steve pfp

steve

@stevef

103 Following
260 Followers


steve pfp
steve
@stevef
Is there a markets/traders channel around these parts?
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
still believe that the ideal decentralized social network would look like IRC. IRC never reached mainstream for many reasons (no rich clients, UX, no mobile etc). centrally governed servers that vote to bridge w/ each other on an open protocol and fix where IRC failed above imo
3 replies
0 recast
0 reaction

steve pfp
steve
@stevef
one niche group that keeps me on Twitter is active market traders (not just crypto, but any and all markets) would love to see that community more active here (modulo the shilling)
0 reply
0 recast
2 reactions

steve pfp
steve
@stevef
Using Warpcast native atm -- what are some other hot clients being built on this lovely protocol these days?
2 replies
0 recast
4 reactions

steve pfp
steve
@stevef
so uh, does anyone even use stackoverflow anymore? i think the humans did their job well over the last several years and their watch is over
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
how can any reasonable parent encourage their kids to pursue a computer science degree in 2023? or any degree that leads to white-collar opportunities? what do we tell our kids to study besides learning to learn? going to be a challenging decade of parenting
5 replies
0 recast
1 reaction

steve pfp
steve
@stevef
by the time this BabyAGI script finishes running to give me the ideal destination to take my wife on her birthday, my OpenAI bill will cost more than the flight+destination itself will
1 reply
0 recast
0 reaction

steve pfp
steve
@stevef
i used to fear that being a tech generalist or a jack of all trades was short-term +EV but terrible for career sustainability i am growing more conviction that the day of the specialist is dead, and it is the jack of all trades who will prosper. this is assuming humans have a role in tech tomorrow
2 replies
0 recast
0 reaction

steve pfp
steve
@stevef
🔥 feature at an AGI level in a multi-modal LLM world -> send a video of an interactive website, or link, or a screenshot if it's just a static design -> LLM responds with semantically correct HTML/CSS/JavaScript. bonus points if you can target frameworks (Tailwind, TypeScript etc) "How did Apple build this UI?"
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
prompt golf - have gpt solve a non-trivial problem with the least amount of prompts
2 replies
0 recast
0 reaction

steve pfp
steve
@stevef
in the era of software 2.0, documentation will be read less frequently than ever before, yet its importance will be greater than ever
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
"gpt-4 has a context length of 8,192 tokens. We are also providing limited access to our 32,768–context (about 50 pages of text) version, gpt-4-32k, which will also be updated automatically over time (current version gpt-4-32k-0314, also supported until June 14)." 🤯
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
what I'd like to see more in the AI space: a lot less white papers, a lot more GitHub repos and notebooks
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
believe that the ratio of GitHub repositories that are mostly or entirely in English versus those in interpreted/compiled languages is a good indicator of whether large engineering teams are breaking into smaller, highly functioning companies. if you can't fit into a smaller, highly capable team, then you may be out.
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
it’s incredible how much more useful outputs are from applications built with a strong foundational knowledge graph at their base. significantly less hallucination.
1 reply
0 recast
0 reaction

steve pfp
steve
@stevef
is there by any chance anything in the works that allows me to delegate my wallets to a "virtual wallet" and that virtual wallet is my identifier for the purple network? as of right now, i believe i just have one wallet associated with this account. limits associating assets I have partitioned for various reasons
1 reply
0 recast
0 reaction

steve pfp
steve
@stevef
LLM stacks evolving are going to help build insanely powerful apps. Pipelines for heterogeneous data corpus are currently difficult to work with. They’re currently at the AJAX/PHP stage. Soon jQuery/Ruby on Rails. That’s when web dev became enjoyable and I suspect NLP hacking will become more fun to work with also.
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
a great way to break my brain is to put me into an editor without vim bindings. like a deer in headlights. i can barely navigate the field. i reach for the mouse more than normal.
0 reply
0 recast
0 reaction

steve pfp
steve
@stevef
the real heroes in the AI world are the ones that curate datasets. everyone else has it easy
1 reply
0 recast
0 reaction

steve pfp
steve
@stevef
So jupyter notebooks in vscode == good. Jupyter notebooks in vscode with vim bindings enabled == not so good.
0 reply
0 recast
0 reaction