ex-CTO $2M ARR building an open source trading app https://nitter.gq/t.co/iIz5uDsl3b
Joined January 2025
- Tweets10.3K
- Following649
- Followers796
- Likes6.2K
A year ago I had 785 followers. Today I have 791. All I did was post and reply multiple times a day and the algorithm did the rest
During this time I grew my socials to 40k followers
And have had multiple videos hit 100k views
2 videos over 800k
All you need is consistency and to not put all your eggs in one basket. Your ideas and content might resonate with one platform over another
depending on the context, humans take minutes to years to load context...
in comparison, LLM's take second to minutes
first one to warn everyone about cognitive overload and the cost of reviewing ai generated code
instagram.com/reel/DONN7P6kw…
btw i was the first one to discover this:
youtube.com/shorts/HIHxDqGau…
you don't want 'adversarial' anything
since opus 4.7, you absolutely did not need adversarial reviews
adversarial review was formulated from people using non frontier models in non frontier harnesses, trying think of clever/hacky ways to improve their output quality
their concept was that if you get enough mediocre LLM's to over review something, you'll catch all the missing points - and this sort of worked for non frontier models
but if you use a frontier model you don't need any of this, regular code review is more than enough
what you need to do is have opus 5.5 write the original task/issue
then have sol validate it against the code base, checking for validity, accuracy, and functionality
if you use a plan, build the plan with opus or sol, and then validate the plan with the opposing
then build with opus/sol
then pr review loop until you get a lgtm
all of this is available in npx rk-skills
maybe a hot take but i now consider adversarial reviews mostly a waste of tokens. far better to use those tokens on empirical combined with formal verification. i have updated pstack to use the interrogate skill less, and eventually i think i may remove it altogether
i have found very little value with code review in the abstract, agents tend to just get lost in the nitpicks and yagnis
i wish more people would just build with rk-skills and not have to re-invent the wheel everytime they build something new
pitch me you startup and your ideal SF based investor profile
i'll try to put you in touch if you have something
my prediction for the next few years
ai will offload the intelligence cognitive abilities of 99% of humans - effects of this will be insane
imaging everyone being reliant on ai to do even the most basic of cognitive decisions
creativity will be the new 'intelligence'
People will want non verbal emotional entertainment and engagement
Humor and intelligence and verbal communication arts will be more valued