances.
English
0
0
0
15
Shrinath V
Shrinath V@shrinathv·
I love em-dashes. They have been around long before they were vilified as LLM writing - even featuring on the first page of one of my favorite books. Let's embrace them with joy in our writing.
Shrinath V tweet media
English
0
0
1
17