Unpopular opinion:
I’m tired of hearing “unpopular opinion” when the opinion is not at all unpopular.
2
18
1,327
I'm tired of a culture where the creation of one thing all too often means the destruction of another.
1
217
Updated hiring criteria for my startup:
Technical skills: Preferred
Experience: Negotiable
Opinions on Cursor vs. Windsurf: Strong
Number of unread Substack newsletters in inbox: 300+
Number of stickers on laptop: Minimum 7, bonus if one is from a failed unicorn
DM to apply.
1
3
311
Me: "So how’d you get into VC?"
VC: “Funny story, actually…” ▶︎ •|။||။|။|||||||။|။||။|။|။| 1:37:12 (sponsored by Athletic Greens)
2
6
357
It seems likely that whoever creates the AI ecosystem will be a billionaire.
Consumers and enterprises alike want to give directives - sue my competitor for espionage, do my taxes, book a doctor appointment, move this shipment from location X to location Y, negotiate this domain purchase, whatever - and have it executed in line with the party's unique preferences: context, style, communication methods, billing protocols, security considerations, approved vendors, etc.
Many of these tasks are already possible with frontier models, yet AI is not currently being widely leveraged to execute them. Why?
1. There are many companies building atop these models, but distribution is still a problem. Selling to enterprises is hard because there's a lot of red tape to cut through, and selling to consumers is even harder because you need to educate them & the math often doesn't work out without subsidization.
2. Many tasks require a multi-layered approach to execution. Some things need to happen in the physical world, some need to use a computer, some require multiple agents, some require a human, some need to connect to complex legacy systems, some require a firm with licensure, etc.
An AI ecosystem would allow enterprises and/or consumers to define their task along with a budget & time allotment, and it's done - like magic. The system understands the preferences of their organization and lays out a plan to make the action happen. This could happen through a UI for one off tasks or an API call for recurring tasks.
Technologically, this doesn't seem all that hard; reasoning models are already capable of this type of complex planning, and LLMs are already capable of doing a surprising amount of knowledge work with a high degree of accuracy. The hard part is in partnerships and distribution:
- Working with partners to make things happen in the real world, hiring humans, finding firms with licensure to handle complex tasks, picking the best foundation models, etc.
- Selling into enterprises and guaranteeing you will protect their data, only work with parties who meet their standards, etc. OR branding as a personal assistant and selling to individuals: getting users, onboarding them, and retaining them and increasing usage.
It's possible this won't be a single, unified ecosystem but instead thousands of SAAS companies. Maybe enterprises will still want to cherry pick vendors and build workflows themselves. This is a harder argument to make for individuals, but so is attracting consumer usage. Realistically, at a large scale this product would require so many integrations it could create thousands of profitable companies who plug into the ecosystem and don't even need to think about distribution.
Is anyone already building this? It seems like the perfect opportunity for an ambitious, smart team.










