Graham claims the objection shows investor ignorance or bias.
"What if the model companies do this?" is the new "What if Google does this?" I.e. the meaningless question investors ask that shows either that they're stupid or that they dislike you and are looking for ways to find fault.
Positive users praise Paul Graham for calling investor questions about AI model companies copying big players uninsightful, while negative users dismiss his take as dumb or lazy.
@paulg but ...
@aphysicist I'm not saying that the model companies won't kill startups, just that it's so unpredictable which they might kill that it should rarely be a factor in investment decisions.
@paulg It sure seems like human VCs will be out of business long before startups. The inefficient process to identify opps and allocate money won’t be missed. If I were that VC, I’d spend more time answering my own question for myself.
@AdamTahir_1 Mostly by ignoring such investors.
@paulg and just like google, they'll announce the competing feature, half-build it, then quietly kill it 18 months later
Blue pill or red pill?
@paulg I feel like the question misses founder-market fit
if you’re truly obsessed with a niche, you notice workflow details, edge cases, language, and trust gaps that a model company won’t care enough to get right
they can copy the feature they rarely copy the obsession
Two things wrong with this… 1. LLMs are constantly putting out new features that disrupt an existing wrapper start up. 2. The LLM vertical application wrapper is pretty unique to this era of innovation. Yes, “wrappers” have always existed, but not quite like this
So I think the question is valid
@paulg Founders should probably reply: "What if a bank does what you do?"
@paulg true!
@paulg Fair, though I doubt anthropic and openai's graveyards will match google's
@paulg the funniest part is half the time the model companies actually do ship it three months later and the answer is still "so what, you built something on top of it"
@paulg Google asked it about Facebook. Facebook asked it about Snapchat. Snapchat asked it about TikTok.
The answer is always the same: execution beats paranoia.
@vernic_daniel @paulg "What if Google builds this?" was also always a valid question. Many startups failed to compete with Google.
@paulg the question was never about Google or models - it's about whether the investor understands moats, and asking it out loud answers that
@paulg Considering the volume of startups working on task-agnostic model adjacent functions (like agentic memory, pincites, and execution) the question is pretty logical.
@paulg real problem is when your users asks this on day1 when you have just started attacking a problem.
why can't i do this on claude / why can't we build this internally?
answer i believe is prove outcomes.
@paulg Back office tasks and implementation are the most exposed aspects of any potential expansion for LLM to produce applications.
Front office tasks needs localization, and every big incumbent in the past couple of years, they always looked for localization to boost their offer.
@paulg "they're stupid" This might be in most cases cause the investor is not really looking for the ideas where he has the knowledge but ideas where he can make money so he might know anything about your field so come up with questions which might not sound reasonable.
@paulg Perhaps a better question is: “What if the model becomes smart enough to do this on its own?” A $20-per-month subscription that everyone already has would beat whatever you’re selling.
Graham claims the objection shows investor ignorance or bias.
"What if the model companies do this?" is the new "What if Google does this?" I.e. the meaningless question investors ask that shows either that they're stupid or that they dislike you and are looking for ways to find fault.