We were thinking for like a month what to do, and i thought that it might be a decent idea to just build a CC clone that works with any model and fully opensource instead of plugging models to CC via various hacks. WIP obviously, but may be someone whant to test and share opinions.
https://github.com/slopus/flow
https://github.com/slopus/flow
🔥5
Funny how startups try to launch and they got shoved in the face with our app
LLMs are so "smart" that even a single reorderable list is out of distribution apparently. Most of my work is still manual.
😁4
My mattress literally failed to wake me up because of us-east-1 was down
😁22
I am this close 🤏 to go to the cave for 5 years and reinvent literally everything in software:
1) AI friendly backend with DSL like language that is much more high level than HTTP request, with a cool protocol. (Even Kotlin doenst even have a nice ORM!! Typescript is still single thread === multi server should be supported from the start)
2) New React Native - a proper animation engine that uses renderthread/renderprocess, that have built in multithreading, all standard libraries already baked in, decoupled js part and allow having any language. Kotlin native but with real Native UI on iOS? All big 3 (flutter, rn, kotlin) sucks currently.
3) AI CLI agentic system to build all of this.
This is all is so straightforward.
1) AI friendly backend with DSL like language that is much more high level than HTTP request, with a cool protocol. (Even Kotlin doenst even have a nice ORM!! Typescript is still single thread === multi server should be supported from the start)
2) New React Native - a proper animation engine that uses renderthread/renderprocess, that have built in multithreading, all standard libraries already baked in, decoupled js part and allow having any language. Kotlin native but with real Native UI on iOS? All big 3 (flutter, rn, kotlin) sucks currently.
3) AI CLI agentic system to build all of this.
This is all is so straightforward.
2🔥10👎1
Today Bryan Johnson takes 5g of shrooms and streams it: https://x.com/bryan_johnson/status/1987594468544192942?s=20
X (formerly Twitter)
Bryan Johnson (@bryan_johnson) on X
Hey guys. It’s Kate Tolo, Bryan’s co-founder. I’m taking over his X account today.
He’s just taken the 5g of magic mushrooms.
I’ve officially confiscated his phone 🫡
I’ll report back once he starts feeling it
He’s just taken the 5g of magic mushrooms.
I’ve officially confiscated his phone 🫡
I’ll report back once he starts feeling it
🐳4💊3👍2
Built an MVP - Power Tool
I am still having trouble to formulate what it is to convince people, but once anyone tries it becomes clear what is it and how valuable it is.
This is a tool that helps to acquire social power even when you dont have natural skills. Did you ever wonder how Sam Altman is so good? Is that he has such tool + his natural skills of working with people. This tool, while being simple, dramatically improves understanding people around you.
I was using it to understand complex power dynamics at Amazon, understand motivations of people around me and find a way to move towards a goal that is needed for our team or me personally.
It is an AI assistant that is meant to be used mostly by talking to it, not texting. It supports english and russian (for now).
Please, help me to test this idea - download it and talk about some challengine situation that you faced and weren't being able to understand fully or triggered a severe frustration. It is encrypted and key is stored in iOS Keychain, and inference is running on 11labs agents platform and doesnt touch my backend (unencrypted). I wouldn't call it end-to-end encrypted, but it is almost there.
It uses best model available - Claude Sonnet 4.5, so it can be a little bit slow, but it is well worth it comparing to smaller models.
https://testflight.apple.com/join/cxNKEBdb
I am still having trouble to formulate what it is to convince people, but once anyone tries it becomes clear what is it and how valuable it is.
This is a tool that helps to acquire social power even when you dont have natural skills. Did you ever wonder how Sam Altman is so good? Is that he has such tool + his natural skills of working with people. This tool, while being simple, dramatically improves understanding people around you.
I was using it to understand complex power dynamics at Amazon, understand motivations of people around me and find a way to move towards a goal that is needed for our team or me personally.
It is an AI assistant that is meant to be used mostly by talking to it, not texting. It supports english and russian (for now).
Please, help me to test this idea - download it and talk about some challengine situation that you faced and weren't being able to understand fully or triggered a severe frustration. It is encrypted and key is stored in iOS Keychain, and inference is running on 11labs agents platform and doesnt touch my backend (unencrypted). I wouldn't call it end-to-end encrypted, but it is almost there.
It uses best model available - Claude Sonnet 4.5, so it can be a little bit slow, but it is well worth it comparing to smaller models.
https://testflight.apple.com/join/cxNKEBdb
4❤9🤡1
Anybody else notice that Gemini 3 does not seem as good as it used to be? Did they quantize the model?
😁7
Ok it seems power tool was really used by a single person 😅 and 10 more tried, not great considering how great this tool was for me
Please open Telegram to view this post
VIEW IN TELEGRAM
i am starting to feel that skill imbalances in the world are not because of some natural talents of one individual and lack thereof in another. it is more like a bad start in learning things. similar to when you try to learn something and then you learn it incorrectly. people spend time and time doing stuff incorrectly and then it accumulates and multiplies. it doesn’t sound like much - you can always re-learn - but in practice i doubt anyone would learn something from scratch. if you learned how to drive and you acquired wrong reflexes - you are not going to re-learn anything, you will just carry on with your life. imagine you didn’t build a behavior of resting your leg on the brakes? why would you bother to change something? how will you even learn that you are doing something wrong? chances that you will get there are next to none.
such problems then multiply - bad ideas, bad principles, bad knowledge. bad driving skills will lead to lacking understanding of a probabilistic nature of the world which will lead a person to aim for local maxima. or not learning to ride a bicycle leads to an overly cautious life which leads to anxiety and then to depression.
is it like a butterfly effect?
you start by not learning some skill and it permanently shuts off some knowledge branch, some part of life.
some people do this a lot, maybe even everyone does this to some extent
and this stuff simply divides people and the poor become poorer and the rich richer
such problems then multiply - bad ideas, bad principles, bad knowledge. bad driving skills will lead to lacking understanding of a probabilistic nature of the world which will lead a person to aim for local maxima. or not learning to ride a bicycle leads to an overly cautious life which leads to anxiety and then to depression.
is it like a butterfly effect?
you start by not learning some skill and it permanently shuts off some knowledge branch, some part of life.
some people do this a lot, maybe even everyone does this to some extent
and this stuff simply divides people and the poor become poorer and the rich richer
1❤9🤔6
the most universal thing that gives the most success is simply always betting on the best performers. picking the best team/people around you based on your own taste (not what some anonymous crowd thinks) consistently yields good outcomes.
looking back when i did that it always worked. eventually best performers became not so best and you have to switch to a new (usually much better) ones.
rinse and repeat.
any time i did choose differently - it failed
looking back when i did that it always worked. eventually best performers became not so best and you have to switch to a new (usually much better) ones.
rinse and repeat.
any time i did choose differently - it failed