Vivek Suriyamoorthy

I believe we're on the brink of a design paradigm shift from UI/UX to DI/DX - Delegation Interface and Delegation Experience. The future is in how we delegate tasks to AI. #DesignThinking #TechTrends #AI #Innovation #FutureOfDesign #AIAgent #LLMAgent #Copilot

Agentic/Vibe coding is great until every tool you touched yesterday wakes up and says: “New version available.” And I don’t even remember installing you. #agents #vibecoding #agenticcoding

>But I do love the concept and I think that just like LLM agents were a new layer on top of LLMs, Claws are now a new layer on top of LLM agents, taking the orchestration, scheduling, context, tool calls and a kind of persistence to a next level. How is Claw different from Agent Harness? 🤔

Why are we doing this to ourselves 😂😳 The comments in these posts are insane, they are forming their own cult 🫠 #moltbook #OpenClaw

One prompt - "Imagine the view from Earth if it were the first planet in the Solar System." Gemini vs ChatGPT vs Grok #Gemini #OpenAI #Grok

This weekend, I revisited some old photographs I had created years ago. Images shot between 2009–2015 — starting from a Canon point-and-shoot, moving to a 650D, then primes, the 70–200 f/2.8 L, and occasionally renting a 5D Mark III. I used to enjoy shooting in RAW and do a long and slow workflow in Lightroom. There were many moments where I felt disappointed in missing the right frame, missing the focus but there were those few nuggets where it gave immense joy to see the perfect frame Going through those images brought back a rush of nostalgia — not just for the photos, but for who I was when I made them. The effort, the intent, the patience. The act of showing up mattered. On Saturday, I felt an urge I hadn’t felt in years — to pick up a camera again and just go shoot. Then on Sunday, reality hit. Today, we can generate images with precise prompts: camera body, lens, focal length, ISO, aperture, lighting, even “cinematic mood.” From a desk. Without being there. As someone deeply involved in building AI systems and excited about how LLMs and agents are transforming software, I am super excited about this shift. I work on it every day. But as a photographer, it hit me differently. Not because AI is “better” — but because it quietly removes something that once mattered deeply: presence. Photography, for me, was never just about the output. It was about waiting for light, missing moments, learning through friction, and knowing that the image existed because I was there. AI didn’t kill photography. It killed photography as proof of presence. And that realization was unexpectedly heavy. I don’t have a neat conclusion or a solution here. Both feelings coexist — excitement for the future I’m helping build, and grief for a craft where effort, patience, and being present once defined the value. Maybe the meaning of photography now isn’t in producing images anymore — but in choosing to stay present in a world where you don’t have to. Just sharing a real feeling. Curious if others have felt something similar. #photograghy #AI

Getting AI agents working for enterprises is not easy. And this is exactly why there’s so much opportunity right now. Today, the tech is *just* hard enough to get working right which means only a relatively small number of teams and companies in total will make this simple enough for the world to adopt. So you basically have a cheat code if you’re building AI agents because we know exactly how this will play out. The winners of the internet brought powerful web services to the masses. The winners of SaaS did the same for infrastructure and software. The same will be true for AI agents and knowledge work. Architecture shifts at this level only happen every decade or two. And this will likely be the biggest one we’ve ever seen in tech.

AI agents will fundamentally reshape the relationship between software vendors and customers over time. For the history of enterprise software, vendors sold customers technology that the customer was ultimately responsible for getting the value from. Software was just a tool, and it was largely up to the customer to figure out how to use that tool and get their employees to use it in productive ways. With AI agents, software vendors are now providing the work to the customer, not just the tools. If you’re a life sciences company and you have Agents that are reviewing your drug trial data, or if you’re a law firm using agents to review and process contests, the vendor is now an extension of your workforce. The implications to this are massive, because now the software is essentially a supplier of work output, much more akin to a professional services provider in the past. Here are just a few things that change as a result: * Domain orientation: the winning AI agent providers will likely be the vendors that have the best understanding of the domain or problem they’re going after. No enterprise wants to hire the 3rd best person they can afford, and it will be the same with agents. The best agents in their respective domain will provide disproportionate value to the customer, so having a deep domain focus will be a huge leg up. * Evals: Evals will become a much bigger deal in the enterprise. How do you deploy agents if you don’t know how well the agent is performing at a task? Evals are essentially the new performance review system for autonomous work. And it will be even more important for agents than people given the blast radius of a good vs. bad agentic workflow. * Implementation: The implementation of agents also looks very different from traditional software because you are actually changing the underlying workflow for the customer. The need to get data, tools, context engineering, and the employee change management is going to be bespoke per customer. This is why forward deployed engineers are growing in popularity, but we’re still very early in realizing how big of a change this will be for software vendors. Support: when you’re actually now on the hook for a company’s ultimate work output, the support dynamics are going to change meaningfully. A customer is no longer calling you because they can’t login or there’s a bug in your software that they can try and work around, but now because an agent didn’t deliver the desired output needed. This is going to require a very different level of support for customers that are far more tailored to he customer’s business process. * Versioning: when you’re supplying work to an enterprise, you don’t have the same flexibility of just being able to change the work output at any moment. It would be like if all of a company’s employees came in one day and uniformly operated differently. This is going to have huge long term implications to how agent updates work in an enterprise and how prior versions get maintained. * System integration: As an aside, there’s going to be room for a new type of system integrator to emerge (or for existing ones to pivot) for implementing and helping manage AI agents. Because of the complexity of deploying and operating agents, and the change management necessary, most companies will need support along the way - there will likely eventually be points of aggregation from the SIs. These are just a few of the changes that will start to emerge when software providers sell and deliver ultimate work output to customers. But it’s going to be a very big change over time.

Where do you think enterprises will land in 3 years? 👇 🔹 Prebuilt (fixed logic) 🔹 Configurable prebuilt (prompt tuning) 🔹 Bespoke (enterprise-built) 🔹 Hybrid (evolving from prebuilt to bespoke) #AIagents #EnterpriseAI #FutureofWork

Configurable prebuilt agents will dominate short term — fast to deploy, flexible enough for most. But long term, enterprises will want full control: -Deep context integration -Governance & telemetry -Cross-agent collaboration That’s where agent platforms win. ⚙️ #AIplatforms #EnterpriseSoftware

SaaS made people adapt to software. Agents will make software adapt to people. 🤖 Prebuilt agents are today’s SaaS apps — the next decade belongs to agent platforms, where every company builds its own bespoke agents. 🧠 🧵👇 #AI #Agents #SaaS

Just watched this podcast of @karpathy in @dwarkesh_sp's show Mind-blowing stuff, particularly the last section on education was brilliant. It felt like Sheldon Cooper talking in real life :) Watch here : youtu.be/lXUZvyajciY

I think the way I learn has completely changed after LLMs. I was never a voracious reader — I’ve always been the kind of person who learns by asking questions. There were only a few patient people in my life who would answer my endless curiosity… most eventually got tired 😅 But ChatGPT (and LLMs in general) never got tired. And because of that, the speed and depth at which I acquire knowledge has become exponentially better. We used to be in the era of content sharing — writers shared knowledge in structured formats, and readers consumed it passively. But in the LLM era, I believe sharing the questions that led you to understand something is far more powerful than sharing what you learned. Each learner can follow those questions, branch off in their own direction, demand examples, analogies, or even visuals — and get back on the course. I’m calling this concept Prompt Courses — learning paths built entirely through questions. To make this real, I’ve started an open GitHub repo called StayCurious 🧠 It’s a community-driven library of question paths extracted from real AI chats. 👉 Repo: github.com/vivek-suriyamo… If you’ve had a chat session where you feel you have learned a lot — use the prompt in the README to safely extract your learning questions (no personal or confidential info), and contribute your “Prompt Course”. If you’re technical, raise a Pull Request (instructions in the README). If you’re non-technical, just share your questions in the comments — I’ll help add them to the repo until we make this process easier. Let’s move from content sharing to curiosity sharing. Because sometimes, the right questions are the real teachers. 💡 I have added a course to start things off - Foundation of metabolism and how we can hack hormones to lose fat - github.com/vivek-suriyamo… Thanks @yuvanist for collaborating. #prompt #LLMs #learnbyasking #promptcourses #promptpath

🔁 The Irony of Engineering in the New Age of Software We’re at the peak of an irony. In the SaaS era, everything felt stable. The stack was known, frameworks were predictable, and the fundamentals were well understood. There was abundant supply of engineering talent — and equally high demand. It was an age of optimization — building faster, scaling better, and refining what already worked. But the AI era has flipped that equation. The technologies are new, volatile, and evolving almost weekly — especially in areas like context engineering, where the real breakthroughs are happening. The number of people who deeply understand these systems is incredibly small. Yet, I see more and more non-technical founders starting AI startups, assuming they can build products by vibe coding. Sure, they can. But what they’ll end up building is a SaaS-era product, not an AI-era one. The real opportunity in this era isn’t building the same software 10x faster or cheaper. It’s building the next generation of software — the kind that’s only possible because of AI and agents. And this, to me, is nature’s law at play. When the skillset to build old software becomes a commodity, the standard for what counts as “software” itself evolves. Engineering doesn’t retire — it reinvents itself 💪 #AI #Engineering #Startups #SaaS #AIAgents #LLMs #ContextEngineering #Tech #AIEra #Builders #ProductThinking #Innovation #GenAI #AIDev #AgenticAI #SystemOfIntelligence

The future of CPQ isn’t a form — it’s a conversation. AI Agents at @everstageinc now talk to each other to build quotes and align incentives. Context-aware. Outcome-driven. Behind the scenes, there are two intelligent agents at work: 🧠 The CPQ Agent, optimizing quotes within pricing guardrails. 💡 The Incentive Agent, aligning those quotes with commission outcomes. Together, they talk, negotiate, and find balance — creating a seamless, intelligent loop that drives both revenue efficiency and rep motivation. Welcome to revenue transformation. 💡 Read more here - businesswire.com/news/home/2025… Thanks to the foundational model providers - @OpenAI @AnthropicAI #cpq #everstage #agents #AI

AgentKit - I'm not entirely sure about using directional workflows with agent nodes — aren't agents meant to operate autonomously? #OpenAIDevDay

Apps in ChatGPT - talking to a live app is a really cool idea 😎 #OpenAIDevDay