inshard

so.... Weight-space programming may unlock entirely new architectures or self-improving loops.

The GitHub repo makes it seem like it’s just one dude and a Claude subscription.

Do memes have a concept of proto memes? Like primes are the proto numbers that all other numbers are made from? Could there be similar ontological primes for all concepts / semantic space? Base reality concepts that are irreducible so to speak.

More of this but Singapore and CCP core politburo should also be acknowledged.

If I had interest in FAANGS right now. I might sell now in anticipation of reallocation to Anthropic and other new entrants to the public markets in the AI space. I wonder if some of this is partly the liquidity shock an Anthropic IPO may cause for retail investors.

Bunch of dudes, on mission, hardcore, locked in, and grinding. No pun intended.

I need 64GB RAM and 1TB SSD suddenly having been more than comfortable with 16GB and 256GB respectively. My agents have sessions which have workteees and scratchpads and simulators and build runs and test suites and wow. The hours I’ve been spending for memory swaps and quad simulator calls to run through. I am compelled to view a $4k machine as a necessity. Also a cool thing about M5 Pro is that it’s very close to Max on the CPU performance side with identical core count if you get the 18 core variant. The Pro’s Geekbench scores are also in similar ballpark as the Max. You rather get the Pro for most use cases and use the extra cash on RAM.

So world models capture inherent complexity of physical reality to help teach AI about hidden laws which probably helps train both LLMs and multimodal models used for autonomous vehicles and robots? And AMD acquiring World Labs is a move to outdo Nvidia’s work in this space. And Google probably sees this as a gateway to a new platform of autonomous vehicles and humanoids that they could rent out as a CSP? Competing with Nvidia, Tesla and complementing Waymo. Am I getting that right?

Interesting that he pivoted to something more akin to a human in the later versions. And I feel akin is an interesting word here. Kin selection is well-studied aspect of how genes work at population scale; and I wonder if by just having human-like traits like in the direction your knees bend, helps make robots more agreeable to us. And the UX of full submission as a human approaches, however the optics, seems to be a pretty strong way to indicate to the human user of a non-threatening environment in their control. I should be able to clearly see if something isn't as it should be if the humanoid fails to do this upon approach and take precautions.

Superdeterminism... quantum noise and the which way paths are all correlated.

Researching this and imagining what the Romans saw in their campaigns up north, and also Asterix comics, I found this tidbit about adaptation in general: "Wolves generally follow Bergmann's rule, a biological principle stating that populations of the same species living in colder, higher-latitude environments tend to be larger than those closer to the equator."

I wonder what Darwin would observe and conclude if he were somehow brought back to Earth today. Particularly around information theory, Dawkins’ memes and selfish genes, and now a seemingly new class of something in AI.

Slightly tangential: Michael Levin’s work on bioelectricity and cell reprogramming is pretty interesting in context of emergent behaviors in groups of cells. https://en.wikipedia.org/wiki/Michael_Levin_(biologist)

Second para, line 1 in that Wikipedia link: “ Common side effects include sleepiness,”

Could explain one pathway on how Guanfacine, an ADHD-approved, blood pressure-lowering, sleep-inducing drug, works.

That paywall is techno imperialism

Nicely done Mr. Gebbia.

Vancouver Island is a special place.

My new sub-benchmark is which combinations achieve the hook at the end of the upper beak. Right now just 4: Astra Max, XHigh and Medium; GPT 6 Sol Max

Not as good as Astra or Fable 5.1 on this test as far as I can see. I wonder if any benchmark exists for artistic taste, visual sophistication etc. I think your Pelican test does touch on these aspects of a model and is useful for developers trying to build rich digital experiences (includes games, interactive websites and apps). These benchmarks are subjective so it may not be easily established and will have polarized reactions before it gains legitimacy. May even need human judgement layers adding to the cost of running it.

Thanks I meant discretionary.

I am not seeing the connection. Whenever I hear capital and elites, it’s a clear red flag around lack of understanding. And just laziness about money. I think a lot of these voices would change if they just had 20% of [edited: discretionary] income auto-invested in equities from an early age.

I just feel apps like linear are increasingly getting in the way of full send agentic development where sub agent orchestration is done through agent to agent messaging, work trees, on demand git restructuring and epoch specific coordination plains, often .md files. The smaller the human component of total product development gets, the more this may be the case.

Any real world experience with Grok Ultra $300 monthly subscription vs Claude Code Max in terms of overall built work mileage, or general token limits?

I use SelfControl for Mac and set the down time to around 15 days. No Youtube no news sites etc. Except HN. It's helped a lot, after a week you can notice measurable change in attention and habit and productivity. SelfControl lets you add 48 hrs to your blocked time repeatedly, so you can get to a 15 day mark pretty easily through repeated mouse clicks.

LOL. "I'm open to contributions, such as if you want to support exfitration using, like, power grid voltage fluctuations or something."

Not just humans… all mammals.