 | TL;DR: JD Grèze just told Harry Stebbings something about the AI-assistant market that should worry anyone still counting on a head start.
Here's the actual claim: the interval between finding something that works and someone else shipping it has gone from roughly a decade to two-to-four weeks, industry-wide, and it isn't slowing down. That's not a complaint about competition — competition was always the given. It's a claim about time itself as an asset class: the thing an under-resourced builder used to be able to bank — years of compounding learnings a slower incumbent couldn't touch — no longer exists as a form of capital. Grèze isn't describing a rough patch. He's describing a structural repricing of speed, live, from inside a company (Town) that's actually inside the race he's naming. For a founder whose entire moat theory ran on "I found this before you did," that repricing is the whole argument — being early was never the advantage; being early with time to compound was, and the compounding window is what just got taken away. A piece making the rounds this week (Artiverse, on the AI moats big tech can't copy without breaking) is asking the same question in real time: when 97% of award-nominated products are already AI-integrated, what's actually left to defend?
I remember watching this Youtube clip about this guy recounting his prophetic dream of this young lady walking down the street. She stopped by a billboard, showing a pretty model in a beautiful dress. And then, with her eyeballs, she selected the "test" icon. And the dress was "projected" onto her body, as if she was actually wearing it in her living room watching herself through the standing mirror. All these imagery projections are powered by AI into her mind. She admired herself in the dress, basking in sweet elation. She was thinking of wearing it to show her special someone A, in some upcoming event X. Of course, all these are just in her head. She's still standing in the street walk in her own clothes, in front of the billboard, with its AI projecting the image onto her mind. And then – also in her mind – her AI assistant "stepped in". It reminded her of her body shape, and her poor diet. It also pointed to the part of her body where the dress wouldn't fit. And then, it kind of say, "if you really like this dress, I'll formulate a health plan and exercise regime, for you to follow" - and projected a recommended diet and time schedule into her mind, including workouts, such as Pilate, Yoga, Strength Training, etc. – "in time for you to wear it for your A at this event X." All of a sudden, the elation in her fled. The enthusiasm was drained out of her instantly. There was a tinge of humiliation and guilt brushes over her face, as she hurriedly walks away. That's the scene. By the way things are moving now, this scene isn't that far-fetched from reality. I can see an Elon Musk's Neuralink-like personalized Claude AI LLM, residing in the cloud, linking to her mind. This AI, sounds less like her assistant, and more like her... What do you think? ____________________________ Every rung people insist can't be skipped keeps getting skipped anyway — usually by whoever has more capital, not more insight. That part isn't new. What's new is how fast the skip happens now.
Actually, this reminds me of something — a while back I broke down a VC's take on why "the labs will build it" is the wrong startup fear — her argument was that incumbents don't chase everything, only what's already proven. Grèze's version of the same room is uglier: "proven" now takes two to four weeks to notice.
Founders here — if your moat used to be "I got there first," what replaces that once first stops meaning anything? Drop your take.
Clip credit: 20VC with Harry Stebbings — full conversation with JD Grèze on their channel. DM for credit or removal requests.
[link] [comments] |
Predictive technology: computational social science
