Welcome to Venture Logbook, your bi-weekly guide to building venture in a new era. Today, we have Trend Spotlight on the rise of AI agents and what OpenAI’s AgentKit means for automation, Startup Unboxed with a fresh take on mental health tools in the workplace, and Tech Dictionary to help you decode the difference between AI agents and agentic AI. Stay tuned to see how these shifts could reshape your investment strategy and the future of work.
Trend Spotlight
Every year after OpenAI’s DevDay, people say that OAI has killed off countless startups. But if you look closer, we always see strong teams like Cursor emerging and thriving.
From a product perspective, they’re aiming to build a personalized OS, a default gateway for consumers in the AI era.
With daily productivity features like ChatGPT Pulse, social experiences like Sora2, and an external ecosystem app SDK, it really brings to mind how Yahoo once owned the gateway to the internet for an entire generation.
Will they become the go-to portal for the next decade, or will a rising star like Google come along and steal the spotlight, just like in the old days? We’ll have to see how things play out.
But today, I want to talk about OAI’s latest feature: Agent Kit.
They’re not the first to develop workflow automation—nor are they the current market leader. Drag-and-drop visualization isn’t exactly a radical innovation either.
So why all the attention? The real question is: Will this “AI agents” model become the mainstream path for automation in the future?
From long-standing players like Zapier, to open-source options like n8n, and even Google’s originally planned Opal, the mission has always been to lower the barriers to automation, to make no-code workflows an everyday reality.
While many people don’t agree this is the “AI agents” they’ve imagined (and I get you, my geek friends), we can’t deny that this is the approach closest to true market fit.
The AI agents market is now in an awkward phase:
- Large tasks rely on context window size
- Complex tasks depend on foundation model capabilities
- Long-process flows hinge on memory
Honestly, those “big three” types of tasks usually land with only the most talented folks in real life. Agent Kit’s bite-sized, time-saving workflow automation is exactly what most people actually need. Smaller task chains might seem like low-value at first, but don’t underestimate the total value they create when scaled up.
It’s too soon to predict who’ll take the number one or two spots in the market. But one thing seems inevitable:
Office jobs built on repetitive tasks will be 99% replaced in the next 1–2 years.
So whether you’re an investor, an executive, a founder, or just an average person, here’s the one thing you need to start doing right now:
- Find the inefficient, high-value activities in your personal productivity matrix and automate them
- Double down on the high-efficiency, high-value stuff, and make sure others know what your moat is
- Hand off the efficient-but-low-value and inefficient-low-value work to others.
We are living through a historic moment.
No matter where you are on this planet, you’re sure to be touched by this wave—even if you’re standing all the way up on a mountain top.
Startup Unboxed #1
Just attended a university incubator demo day and one startup totally caught my eye with their gorgeous UI design. They built an enterprise dashboard with a "stress mode" toggle where employees can instantly switch to chat with an AI assistant when feeling overwhelmed. The interface is so beautifully designed it literally feels like having TIDE meditation app embedded right in your work dashboard.
And the questions I had were:
-
Would people actually use a "relax mode" while they're supposed to be working? There's something psychologically weird about switching to stress relief in the middle of your workday.
-
From a business model perspective, would employers pay for the product for their employees' stress? I saw they showed data that overwork index will lead to enterprise loss. However, I'm not sure if enterprises will notice that or care about those metrics. Yeah, maybe it's HR's metric, then it should be separate from the workspace, not combined.
If I were the team, I would separate the wellness assistant from work tools entirely. Sell it directly to HR departments as a standalone AI mental health solution. Way simpler, no workflow integration headaches, and you can keep that gorgeous UI without compromise.
Startup Unboxed is a series that I met startups, journal the takeaway and share my thoughts.
AI Agents vs. Agentic AI
I remember hearing from a VC friend that everyone in San Francisco is saying this is the “year of the agent.” Not too long ago, there was a lively debate in the ecosystem about the definition and difference between “agents” and “agentic AI.”
But now, with OpenAI’s release of Agent Kit, the line between the two concepts is clearer than ever.
- AI Agent: An AI that can autonomously execute commands—single-purpose, single-direction interaction.
- Agentic/Agentic AI: A term centered on whether the system truly possesses things like choice, planning, interactivity, long-term autonomy, and dynamic collaboration. It describes a whole range of capability from weak to strong.
You can think of Agentic AI as a mentor guiding an entire classroom of students, systematically planning, orchestrating, and assigning tasks based on experience.
By contrast, an AI Agent is more like a class monitor or student leader who takes on specific tasks assigned by the mentor. Imagine a teacher leading a class through a group project: the teacher organizes everything and delegates tasks, while “team captains” are in charge of guiding (or even completing) assignments for their teammates using set methods.
— Andrew NgAll new AI systems can be viewed as agentic, each possessing different levels of autonomy. Don't waste time arguing about rigid definitions of what counts as an agent. What matters is what these systems can actually accomplish in real-world scenarios.
One of my favorite takes on this comes from Andrew Ng. He pointed out that “agentic” is really about the degree of autonomy and capability, not a black-and-white “is it or isn’t it an agent” question.
In other words, whether an AI system is “agentic” sits on a spectrum, a degree of agency, rather than a simple yes or no.
🧠 Tech Dictionary helps you decode common tech terms so clearly, even your grandma would get it. Quickly find out what matters & why. So you're never lost in the tech talk.
👋 Thanks for reading! For more tailored insights on venture building in the AI era, subscribe at Venture Logbook. Disclaimer: This newsletter is for informational purposes only and does not constitute financial, investment, legal, or tax advice.