Get live statistics and analysis of Dave Friedman's profile on X / Twitter

I mean, this is insane. I uploaded Apple's most recent set of financial statements to ChatGPT, told it to build a discounted cash flow model in Python and provide the output to me in tabular and graphical form, and less than five minutes later, I had my analysis.

I keep seeing accusations on here that various investment funds are engaged in fraud because they have paused or reduced distributions to limited partners. This doesn't strike me as the correct interpretation. Distributions to limited partners generally decline or are stopped because the fund isn't generating sufficient cash to make those distributions. And in the context of real estate investment, where many of these accusations arise, investment funds frequently use debt to finance acquisitions. Debt of course is senior to equity, which means that lenders get paid before limited (or general) partners get paid. Seniority of payment is not evidence of fraud. And debt payments have increased because interest rates have increased. All of this suggests that GPs should have much more sophisticated interest modeling operations than they do, but again that is not tantamount to fraud.

I'm never leaving the bird app

I keep seeing startups trying to sell software to small businesses. Think a dentist's office or a HVAC company or a florist or any other number of thinly capitalized companies with relatively few employees. Thing is, these companies don't have the bandwidth or cash for complicated software. They need something dead simple that is plug and play. Literally something like, say, MS Office. If your product or service is not that simple to set up and install, you're not going to get any traction selling to SMB.

A good heuristic is that if you found out about Aschenbrenner's manifesto via Ivanka, you're hopelessly late to the game.

I’ll never understand people who go to the gym and spend more time jabbering and socializing than lifting weights. What is the point? Seems so inefficient.

An observation, based on a recent experience to which I was adjacent: it is very hard for most veterans of large companies to thrive in startups, and it's even harder for most veterans of large companies to successfully found startups. Note the use of the word 'most' in the above: there are, of course, exceptions.

What in the McMansion is this

Run the same prompt through GPT4 & Claude 3 Opus and get vastly different responses. I never know which one will be better for a given prompt. Yes, non-deterministic, etc. But one reason so many people are so dismissive of generative AI is that they're unwilling or unable to put

Every time I hear stories like this I am reminded that few things are more high risk than thinly capitalized small businesses. A lot of thinklfluencers on this site extol entrepreneurship, but at the scale these people are operating at it is no way to manage risk.

Hard to see how OpenAI wins enterprise when its service goes down so much.

It seems like AI agents will be incorporated into foundational models. So I don't see much of a future for pure play AI agent startups. One exception here might be a startup pursuing a particular vertical market. Most of the AI agent startups seem to be pursuing a horizontal market, and that's deadly in commoditized technology.

Yeah I’ve seen this in person before, and it’s somewhat unnerving and off putting.

One way to do Deep Research that I don't see anyone talking about is: multiple tabs, one report per tab. Review all reports when complete and combine pieces from each report and ask Deep Research to compile a synthesis of the information presented. Works when you are trying to understand how adjacent fields relate to each other.

The funny thing about intellectuals is their predilection for making up really long words to explain simple concepts. We don’t advocate letting bad businesses fail—we advocate liquidationism.

I mean, this is insane. I uploaded Apple's most recent set of financial statements to ChatGPT, told it to build a discounted cash flow model in Python and provide the output to me in tabular and graphical form, and less than five minutes later, I had my analysis.

What in the McMansion is this

I keep seeing accusations on here that various investment funds are engaged in fraud because they have paused or reduced distributions to limited partners. This doesn't strike me as the correct interpretation. Distributions to limited partners generally decline or are stopped because the fund isn't generating sufficient cash to make those distributions. And in the context of real estate investment, where many of these accusations arise, investment funds frequently use debt to finance acquisitions. Debt of course is senior to equity, which means that lenders get paid before limited (or general) partners get paid. Seniority of payment is not evidence of fraud. And debt payments have increased because interest rates have increased. All of this suggests that GPs should have much more sophisticated interest modeling operations than they do, but again that is not tantamount to fraud.

Run the same prompt through GPT4 & Claude 3 Opus and get vastly different responses. I never know which one will be better for a given prompt. Yes, non-deterministic, etc. But one reason so many people are so dismissive of generative AI is that they're unwilling or unable to put

I’ll never understand people who go to the gym and spend more time jabbering and socializing than lifting weights. What is the point? Seems so inefficient.

An observation, based on a recent experience to which I was adjacent: it is very hard for most veterans of large companies to thrive in startups, and it's even harder for most veterans of large companies to successfully found startups. Note the use of the word 'most' in the above: there are, of course, exceptions.

I'm never leaving the bird app

I keep seeing startups trying to sell software to small businesses. Think a dentist's office or a HVAC company or a florist or any other number of thinly capitalized companies with relatively few employees. Thing is, these companies don't have the bandwidth or cash for complicated software. They need something dead simple that is plug and play. Literally something like, say, MS Office. If your product or service is not that simple to set up and install, you're not going to get any traction selling to SMB.

It seems like AI agents will be incorporated into foundational models. So I don't see much of a future for pure play AI agent startups. One exception here might be a startup pursuing a particular vertical market. Most of the AI agent startups seem to be pursuing a horizontal market, and that's deadly in commoditized technology.

A good heuristic is that if you found out about Aschenbrenner's manifesto via Ivanka, you're hopelessly late to the game.

Rippling must be desperate for customers…they want to give me $150 for listening to a sales pitch. Doesn’t inspire confidence.

Hard to see how OpenAI wins enterprise when its service goes down so much.

Every time I hear stories like this I am reminded that few things are more high risk than thinly capitalized small businesses. A lot of thinklfluencers on this site extol entrepreneurship, but at the scale these people are operating at it is no way to manage risk.

The funny thing about intellectuals is their predilection for making up really long words to explain simple concepts. We don’t advocate letting bad businesses fail—we advocate liquidationism.

Yeah I’ve seen this in person before, and it’s somewhat unnerving and off putting.