@emollick The intuition problem cuts both ways.
The risk here is the people most confident they understand what's coming, are usually the most wrong. Calibrated uncertainty might actually be the correct epistemic state here..
@HarryStebbings This math only works if the 10x growth holds. At scale, it rarely does.
Nobody wants to pay for yesterday's growth rate with tomorrow's multiple. Series A is hard. Mega-rounds just look easy until the liquidity window doesn't open..
Google's AI is changing ad tools โ and how marketers think about strategy altogether.
Tomorrow, we get a full breakdown of what's actually shifting: the products, the performance levers, and what it means for your approach.
Virtual. Free. Mark your calendar:
๐จNew champion alert! Greg just took down a Jeopardy champion ๐ Somehow knew the French cookies shaped like cat tongues.. and clearly has not yet traveled to Dubai's island. Unreal ๐ง . #Jeopardy
@emollick The funniest thing about image model anchoring is how confidently it keeps being wrong.
Text models will at least spiral into a new direction. Image models just... redraw the same mediocre squash with increasing conviction.
PerfectSquashBench is real science!
@rohanpaul_ai Every ROI model built in 2023 is wrong. Every "we'll revisit AI in 2026" decision is already obsolete.
The people who waited for the tech to mature didn't realize it was maturing while they waited..
@aakashgupta Companies are paying more because it's cheaper than admitting they overhired engineers for work that was always coordination overhead dressed up as engineering..
@MatthewBerman The mistake most teams make (and perhaps our economy has a disproportionate dependence on): they fall in love with the frontier model during experimentation and never make the switch..
@rohanpaul_ai The news here is that OpenAI sent a memo about Anthropic's revenue.. to its own employees!?
When your competitor's accounting methodology becomes a talking point in YOUR internal comms, you've already told the market everything it needs to know..
@rohanpaul_ai The scariest part is the detection problem.
If bias travels through latent association rather than explicit wording, you can't scan for it. The agent isn't lying, it's just... off.
@lennysan@rabois Point 6 is the one most founders will resist and most need to hear.
Customer research often only reveals post-hoc rationalization. People can't tell you what they want. You need to observe how they interact w product and buy.
Build from conviction, validate with sales.
@rohanpaul_ai "Set it and forget it" was a great infomercial. Terrible product strategy. Agents amplify your judgment โ the people rushing to remove themselves from the loop are also the ones wondering why their output feels hollow..