Stan Kirdey

clark

@nomnom2077 could happen, but not until all edge cases of software and product can be solved by coding agent 100% of the time, which is almost impossible with current autoregressive paradigm

@stan_info Or reduce the count of software engineers....🥴

there are ~2M software engineers in the US there are maybe 40 million (and growing with vibe coders) in the world and it creates significant compute constrain on GPUs imagine when entire world moves into agentic AI we either need to get a lot more efficient intelligence, per byte or i don't know

1.5T seem to be a sweet spot for all major labs

@wryzx maybe not, but if it is as easy as talking to a human with agency then maybe yes

@stan_info you cannot make my mother "move into agentic AI" ... the disconnect to the real-world is huge in AI Twitter

our civilization propelled after inventing language and then inventing wheel, now we invented large language models and robots i think robots will only need to be slightly smarter than a wheel to be useful - heavy duty manufacturing and repetitive mechanical labor 24/7 language models on the other hand can truly unlock something that suppose to come next in this quest

@Ravenismeee that is called congress

Realistically, what’s stopping someone from taking out a huge loan at like 90 years old and spending it all before they die

VCs snarky question of "what if google builds it" has never been less relevant

@VraserX both, plus a mega echo chamber where messaging is amplified to levels never before seen. everyone will have their own version of the matrix.

When AI can generate an entire Netflix season, a AAA game and a custom actor just for you in minutes… does entertainment become more meaningful, or do we drown in infinite content? 🤔

the nature of enterprise engineering was often been writing internal systems and tools slop. So LLMs naturally accelerated ability to create one more service, one more internal prototype for accounting team, one more framework to combine all other frameworks in one place. But the value of these outcomes are not clear. There is just not as much high states engineering work at enterprises. Where we will see value is LLMs unlocking billions of dollars in new revenue or obvious indisputable savings. The only way to achieve it know is to limit workforce and keep only strong AI "supervisors" that could support projects. The next stage is AI actually getting into artifact generation that generates unique, hard to reproduce value - and this is a very moot point right now. I am not sure what it will look. For compute intensive enterprises - maybe new state of the art optimizations in compression, cpu allocation, gpu optimizations. For human labor intensive enterprises - still day 0.

is anyone solving quantum computing with current language models?

LLMs are interesting continuation of human intelligence - not perfect, can not really be trusted, occasionally useful but requires rigorous peer review.

LLMs still are extremely expensive for general public to automate their lives. But we will get there.

@SamRo it will go up because AI

Of all the trillion dollar companies, SpaceX has the best livestreams. Will be interesting to see how the stock trades

@alokbishoyi97 so true 😵‍💫

@stan_info another way to look at it is research will become running and orchestrating autoresearch :)

autoresearch will replace most of research

do not work on another saas