1-shot prompt on doppelganger:
take a look at this: https://t.co/Y9RB9kM0q3
I wanna create something similar but as a dog version. use the analyze media tool then create a prompt to use the generate media tool to create such a video just with a dog, hello dog
Today, we raised a $13M series A to solve the GPU shortage
Trillions are being spent on GPU CapEx while 80% of it sits idle
The capacity already exists
Virtualizing GPUs unlocks it
Backed by Matrix, YC, and CEAS
pump and dump 2026 edition:
1. invest into big ai labs
2. invest into applied ai startups
3. scream about token maxxing
4. startups are using investments for token maxxing
5. big ai lab makes $$$
6. big ai lab ipo
7. dump everything
8. congrats you rich rich now
1) you can sync your cookies through tailscale - (check out agentcookie by matt van horn)
2) its not enough. you don't want to give you agent access to your chase bank cookie.
the solution is a very secure permissions layer for browsers - sort of like @executorsh, but for browser-use. But its a catch-22 - its not possible
a) websites don't have a defined set of tool calls - your agent has to click around and figure out what it can and can't do.
b) you can't deterministically ban browser-use actions without knowing what the actions are.
legitimately the only way to solve this is to have some agent index a bunch of well-known sites in a sandbox and make MCP tools for them.
thoughts about the future of ui/ux of apps:
1. most things can be easiest described via words.
2. most things can be easiest understood via graphics.
this means, any interactive interface requiring clicking buttons can & will be replaced with a chat bar, but any interface which is serving information will still benefit from delightful ui/ux.