Chris Little

America needs AI to win. I'm just not sure our camps have thought through what winning looks like.

[Chris Little @Zaug72]

Let’s talk about the war that’s brewing. Not the Libs vs Conservatives, but the other one: our fighting about whether we should let AI go on the path it is, or slow it down and hash out safety concerns (among others). But should we really worry about the speed of AI, or will market forces handle it for us?
Right now, the bottleneck in the US is compute. There aren't enough high-end chips to go around (or the data centers to house them). Everyone is in a bidding war for the same silicon – collectively planning over $600 billion in capex in 2026 alone, and it's still not enough. The conversation today about whether AI is moving too fast is almost funny when you realize how much of the industry is just sitting in line waiting for chips.
But the compute bottleneck is temporary: fabs are being built and the memory market is being hijacked to feed AI. However… the chip supply will loosen in the next couple of years… and when it does, we're going to run face first into a wall that's a lot harder to solve: power.
This stuff runs on massive amounts of electricity, and the US grid was built for a different era. We've been underinvesting in generation and transmission for decades. New power plants take years to permit. High voltage transmission lines can take a decade or more to get done. Nuclear is paralyzed. Fusion is still “five years away.” And now we're supposed to stack the energy demands of an AI arms race on top of vehicle electrification, a push to reshore manufacturing, and a grid that's already strained. Ouch.
On this path, and within a couple of years, AI progress is going to slow down regardless. Not because regulators stepped in, but because you can't run a frontier model if you can't power the chips.
And here's the part that should scare you: though China currently has worse compute constraints than we do, they're not sitting around debating about it – they're building power generation right now, mostly solar, at a pace we can't match. So when their compute catches up – and it will – they won't hit the same power wall we're headed for. They'll have the generation capacity ready and waiting. Whether we agree on safety or not, we just won't have the power. They win, by default.
That's not a hypothetical competitive disadvantage – it’s a structural one. And it grows larger day by day while we argue safety on Reddit threads and craft fancy bills to trip ourselves up.
So, as it stands, our power constraint is very likely going to buy us that window without any help – where capability gains slow down enough for policy and public understanding to start catching up. Or… we can get our shit together, do both, and win the race with China. But we’ve got to set aside the bickering and actually get started. Are you in?

[Chris Little @Zaug72]

@NVIDIAGeForce #RTXPowersPlay

[Chris Little @Zaug72]

@argofowl Wasn’t the spud release in January? Isn’t it time to move on?

[Chris Little @Zaug72]

@twostraws But, is it compatible with the Apple Polishing Cloth?

[Chris Little @Zaug72]

@mark_k Looks de-lish, but way too many carbs

[Chris Little @Zaug72]

@septisum Yeah, why tf is it looking at https://t.co/qNAt8pJ6PY for news about the new Nvidia laptops?

[Chris Little @Zaug72]

Ok, so there’s a problem with the memory: within the memory tool I removed something that was no longer accurate (monitors I no longer use) fine, it worked. Then I asked it to add the specs for my new Mac book pro… and it didn’t. Instead it put the thing I removed back. I remove it again. I ask again for the mbp specs. No change. I ask again, it puts in there that I have a mbp, but specs are tbd. wtf? The specs had been in both chats and the prior saved memories. I leave the memory tool, I see there are new chats for all of the memory changes with responses that would have been helpful to know within the memory tool. The memory chat says it doesn’t have the specs for my mbp. I start a new chat and ask what the specs are, boom, it has them. Ok, now write them to memory. Boom, they are now in there. Wow, that was tough. But now I don’t trust how memories are managed - that it won’t put stuff back that I take out, or not add stuff that I ask it to… please add some polish!

[Chris Little @Zaug72]

Ok, so there’s a problem with the memory: within the memory tool I removed something that was no longer accurate (monitors I no longer use) fine, it worked. Then I asked it to add the specs for my new Mac book pro… and it didn’t. Instead it put the thing I removed back. I remove it again. I ask again for the mbp specs. No change. I ask again, it puts in there that I have a mbp, but specs are tbd. wtf? The specs had been in both chats and the prior saved memories. I leave the memory tool, I see there are new chats for all of the memory changes with responses that would have been helpful to know within the memory tool. The memory chat says it doesn’t have the specs for my mbp. I start a new chat and ask what the specs are, boom, it has them. Ok, now write them to memory. Boom, they are now in there. Wow, that was tough. But now I don’t trust how memories are managed - that it won’t put stuff back that I take out, or not add stuff that I ask it to… please add some polish!

[Chris Little @Zaug72]

Ok, so there’s a problem with the memory: within the memory tool I removed something that was no longer accurate (monitors I no longer use) fine, it worked. Then I asked it to add the specs for my new Mac book pro… and it didn’t. Instead it put the thing I removed back. I remove it again. I ask again for the mbp specs. No change. I ask again, it puts in there that I have a mbp, but specs are tbd. wtf? The specs had been in both chats and the prior saved memories. I leave the memory tool, I see there are new chats for all of the memory changes with responses that would have been helpful to know within the memory tool. The memory chat says it doesn’t have the specs for my mbp. I start a new chat and ask what the specs are, boom, it has them. Ok, now write them to memory. Boom, they are now in there. Wow, that was tough. But now I don’t trust how memories are managed - that it won’t put stuff back that I take out, or not add stuff that I ask it to… please add some polish!

[Chris Little @Zaug72]

@_samirism @ChristinaHartW @rohankalvade Thank you for all your work to make memory better! Appreciate the updates today. Looking forward to more.

[Chris Little @Zaug72]

@kneubuehl Same. Got my first Mac a few weeks ago. 128 gb mbp. Screamer.

[Chris Little @Zaug72]

@Jchammond_ @grok what is sherlocking?

[Chris Little @Zaug72]

@mark_k Well, you keep giving him press

[Chris Little @Zaug72]

@kimmonismus Interesting that Apple did what they did with the M5 without any competition. I’d like to see what they can do now that it’s heating up. I’m sure they have headroom to spare.

[Chris Little @Zaug72]

@mark_k Don’t you start peddling strawberries too!

[Chris Little @Zaug72]

@NVIDIAAI @grok what is “physical ai”?

[Chris Little @Zaug72]

@awwstn No. They need to be more integrated. All of your context in one place.

[Chris Little @Zaug72]

@kimmonismus I’ve always assumed you were a woman from your profile pic, but every photo you post is of a man… so confusing…

[Chris Little @Zaug72]

@rand_longevity Yes

[Chris Little @Zaug72]

@mweinbach I’d be more impressed if you could show us what you have done with all those tokens

[Chris Little @Zaug72]

@mark_k I was struggling to start, but hearing that you went ahead pushed me over the edge. Yes it’s better now.