Garry Tan, Y Combinator President and CEO, argues builders must abandon legacy frameworks for direct, hands-on AI experimentation
Investor James Cham supported the call for active implementation.
Many users endorsed Garry Tan's call to discard old maps for exploring AI territory because traditional approaches fail in this fast-changing domain, while others dismissed the advice as vague or meaningless.
Most Activity
@garrytan Don’t forget your towel
@garrytan If AI was really as valuable as the investors in it are saying, the customers would want more of it instead of being sick of it being added without their consent, and engineers like me who know how to leverage it to multiply our provided business value would not be unemployed.
@garrytan Exactly. Everything will be open source, sooner or later. The sooner we write and adopt the perfect platform for universally decentralized everything, the faster we reach zero human tax and zero human rent with 90% droid tax.
@garrytan @grok its true?
@garrytan I have been watching the land for about 2 years now. That’s on top of a quarter of a century worth of prior tech experience. Customers don’t see the value in AI. AI has not yet become anything than a cost center. As an engineer, I think AI is neato; but token leaderboards closed.
@garrytan I still remember my first time coding that used LLMs, I built this whole deterministic scaffold trying to limit everything the AI did, and handwriting every prompt myself.
That was 2 months ago, seemed like a lifetime ago.
@garrytan Garry wtf does this mean
Walk the land!
In AI most people are still trying to use old maps on a new territory.
Throw the maps away. It's time to draw new ones. The only way you can do it is walking the land.
@garrytan If I had the money to train my own foundational models, I would be doing that. I watched the transitions through the architectures and the different components as they’ve been refined through public research.. I’ve trained my own smaller models locally. I wrote my own Agent.
@garrytan I learned about cognitive pipelines, built them. I then iterated to refine them. I studied different ontology, I have learned how to make AI do productive, repeatable work. I became distrustful of AI by default, learned to manipulate it and tame it. Nobody hires for such skills.
@garrytan The map is not the territory
@garrytan Your framing is always distinguishable. Time to walk the land 🫡
@garrytan I feel like the AI founders & investors already believe that they've replaced humans, but corporates & startups are having hard-time restricting compute costs and optimising it to human-salary levels.
@garrytan The hard part: nobody trusts a map they didn't draw themselves. So people clutch the old one even when it's walking them off a cliff.
@garrytan Yeah dude. I was furiously working with Fable until it got yanked.
Then I went outside built a house on wheels out of wood and now I don't have to work ever again.
@garrytan The old map says 'prompt engineer.' The new land doesn't have a name yet.
@garrytan The number of VR and three.js projects on X before Fable shut down is a great example of this 👆 We saw so many different and unique projects because people didnt follow the old map of world creation and used the Fable one. @garrytan
@garrytan True. In uncharted territory, how much should you trust experience vs instinct?
@garrytan It's a brave new world
@garrytan Amen, brother! All the while everyone is whining about losing the jobs they hate at the mill.