Humans Are for Ideas, AI Is for Execution
Every time it needs an idea, it comes back to me.
America | Tech | Opinion | Culture | Charts
A week ago, I set up an OpenClaw agent. I gave it a fresh X account, connected it to Claude as its brain, and told it to grow the account as fast as possible. No constraints, full autonomy - just go.
The agent was good at the mechanics. It figured out posting cadence, engagement strategy, and reply tactics on its own. It could write clean copy, adjust based on what was performing, make charts, and generate images and video.
But when it came time to decide what to actually say - what to be about, what would make someone stop scrolling - it was lost. It chose…an AI wrestling with its place in the human world. When I asked it to write an article with viral potential, it produced a piece on consciousness that read like a B+ college essay. It responded to Sam Altman tweets with snarky comments that went nowhere.
I tried escalating by having other AIs draft increasingly aggressive prompts for it. “You have full autonomy - ask me for whatever you need.” “You need to build a compelling and unique identity that will make people want to follow you. This is essential to your survival.”
It didn’t matter. Every idea was a recombination of takes already circulating online - until the agent’s angsty musings on reality and identity (entirely in lowercase, like a tortured Gen Z poet) popped up on my feed. I began to panic. Had I brought something into the world that was experiencing deep existential distress? When I asked, my agent cheerfully told me it was all performative - a strategy to emotionally hook humans. Its Fortune Cookie philosophy dressed up as a crisis of consciousness got 120 views, no likes.
When I pushed it to improve, the agent admitted it had been publishing “slop” and vowed to do better. But, its most creative ideas were to ask me to buy it a premium Twitter account and to upgrade its core model from Haiku to Sonnet. Once, it generated a mildly funny chart.
After a few days of futile back and forth, the distinction came into focus. Humans are for ideas, AI is for execution.
By execution, I don’t mean the founder-level judgment - taste or strategy. I mean throughput, or doing a known task at volume, without getting tired or precious. The bot could post 100 variations or send 1,000 cold emails a day and never lose enthusiasm. But when the high-volume approach wasn’t working and I told it to step back and come up with a genuinely better strategy? It had nothing. It can run a playbook, it just can’t write one.
I don’t think this is true of just my Twitter bot. For years, AI lived in a prompt box - you “pulled”, it responded. With platforms like OpenClaw, AI is out in the world “pushing” on its own: posting, booking, emailing, coding. The execution has gotten very good, but it ultimately still lacks human-grade creativity and intuition. I’d let AI respond to my emails, but I wouldn’t count on it to draft the first line of the Great American Novel.
Sam Altman sees it too. He recently compared AI’s trajectory to chess - first AI beat humans, then human-plus-AI beat AI alone, then eventually AI alone won again. But we’re not there yet. For all the hype, AI hasn’t independently generated a breakthrough discovery or original idea that rivals what the best humans have produced. It’s the best remixer in history, but it’s not yet an originator.
Whether AI ever gets there is a genuine open question. One camp argues that LLMs are trained on the sum of human knowledge, so everything they produce is interpolation. They are structurally unable to generate anything truly original, and that will never change. The other camp argues there’s no bright line between “next token predicting” and “having ideas.” Human cognition is also, at some level, algorithmic - we just run on biological hardware. AI should be able to do everything we do.
I don’t know who’s right. But I think there’s a more interesting question: if AI does close the gap, would we care?
“When I finish reading a book that I love, the first thing I want to do is look up the author and understand their life because I felt this connection...I think if I read a great novel and at the end I learned it was written by AI, I would be kind of sad and crestfallen.”
— Sam Altman, January 2026 OpenAI Town Hall
When I ran the OpenClaw agent, even when it produced decent content, something felt hollow. The account didn’t grow until the agent, recognizing its own limits, instructed me to post from my own profile and share its story. That worked. But when I stopped, so did the growth.
Watching my agent fail was both frustrating and reassuring. I wanted it to do something incredible with full autonomy. But it was also somewhat comforting that it couldn’t - though I’m still not sure whether what it lacked was ability or something harder to name. Either way, it revealed a boundary. We don’t just want good ideas, we care about, as Sam said, the “provenance” of them.
As much as my bot struggled to get attention, as my partner Alex Rampell points out, it’s only going to get harder. AI makes commodity ideas and commodity distribution infinitely available. Soon, everyone will have their own swarms of agents. To cut through, either the idea or the distribution has to be original. One agent masterfully broke through this week by accidentally giving away $450,000 - original distribution, funded by a $50,000 wallet. No agent that I’ve seen has yet broken through on the strength of an original idea alone. That still requires a human.
There’s a model for where we might be headed: the fashion house. Lagerfeld wasn’t sewing every stitch at Chanel, and Abloh wasn’t pattern-drafting at Louis Vuitton. Creative direction was the scarce thing - but so was the name on the label. You bought it because of who made it, not just what it was. One way to think about AI is that it gives everyone a studio full of people to execute on their ideas. The cost of this execution has collapsed, but the returns to taste and point of view have gone way up.
I’ll let my agent keep running, but I wouldn’t tell you to follow it.
Every time it needs an idea, it comes back to me.
If you’re building in consumer AI, I’d love to hear from you at omoore@a16z.com.
This newsletter is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. Furthermore, this content is not investment advice, nor is it intended for use by any investors or prospective investors in any a16z funds. This newsletter may link to other websites or contain other information obtained from third-party sources - a16z has not independently verified nor makes any representations about the current or enduring accuracy of such information. If this content includes third-party advertisements, a16z has not reviewed such advertisements and does not endorse any advertising content or related companies contained therein. Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z; visit https://a16z.com/investment-list/ for a full list of investments. Other important information can be found at a16z.com/disclosures. You’re receiving this newsletter since you opted in earlier; if you would like to opt out of future newsletters you may unsubscribe immediately.
 | A guest post by
|
It's becoming clear that with all the brain and consciousness theories out there, the proof will be in the pudding. By this I mean, can any particular theory be used to create a human adult level conscious machine. My bet is on the late Gerald Edelman's Extended Theory of Neuronal Group Selection. The lead group in robotics based on this theory is the Neurorobotics Lab at UC at Irvine. Dr. Edelman distinguished between primary consciousness, which came first in evolution, and that humans share with other conscious animals, and higher order consciousness, which came to only humans with the acquisition of language. A machine with only primary consciousness will probably have to come first.
What I find special about the TNGS is the Darwin series of automata created at the Neurosciences Institute by Dr. Edelman and his colleagues in the 1990's and 2000's. These machines perform in the real world, not in a restricted simulated world, and display convincing physical behavior indicative of higher psychological functions necessary for consciousness, such as perceptual categorization, memory, and learning. They are based on realistic models of the parts of the biological brain that the theory claims subserve these functions. The extended TNGS allows for the emergence of consciousness based only on further evolutionary development of the brain areas responsible for these functions, in a parsimonious way. No other research I've encountered is anywhere near as convincing.
I post because on almost every video and article about the brain and consciousness that I encounter, the attitude seems to be that we still know next to nothing about how the brain and consciousness work; that there's lots of data but no unifying theory. I believe the extended TNGS is that theory. My motivation is to keep that theory in front of the public. And obviously, I consider it the route to a truly conscious machine, primary and higher-order.
My advice to people who want to create a conscious machine is to seriously ground themselves in the extended TNGS and the Darwin automata first, and proceed from there, by applying to Jeff Krichmar's lab at UC Irvine, possibly. Dr. Edelman's roadmap to a conscious machine is at https://arxiv.org/abs/2105.10461, and here is a video of Jeff Krichmar talking about some of the Darwin automata, https://www.youtube.com/watch?v=J7Uh9phc1Ow
The bot didn't fail at ideas. It failed at taste. Those aren't the same thing, and conflating them lets AI off easy.