We at @Flybridge couldn't be more excited to be on this journey with Brian, Sid, and the entire @blitzyai team. They've created a truly generational company and have developed technology that's unparalleled.
In this age of AI, so many things are simply promotable into existence, but what Sid and Brian have built is an N of one. Let's f**king go! π π€
Can AI Build a Personal Family Office for Everyone?
In this episode of AI Without Borders, I sat down with Bruno Koba, founder and CEO of Astor (YC S25), an AI investment advisor rethinking how personalized financial advice gets delivered. Astor announced today their $5M seed round led by MONASHEES, with participation from Y Combinator, Goodwater Capital, and Gilgamesh Ventures.
https://t.co/DRLzwZl5UW
We (@astorinvest) raised a $5M seed led by Monashees, with participation from Goodwater, Gilgamesh Ventures, and other incredible investors.
In less than two months after launch, thousands of users have connected over $200M in assets to our platform. We're just getting started.
NYC is *a* new home for so many people building incredible things in the AI ecosystem. Hosting an intimate dinner with @AnthropicAI soon. We want to find the best transplant founders building the future in the city. Tag the most exciting new arrivals below. cc:@shaig@brettperl@andruyeung@lindsaykap@TeddySolomon11
The biggest mistake people make with AI agents? Over-constraining them.
In my conversation with @rauchg, we talked about what he described as the bitter lesson for agents: teams over-constrain agents to have a feeling of control, boxing them in with hand-written rules and narrow permissions. But that just leaves capability on the table. The real progress comes from giving agents proper tools and enough context, then letting them operate inside sandboxes where they have full freedom within a secure boundary.
We also spoke about a world where building software has become dramatically easier. The value shifts away from the software itself and toward the judgment and taste embedded in the prompt. Vercel is already experimenting with this: rather than paying subscriptions for software, you purchase prompt templates, share the recipe with your agent, and spin up your own personalized version.
The tools you thought were right a few weeks ago might no longer be the right ones.
This was @rauchg reaction to @karpathy, and it perfectly captures how fast the AI space moves. If you are not evolving and testing new tools frequently, you are already behind.
I recorded one of the best episodes to date, subscribe to the AI Without Border's Podcast to get it when it drops on Thursday. In the episode we also discuss the Bitter Lesson for agents, how to focus in a world where everything is possible, and more.
I've never felt this much behind as a programmer. The profession is being dramatically refactored as the bits contributed by the programmer are increasingly sparse and between. I have a sense that I could be 10X more powerful if I just properly string together what has become available over the last ~year and a failure to claim the boost feels decidedly like skill issue. There's a new programmable layer of abstraction to master (in addition to the usual layers below) involving agents, subagents, their prompts, contexts, memory, modes, permissions, tools, plugins, skills, hooks, MCP, LSP, slash commands, workflows, IDE integrations, and a need to build an all-encompassing mental model for strengths and pitfalls of fundamentally stochastic, fallible, unintelligible and changing entities suddenly intermingled with what used to be good old fashioned engineering. Clearly some powerful alien tool was handed around except it comes with no manual and everyone has to figure out how to hold it and operate it, while the resulting magnitude 9 earthquake is rocking the profession. Roll up your sleeves to not fall behind.
@maxmarchione Big fan. As you consider the roadmap, some good additions as a user include:
- Ability to use a higher power model for complex queries.
- Expanded offering such as MRI and Dexa scans.
- More useful integration with wearables. I connected my Fitbit and have not seen any value.
What will differentiate the next generation of business software?
Context.
In this episode of AI Without Borders, @franco_pinto_ CTO of @mobilefirstco , explains how they discovered something powerful: you can understand a business remarkably well just by knowing which tools they use. That insight fuels their approach to contextual and proactive intelligence:
- Connect a CRM or scheduling tool and the system instantly understands workflows
- Surface the right information the moment a call starts - Suggest follow ups automatically when a conversation ends
-It is context doing the heavy lifting instead of the user.
Franco also shares why their entire product strategy resembles a consumer company. Self-serve onboarding, plug-and-play integrations, and interfaces that feel simple even when the workflows are complex.
YouTube: https://t.co/Gs2Wj6HGrs
Founders building consumer apps must be intensely aware that users will increasingly ask this build vs. buy question before committing to a subscription, especially as platforms like Replit (@amasad ) make app design easier. Unlike in the enterprise setting, this threat is higher because many consumer apps are simpler and require less maintenance or updating . Consumer should make apps that are so robust and undeniably excellent that even an AI powered DIY alternative feels like a waste of time (even as cost to build apps keep going down).
AI is reshaping consumer subscription purchase decisions.
The way consumers decide to purchase a subscription is changing. Before, the main question was: is this valuable enough? Now, with AI coding making it easier to build apps, an additional factor emerges: "Can I just build this myself with AI?"
This past week, I saw a new productivity app called Halo by @benspringwater. I had previously built a basic habit tracker, but it wasn't great. When I saw Halo, I tested it and purchased the $69 per year subscription. It was well designed with several modules. I could have replicated some features in my app, but it would have cost at least $100 to build and delivered only about 60-70 percent of Halo, so it was not worth it. However, for another productivity app with focus music and work timers, I decided to build it myself. The subscription was $100 to $120, and my DIY version cost only $30 to $40, and it was a simple app.
Why Most Startups Get the FDE Role Completely Wrong
Forward Deployed Engineers have become one of the hottest hires in the AI era, and also one of the most misunderstood. Founders imagine FDEs as silver bullets. But after hosting an internal learning session at Flybridge with Brian Keohane, a experienced Palantir FDE, one truth was clear:
For the vast majority of startups, hiring an FDE is the wrong strategy, and often a very expensive one.
Most companies bring on FDEs for the wrong reasons:
β’ to compensate for weak product market fit
β’ to patch a fragile or immature platform
β’ to support low ACV customers with limited upside
Every great startup story starts with one thing: a connection.
Startups are all about amazing people coming together around a unified vision. Sometimes, all it takes is making the right match.
Over the years, Iβve met so many incredible people: founders, engineers, researchers, and operators. One of my favorite parts of this job is helping them find each other.
So I decided to run a small experiment.
I built a simple site where:
π€ Founders can find co-Founders or early hires
π Engineers, Researchers, or Operators can find exciting early-stage AI startups to join
If thereβs a match, Iβll make a private, confidential intro. Nothing will be shared publicly.
You can fill out the form: https://t.co/umyYnJ5d4J