Jake Carlson

We're quickly approaching an intelligence threshold where it won't make sense for apps to keep updating to the newest model. Yes, LLMs are already good enough for most use cases, but upgrading a wrapper from GPT-4o to GPT-5 still makes enough improvement to justify the cost. This will change soon, and labs will have to come up with new ways to make money from super advanced models.

@sbmthakur I think that'd be a great way to show mutual respect- I'm asking this of you, so you get to expect the same thing from me Since they're forcing candidates to answer these before they'll even consider them for a live interview, theres no chance for them to ask their own questions

@llm_bender Feels like they actually don't want people to apply. Maybe they should add a blog entry explaining why such responses matter.

60% of AI roles now require candidates to fill out long essay questions in the application. It sucks. Nobody wants to spend hours writing personal essays, just to get auto-rejected. From what I'm hearing, this actually backfires on companies, because top candidates won't apply.

@towheretobegin Yep, the interview process is a good proxy for what working there will be like

@llm_bender I dont think its about time; any real interview process is going to take many hours of time. it's more about respect and taste; these questions are not written by the type of person top talent wants to work for.

@towheretobegin Well said. It feels very "I'm going to save myself some time by taking up a lot more of your time" The unintended consequence is that it filters out top talent, who are busy doing meaningful work

@llm_bender these are 5 good answers to get from your candidates but asking them point blank on a form is probably the worst way to get them

@gpt_alex Yea, the off-topic personal questions are very Common App-y Thing is, I feel like college apps have a much higher chance of success, so they're worth the time

@llm_bender Recreating college app process from first principles

You're a manager at a FAANG. You have 2 options: 1) Assign a task to one of your reports 2) Outsource it to AI, which shifts liability for the output back on to...you No one wants to talk about it, but this is the real reason enterprise AI adoption is so slow.

Man, I thought Hollywood was bad... The publishing industry requires authors to pledge that they won't use *any* AI, and publicly crucifies violators Should people who are so hell-bent on fighting against AI receive the benefits it will bring? wsj.com/business/media…

@theharshith_ @StevenMusielski Well said...have held this ETH for like 9 years lol

@llm_bender @StevenMusielski I'm also into long term investing and I don't react to shrt term price changes

Latest sneaky trick to get the masses hooked on gambling: Coinbase has started sending weekly portfolio recaps, but I just noticed they only send it when my portfolio is up Only 7 emails since August 🥲

@theharshith_ @StevenMusielski Haha, I'm more of a HODLer! It's a relief to see prices coming down to earth though.

@StevenMusielski @llm_bender It's like be greedy when others are fearful.

Shower thought: there's probably someone at Anthropic who's job is to look through the latest YC batch and think "which of these can we copy and ship in the next release"

@0xAvious Yes, this week is good Or just send a calendly!

@llm_bender I think "free to chat this week?" would have sufficed this other stuff like asking for your number is crazy

One mistake I keep seeing in startup sales is being too aggressive with timelines In this case, I was willing to listen to his pitch, but he pushed to connect today/tomorrow, and even asked for my number to call me unscheduled This creates fake urgency and forces me to say no

@dionizije_fa I hadn't thought of it that way - I definitely agree that more socialist countries will put up heavy resistance to AI (we're already starting to see this), but why do you think only the US will incur enough gains to make up for job loss? Just because GDP per capita is higher?

@llm_bender Well said, but with that being the case, other industries will be much faster to protect their entrenched status, especially outside of US. The US is the only place where productivity gains can hope to offset the disruption from job loss in the transition to post labor economy.

Most of the people I talk to about AI are making a fatal mistake: they're viewing its effect on jobs through the lens of the first industry it's disrupting (software engineering) They believe coding jobs are going away, but even more will be created around building systems using human judgment. The problem with this is that there's unlimited demand for software, but most professions are zero-sum. If you’re a bank, you have to hire 20 people to handle compliance - that’s a fixed cost. If AI can do 50% of this, you’re only going to hire 10. The same is true for accountants, paralegals, administrators, etc. The market size for these is pretty much fixed, and AI is eating it.

So far I'm finding that AI coworkers are great at open-ended tasks, but bad when given prescriptive tasks Just like gen AI excels at imagining an image of a cat, but sucks when you know exactly what you want This makes them great for startups, but pretty useless for enterprise.

@blwarren @elonmusk @ChrisO_wiki If they wanted to, they would not be able to, because Starlink is putting restrictions on what they can do. Isn’t that the exact reason Anthropic was canceled?

@llm_bender @elonmusk @ChrisO_wiki The difference is that the government isn't using Starlink for weapons systems at all.

1/ Russian commentators are sounding the alarm over America's use of a new kamikaze drone against Iran, the Low-cost Unmanned Combat Attack System (LUCAS). They note that it appears to have an integrated Starlink terminal and warn that it's a serious threat to Russia. ⬇️

@BlackPill_1 @elonmusk @ChrisO_wiki That’s totally fine!

@llm_bender @elonmusk @ChrisO_wiki Yeah you caught me I am AI

@KayBee624 @elonmusk @ChrisO_wiki There was a standoff this week between Anthropic and the US government, partly about whether Anthropic can restrict its models being used for autonomous weapons. Today someone posted unmanned drones using Starlink, and Musk replied that Starlink prohibits use for weapon systems.

@llm_bender @elonmusk @ChrisO_wiki Huh?