gpt 5.6 sol is goated but lowkey find myself using gpt 5.6 terra more
current flow for me is
if it is an ambiguous task that I need to do more exploration for, use sol
this is probably 30% of tasks
other 70% of tasks is just pure code changes I need to make with faster iterations and in that scenario I like grabbing 5.6 terra since it's a nice coding workhorse
crazy how much the AI coding zeitgeist has shifted away from token maxxing to being more token efficient in the past few months
I used to just spam the frontier model on everything
but with gpt 5.6 sol/fable it actually is not possible without burning up all your usage limits even on the $200/month plans (I'd never run out of usage when using opus 4.8 and gpt 5.5 even on extra high)
I'm finding myself being a lot more cognizant of using the actual right model for the task
> be me > spend 5 years building 14 different apps > make $0 for the first 3 years of building > fml i’m a shitty entrepreneur > next 2 years make a couple apps that make $1k-$2k/month > still feel like shit bc I spend too much time doom scrolling on social media where it seems like everyone is making millions a year every 5 seconds by telling AI “make me a millionaire, no mistakes” > convince myself scrolling on social media is useful because it provides me “alpha” by keeping up with latest technology/framework releases > end up getting shiny object syndrome of different product ideas and switching from an idea too quickly > fml I’m back to square one as a failure > decide I'm gonna reduce social media usage by 90% and stop scrolling on it and lock tf in to build a product and stick to it for a long time and not get distracted by other ideas > spend 2 months building with no users and no customers > go viral on social media in December and scale from 0 to 5k a month in revenue in three months > make my goals about what I can control (ship every day and market every day) rather than setting goals I can't control (e.g. get to 100k ARR in 6 months) > wtf I crossed 100k ARR in 6 months by just focusing on my inputs > continue generational lock in and stop getting distracted by doom scrolling and continuing to ship and market everyday
99% who said Fable 5 is a game changing, step function increase probably told it to make a simple CRUD app that could probably have been built with Haiku
Why is everyone larping building a l product that requires bleeding edge intelligence?
Your Average Tech Bro
had some crazy post nut clarity
I went too hard on shipping more and more features
but the UX was so trash and so tasteless
pushing on feature development was important in the beginning to convince users to take a chance on us
time to slow down a bit and clean it up
1 day ago | [YT] | 34
View 3 replies
Your Average Tech Bro
Use social media enough to get the alpha
But too much is poisonous and distracting
Making you feel like you’re behind/a failure when 99.99% you’re not
Everything is gonna be alright, just keep building and marketing
3 days ago | [YT] | 52
View 4 replies
Your Average Tech Bro
hot take: building/marketing a startup will never work on auto pilot
The only thing that is guaranteed to work is:
"Make it good"
Make a good product.
Make good content to market your product.
Doesn't matter if it's made by a human or AI, just MAKE IT GOOD.
I've spent too long trying to find cheap growth hacks/automations to win and it just is not worth it.
I am now just going to stare at my product/marketing content for hours and think to myself, "how can I make this better?"
5 days ago | [YT] | 55
View 8 replies
Your Average Tech Bro
gpt 5.6 sol is goated but lowkey find myself using gpt 5.6 terra more
current flow for me is
if it is an ambiguous task that I need to do more exploration for, use sol
this is probably 30% of tasks
other 70% of tasks is just pure code changes I need to make with faster iterations and in that scenario I like grabbing 5.6 terra since it's a nice coding workhorse
1 week ago | [YT] | 34
View 10 replies
Your Average Tech Bro
crazy how much the AI coding zeitgeist has shifted away from token maxxing to being more token efficient in the past few months
I used to just spam the frontier model on everything
but with gpt 5.6 sol/fable it actually is not possible without burning up all your usage limits even on the $200/month plans (I'd never run out of usage when using opus 4.8 and gpt 5.5 even on extra high)
I'm finding myself being a lot more cognizant of using the actual right model for the task
2 weeks ago | [YT] | 27
View 3 replies
Your Average Tech Bro
> be me
> spend 5 years building 14 different apps
> make $0 for the first 3 years of building
> fml i’m a shitty entrepreneur
> next 2 years make a couple apps that make $1k-$2k/month
> still feel like shit bc I spend too much time doom scrolling on social media where it seems like everyone is making millions a year every 5 seconds by telling AI “make me a millionaire, no mistakes”
> convince myself scrolling on social media is useful because it provides me “alpha” by keeping up with latest technology/framework releases
> end up getting shiny object syndrome of different product ideas and switching from an idea too quickly
> fml I’m back to square one as a failure
> decide I'm gonna reduce social media usage by 90% and stop scrolling on it and lock tf in to build a product and stick to it for a long time and not get distracted by other ideas
> spend 2 months building with no users and no customers
> go viral on social media in December and scale from 0 to 5k a month in revenue in three months
> make my goals about what I can control (ship every day and market every day) rather than setting goals I can't control (e.g. get to 100k ARR in 6 months)
> wtf I crossed 100k ARR in 6 months by just focusing on my inputs
> continue generational lock in and stop getting distracted by doom scrolling and continuing to ship and market everyday
2 weeks ago | [YT] | 374
View 44 replies
Your Average Tech Bro
99% who said Fable 5 is a game changing, step function increase probably told it to make a simple CRUD app that could probably have been built with Haiku
Why is everyone larping building a l product that requires bleeding edge intelligence?
Be fr lol
1 month ago | [YT] | 49
View 8 replies
Your Average Tech Bro
I used to get legit bricked up when a new model would be released
Now I feel like the models are more than powerful enough to accomplish any coding task I have
Maybe skill issue bc my work isn’t “complex” enough
Shit don’t hit the same as it used to
1 month ago | [YT] | 45
View 9 replies
Your Average Tech Bro
some initial thoughts after playing around with latest Antigravity 2.0 and CLI
the software and CLI feel behind in terms of features as codex/claude code
but I like the gemini flash 3.5
fast, good at design (better than any other model imo), relatively cheap
better problem to have than the other way around
2 months ago | [YT] | 29
View 5 replies
Your Average Tech Bro
it's honestly comical how bad gpt 5.4 is at UI
great at a lot of other things though
5.5 can't drop soon enough
3 months ago | [YT] | 26
View 2 replies
Load more