Jake

@teortaxesTex super interesting but again scary bc its potential to make interpretability harder. a man's mind is his last temple

Incredible. a near-frontier model we need research for direct J-space optimization now

@Noahpinion is it possible that a technology as miraculous as AI is just what it takes to preserve the current growth rate?

10x the Industrial Revolution in an equivalent amount of time, or one Industrial Revolution in 1/10th the time, would correspond to a 17.5% annual GDP growth rate.

@paulg i remember @karpathy saying something similar about GDP growth and the emergence of AI; it may be that just keeping the trend necessitated a technology of this magnitude. it may be more continuous than it feels

It's interesting in this case how well AI fits into existing trends. Programming wasn't evolving in this direction because AI was coming. No one knew it was. And yet we end up with what looks like a smooth acceleration along much the same path.

Before vibe coding became a thing, programming was already evolving in that direction. It already increasingly consisted of installing and configuring stuff other people wrote, without reading the source.

@thsottiaux feature request: please enable sol to create luna/terra subagents, so the user can request a subagent that is the right cost/speed/intelligence blend for the task

Build your dreams

ChatGPT Work presents

@amritwt and now that grok build is ... compromised ... cursor's reputation is critical for any enterprise hope

It’s Cursor that put xAI on the map Grok 4.5 is phenomenal Very good

@PatrickHeizer @RuxandraTeslo back to oral culture perhaps? though i bet even verbal expressions will become ai inflected even if not ai generated

@RuxandraTeslo I've often thought of Ted Chiang's short story, "The Truth of Fact, the Truth of Feeling," about the cognitive shift that occurred when humanity moved from oral to written culture. It seems likely to me that we are undergoing a similar shift now, for better and for worse.

I really think the biggest danger from AI is the slopification of our minds. We should be quite socially penalising of people who produce slop, especially if they're high status (like the CEO of one of the most successful young companies), because when high status ppl do things, they're much more likely to get normalised.

@jochenstu @RuxandraTeslo yes but at least then the idiotic shit had more linguistic texture & range

@RuxandraTeslo Always in favor of higher standards, but I am not sure if net slop has been going up or not? People have peddled idiotic shit pre AI (anti-vax, juicy-cleansing, etc). Maybe it's just that now we are becoming more keenly aware and critical? Would be super interesting to study.

@CursiveCrow @goodside i think they're like 200ms slices, so i guess if the model could count its own slices it could report correctly

@jake_010202 @goodside unlikely, they're still transformers.

LLMs seem to have poor intuition for how much time their thinking requires. ChatGPT 5.6 Pro:

"I think Katie, or Lizzie, was describing a holiday on Malta where, she said, the Maltese, with a death-defying insouciance quite beyond comprehension, drove neither on the left nor on the right, but always on the shady side of the road." (rings of saturn)

@scaling01 they should take a picture with everyone on site though just for kicks

even if Anthropic were to build AGI they would still be infinitely far away from the aura of this picture i don't think it will ever be surpassed

@ysu_nlp excellent, had been thinking this inchoately

There's a major gap in this otherwise compelling vision. Lots of interesting signals recently: token budget caps, intelligence sovereignty, increasing competition (Meta/Grok/GLM) that drives the commoditization of intelligence. All seem to point toward a more benevolent future equilibrium where every organization owns its human-AI learning loop through which knowledge and value accrue locally. @satyanadella, @thinkymachines, @dwarkesh_sp all laid out good arguments for it. No doubt that is a more desirable future than the oligopolistic path we are on, but no one seems to be asking about the elephant in the room: Do we actually know how to solve the continual learning problem required to sustain such a local co-learning loop? Right now, there are roughly three buckets of learning methods being entertained in practice: 1) Non-parametric learning, mostly in some form of memory (skills, RAG, knowledge bases, KV cache, etc) 2) Domain-level post-training (e.g., continual pre-training on proprietary organizational data) 3) Task-level post-training (e.g., RL for specific workflows) (More research-y ones like new model architectures are omitted here because there's still a long way to go in both research and validation) Is any of these methods sufficiently general and deep to sustain the desired co-learning loop across all organizations and job functions? The answer is likely negative: > Non-parametric memory is often shallow and has limited control over agent behavior (talk to OpenClaw/Hermes users who struggle to get their agents to learn and follow the rules) > Domain-level post-training remains expensive and has yet to demonstrate broad success outside a few exceptional domains (@Cursor's Composer may be an exception but coding is an exceptional domain in itself) > Task-level RL is engineering-heavy, sample-inefficient, and difficult to apply when success cannot be objectively verified The human-AI learning loop isn’t inevitable. It still needs to be invented. Solving continual learning may be one of the most important problems for building an AI ecosystem where expertise compounds locally instead of concentrating globally.

I go back and forth on this. On the one hand it's not super bitter lesson pilled, nd many big cos cos will prefer not to handle RL and etc internally, just like they went to on cloud instead of on prem bc they didn't want to deal with servers. On the other hand specialization and niche filling feels like an economic and natural law, and it leads to a much more appealing vibrant ecosystem of the future

Cognition and Cursor have both released near-frontier coding models based on OSS models. That is interesting...but more so as an indication of the fact that we will see domain-specific models within the next 6 months for all the high-value domains. Long Open + post-trained.

@teortaxesTex you want to be the country that other countries want to lead

retarded butchery in Vietnam did not prevent either Vietnam or several other nations from turning Communist, so no. A proper nation stops Communism from spreading like this:

@DKThomp i feel like the other thing to watch is govt involvement / investment - see 1873

I would love some feedback about my worries of an imminent AI-related market crash: Industrial bubbles are most common when firms get deep into debt. Even with declining free cash flow (chart 1 below), the AI hyperscalers still have less debt as a share of earnings than the typical S&P 500 company (chart 2). But on the institutional/retail investor side ... that's a different story. Look at Chart 3 (all from JPM). Investor borrowing is going crazy: - The amount of debt that investors are borrowing from brokerages to buy stocks, bonds, and other securities rose more than 50% in the last year to record $1.4 trillion. - Assets in high-risk leveraged exchange-traded funds have quadrupled in the last four years. Am I wrong, or does this make the odds of a major AI-related market crash getting alarmingly high in the coming months/year? Between leveraged ETF rebalancing and margin calls, I feel like one moderately bad earnings call—eg, which points to less forthcoming semi demand—could create a cascade of sell-offs And what makes this interesting is that you could have a significant market correction due to all this investor leverage, but it might not be a decisive judgment about the state of AI, at all, even if lots of people interpret it as a sign of a bubble.

@maharshii read all the great books, have five kids, invest to empower people & projects i believe in, add beauty to public spaces

what would you do if you got enough money overnight, apart from investing and (solo) travelling?

@paulg might have been autocorrect though. that happens to me a lot , super aggravating

Someone sent me an email that used "it's" for the possessive and I thought "At least it's not AI."