The easy way to avoid this trap is for founders to build a company that benefits from 'riding the wave' - i.e. your product / service gets better and better with each new product release from the frontier labs. Very few people seem to do this.

Aaron is precisely right. In most industries, but especially finance, people have a very hard time seeing how AI models WILL outperform humans on most tasks, especially those that involve processing large amounts of granular data. With every model update these people have less

If teens and twenty somethings can start and build multi billion dollar companies, surely some of them can also run a $3M ebitda sell side. To me this is almost entirely a talent quality issue - both at the top and the bottom of the org - rather than about their age.

I think about this a lot - people of my ‘vintage’ are now becoming MDs in finance - what are the 2nd order effects? Will the fresh blood beat out century old ‘traditions’ / ‘culture’? I hope so.

At my company we love using AI to EXPAND. Take a transcript of a 2 hour brainstorming session (we record in person meetings for this reason), take it away and continue brainstorming with a reasoning model as long as you like - tease out tensions, leverage the full context etc.

not to mention going from 80/20 to 90/10 to 99/1 gets logarithmically harder. you can’t really vibe code production grade / enterprise grade apps (at least not yet, not today). building products is still really hard.

This hits hard. Few grinders can actually follow this. Their whole identity is built on agency + sacrifice. Crediting luck feels like you’re erasing your entire story.

this has got to be one of the cooler websites out there. well done PostHog and whoever your cracked designers are!!

they had to add the word "megabank" because they knew OffDeal beat them to the punch 😉 but like i said before, it's not just about the software - you need to innovate on org chart + incentives to get this right.

These are the stats for PE outreach in the landscaping industry. Whoever runs that survey should get into private equity, since they were able to find the 28% of the cos that never get hit up!!

Been getting lots of messages on this morning. Some thoughts: Sam Altman has said this countless times - build with the assumption that the models will keep getting radically better. Spending months building scaffolding to patch current model limitations has been a losing bet

$150/hr is about 4-5x premium to what IB Analysts make per hour today. Feels cheap actually, since presumably this capex would be amortized over millions of hours worth of labor w/ AI.

Zero surprise here, if true. This is why we decided to build our own investment bank de novo - designing the org chart and comp structure for the AI future from day 1. Much better career path for high slope junior bankers than BB imo.

Curious how OpenAI approaches context engineering in this feature. I.e. is it basic RAG (pre-processed) or more similar to Claude Code's "grep' function (just-in-time) etc. Putting the most relevant data into context at the right time is a v hard technical challenge.