🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Wrote two angel checks this month, both used Sydecar for RUV. Do people prefer to angel list?

That's a lot of thinking to not realize it isn't X's turn. (by the way it still gets it wrong if you tell it whose turn it is)

Two potential futures for no code: 1. Prompt a low code tool: AI composes existing components (AI retool) 2. Prompt-to-code: AI writes code from scratch I used to be squarely in the former camp but I’m increasingly wondering if the latter is the more likely dominant paradigm

So far so good

Subtle but important thing I think a lot of people building w/ AI miss— The goal is to make a *product* that can help accomplish a task, NOT make the model itself perform well on the task. Better models make this easier but there are a lot of product levers to pull as well.

Vibe coding is the new 10x engineer

Navigating the agent landscape as an enterprise is tricky as it grows more complex. Enterprise agents need more than reasoning—they need precision, feedback loops, and to live inside well-designed products

I joined SPC six years ago just as they were deploying fund I. In the years since it’s been amazing to see the community and fund grow, change and adapt on its own -1 to 0 journey. Have met so many amazing founders and friends through SPC. Congrats!

The most valuable data for your AI agents isn’t in your enterprise systems—it’s in your experts’ heads. Are you capturing it? In the latest episode of Human in the Loop, we’re talking all about what data enterprises need to build truly effective agents

I don’t think people are properly planning for a world where the cost of writing (at least simple) code goes to zero. The system of record and system of work have always been important but now more than ever. Context is the product.

What company worth under $100b today is most likely to become a $1t company?

❤️🫂

One of the things that made Cursor so successful is that best in class tool for power users was already open source. Browsers have a similar dynamic which is why we’re seeing the competition heat up there. PowerPoint, Excel, Email (maybe) all much more difficult to compete.

This is the type of work I’m talking about for enterprise RL. When you use rubrics as rewards you can expand the set of RL domains dramatically

Who is building Pave but for value based compensation?