silentbob

Karma: 1,759

silentbob 14 Mar 2026 20:40 UTC
3 points
0
on: Probably you won’t be able to perform a data-driven habit stacking for self-improvement
While the comments provide a lot of counterexamples, I think the post still makes a very good point. I’ve done some self experimentation, see eg my melatonin self RCT, and I’m currently running a ~150 day experiment on several mood + productivity interventions in parallel, and I have to say, the power analysis beforehand is always disappointing. Even at 150 days, I’m basically biting the bullet of low statistical robustness of my findings, as I wasn’t willing to commit to doing this for a year or two. Additionally, this experiment can’t be blinded, so I can’t even be certain I’m measuring more than reporting bias (at least for some metrics). If I’m honest, I’m probably mostly doing it because I love data analysis and just look forward to that part. Ideally, I’ll get some insights out of it, but it’s unlikely they’ll be super surprising rather than just weak evidence roughly in the direction I already expect.
I once heard someone make the argument that self experimentation is worthwhile, but if you need statistical tools to evaluate it, then you’re doing it wrong, and you should rather look for effect sizes large enough that you easily and confidently notice them without calculating p-values. Seems like a valid claim to me. As long as there are high-variance things to try that may work amazingly well for you, it surely often makes sense to prioritize these rather than your average “this may improve my mood by 3%” intervention.

silentbob 10 Mar 2026 7:45 UTC
5 points
1
in reply to: Brendan Long’s comment on: silentbob’s Shortform
True, it’s possible larger context windows aren’t even needed and 1M is sufficient for the majority of our economy to get automated.
I also think it’s easy to underestimate how much context humans actually gather over the years though. E.g. in my job there’s a huge amount of information I picked up over time. And I never fully know in advance what subset of that information I might need on any given day. It would be futile to even try to write down everything that I know, because much of that knowledge is latent/fuzzy/hard to put in words/seems irrelevant but isn’t necessarily.
To list a few such things:
- Company culture and structure
- Teams and responsibilities
- Many dozens of co-workers, their tenure, skills, personalities, common memories, what they look like, their voices
- Dozens of tools, how to use and navigate them, when and why to use them, when and why they were introduced
- A huge code base, or at least many many bits and pieces of it
- The product(s) including future roadmap and past development, some known issues and limitations, their design
- Context about how users interact with our software
- Our competition and how we relate to them
I’d assume that my visual knowledge alone (what products, tools, people, logos etc look like) could fill a significant part of a 1M context window (given the current state of the tech).
I recently tried to compile one really thorough readme for LLMs about one project I had worked on. I think it ended up at around 50k tokens, but it was very far from complete, as I have so much latent knowledge about it that I can’t just easily export on demand—it just lives somewhere in my brain, stashed away until some situation arises where I actually need it. That said, it’s possible that “the essence” of that knowledge could be compressed to, say 10-20% the amount of tokens, which would indeed make your argument very plausible.

silentbob 9 Mar 2026 23:45 UTC
24 points
6
on: silentbob’s Shortform
Often, qualitative differences turn out to be quantitative, especially in AI progress. As The Bitter Lesson pointed out in 2019, jumps in capabilities often don’t need some breakthrough or human ingenuity, but merely (much) more of the same, that is, scaling up the compute. And so we went from GPT2, which could produce English text with mostly flawless grammar but not much more, to the multilingual GPT3.5 that could write entire essays, to later models that are coming for most white collar jobs.
This naturally raises the question which other limitations exist in AI that seem qualitative, but end up being pretty much solved by the same thing but bigger. I wonder about three areas in particular:
1. Continual learning
2. Reliability & hallucinations
3. Multi-modality much closer to the human experience (something like audio-visual with depth- and time perception)
For all of these, it’s tempting to claim that they require some big breakthrough or entirely different approach than LLMs, and that the default would be that these current limitations will pose natural upper bounds to the impact of LLMs on our world. And I can well imagine that certain breakthroughs could greatly accelerate progress in these areas. But I also can’t help but suspect that even without major breakthroughs, we’ll inevitably see serious progress on these fronts anyway.
Continual learning: context window sizes didn’t see the rapid progress of some other areas & benchmarks, but even then, today’s frontier models have ~10x the context window of 2023. It’s not the primary thing labs are optimizing, but it seems overwhelmingly likely to me that algorithmic + hardware progress will lead to larger context windows of the years. And if we do reach 10M or 100M token context windows eventually, I wouldn’t be surprised if that (combined with other capability improvements) will be sufficient to make in-context learning capable enough to mostly alleviate the need for true continual learning for most economically valuable purposes. Sure, if somebody figures out true scalable & robust continual learning, then that’s an even bigger deal^[1]. But I’d argue that even if this for whatever reason does not come to pass, merely scaling up context window sizes could eventually be sufficient to surpass the “context persistence advantages” of humans.^[2]
Reliability & hallucinations: some people assume that LLMs will always hallucinate and it will take a fundamentally different approach to overcome this. Maybe they’re right, but at least in agentic coding we see that if you get the feedback loops right and “tether the model” to some verifiable part of reality, hallucinations mostly become a non-issue. It’s unclear to me how far this will actually work & scale in other areas, and Sam Altman’s prediction from 2023 that two years from then “we won’t still talk about” hallucinations certainly turned out to be incorrect. But I wouldn’t be surprised if some relatively marginal changes, such as forms of embodiment^[3] or best-of-n style answers, or whatever other surprisingly simple strategy will be identified in the meantime, end up increasing reliability greatly.
Multi-Modality: in principle, a larger context window might allow just providing an LLM with 100s of images representing some form of livestream from a camera (or two), and appropriate training or reasoning might allow it to “perceive” movement. On the one hand, I’d think that it’s a huge disadvantage for the LLM if the “time modality” is not properly represented in the way its inputs are tokenized^[4]. But on the other hand, it still seems conceivable that even such a suboptimal encoding of movement as “100 separate tokenized still images” could be handled by more advanced LLMs well enough to basically solve current limitations of LLM perception^[5].
I’m not claiming that any of this is what is going to happen. Multi-modality in particular seems like something labs could expand a lot if it was a priority, but they just happen to focus on other areas that are more lucrative on the current margin. Either way, the point of this post is just to point out that I do think that these developments may be a bit of a lower bound of AI progress. Even if no major breakthroughs occur, I’d still assume we eventually end up
- with in-context learning capable enough to surpass humans in many areas where we would currently assume continual learning to be required
- with fewer and fewer hallucinations in many areas
- and with AI models that can perceive the world in very similar ways to us, in so far as that’s helpful for the area they’re deployed in (and in many ways that may go way beyond the limits of human perception)
1. ^
  And to be fair, my best guess is that continual learning will see some breakthroughs in the next 1-3 years and will essentially get solved.
2. ^
  Somewhat relatedly to this, I also get the impression that much of what’s currently happening in the AI coding landscape (around skills, MCPs, agents/claude.md files, memory, context management...) is to some degree “overfitting” on the current margin of AI capability and will in future generations get obsolete once LLMs become better at building & persisting meaningful context themselves dynamically. We’re in this fun phase, where humans can still teach LLMs a lot to make them more useful, but I highly doubt this phase will last very long.
3. ^
  My thought here being that some form of embodiment “nails” the AI to reality and (directionally) prevents it from spiraling out of control in strange failure modes; of course, it might still turn psychotic for various reasons, but having a constant stream of “reality” almost certainly might have some grounding influence compared to its current reality that largely consists of its own thoughts, system prompts and the ramblings of its conversation partner.
4. ^
  E.g., CNNs seem conceptually nice in that they encode a certain prior about the modality of images, in that neighboring pixels tend to be more relevant for each other than more distant pixels. Similarly and reversed, providing frames of a video as entirely separate images just seems lacking, as the temporal connection isn’t really encoded, but just kind of “interpreted into it” after the fact.
5. ^
  To name one example of the limitations I mean here: if you’re working on a website and add some subtle animations to improve UX, this is something today’s coding agents have a very hard time testing. They can generally use browsers, click around, look at different screenshots, but this usually happens “one screenshot at a time” and does not include animations. They can still implement animations, and often do a good job at that, but they’re typically doing this blindly. Any human, on the other hand, who would use this website, would instantly and automatically perceive animations, and notice when they’re off in any considerable way.

silentbob 24 Feb 2026 20:52 UTC
9 points
0
in reply to: Error’s comment on: Large-Scale Online Deanonymization with LLMs
I expect that ⁹⁄₁₂₅ rate to climb quickly
DoxxBench here we come...

silentbob 19 Feb 2026 19:55 UTC
3 points
0
on: What is the most impressive game an LLM can implement from scratch?
About the out-of-distribution game: perhaps they had an internal “competition” for employees to one-shots games and then chose the best ones? I’m not sure what would give you the confidence that such a process would be too cherry-picky for OpenAI. To me this seems like a typical tech company approach, but who knows.
I agree the general question about the best possible games different LLMs can create is very interesting and informative. My experiences so far have been mixed, but I also micromanaged the process a lot. Round-based generally seems to be easier than realtime, as they struggle to get the controls feel right, and can’t e2e test realtime creations properly. And prompting these things is also super finicky. My impression is also that their judgement about which things are important in a game are really unreliable—or maybe I’m just too opinionated, hard to say.

silentbob 11 Feb 2026 7:11 UTC
3 points
0
in reply to: papetoast’s comment on: papetoast’s low quality shortforms
“was not associated” tells us more about the sample size than the effect, as far as I can tell, though, doesn’t it? The 0.82-2.42 CI does not seem very reassuring. Especially given this is just observational—it could well be that people who brush immediately after intake of something that’s bad for your teeth are generally conscientious about their dental health, so if they still end up with worse outcomes in this study (albeit not reaching statistical significance), then brushing quickly after acid intake could potentially be even worse than this CI (weakly) suggests.
That said, the measured odds ratios for fruit/acids between meals were so much larger that it might indeed make more sense to focus on these than on the exact timing of brushing.

silentbob 8 Feb 2026 9:39 UTC
34 points
5
on: silentbob’s Shortform
After using Claude Code for a while, I can’t help but conclude that today’s frontier LLMs mostly meet the bar for what I’d consider AGI—with the exception of two things, that, I think, explain most of their shortcomings:
1. lack of real multimodality
2. context window limitations
Most frontier models are marketed as multimodal, but this is often limited to text + some way to encode images. And while LLM vision is OK for many practical purposes, it’s far from perfect, and even if they had perfect sight, being limited to singular images is still a huge limitation^[1].
Imagine you, with your human general intelligence, were sitting in a dark room, and were conversing with someone who has a complex, difficult problem to solve, and you do your best to help them. But you can only communicate through a mostly text-based interface that allows this person to send you occasional screenshots or photos. Further imagine that every hour or so you lose your entire memory & mental model of the problem, and find yourself with nothing but a high-level and very lossy summary of what has been discussed before.
I think it’s very likely that under such restrictive circumstances, it’s just very hard to not run into all kinds of failure modes and limitations of capability, even for the undoubtedly general intelligence that is you.
So, in some sense, I’d think that there’s an “intelligence overhang”, where the raw intelligence that exists in these LLMs can’t fully unfold due to modality & context window limitations. These limitations mean that Claude Code et al. don’t yet show the effects on the economy and world as a whole that many would have expected from AGI. But I’d argue it makes sense to decouple the actual “intelligence” from the limiting way in which it’s currently bound to interact with the world—even if, as some might correctly argue, modality & context window are just an inherent property of LLMs. Because this is an important detail about the state of things that, I suppose, is neither part of most of the definitions people gave for AGI in the past, nor of the vague intuitions they had about what the term means.
1. ^
  as opposed to, say, understanding video, including sound, and including a sense of time. (This is not to say that vision is necessary for general intelligence, of course; but that’s kind of my whole point: the general intelligence is already there, it’s just that the modality + context restrictions mean AI is still much less effective at influencing the world in the way that a “naively” imagined AGI would)

silentbob 3 Feb 2026 7:47 UTC
2 points
0
on: Empiricist and Narrator
It seems to me that narratives are skewed and highly simplified abstractions of (empirical) reality that then are subject to selection pressure, such that the most viral ones (within any subculture) dominate, where virality is often negatively correlated with accuracy. Yet, when hearing narratives from people we like and trust, we humans seem to have deeply ingrained urges to quickly believe them. This gets most apparent when you hear the narratives other subcultures are spreading that affect you or your beliefs negatively. Like, hearing the narratives of AI skeptics & ethicists (say about AI water usage, about AI not being “actually intelligent”, or about all AI doomers secretly trying to inflate stock prices) really drove home a Gellman-Amnesia-style realization for me, how deeply flawed narratives tend to be, and that this is very likely true for the narratives I’m affected by (without even realizing these are narratives!).
Narratives are usually a combination of an overly simplistic conclusion about some part of the world paired with radically filtered evidence. (And I guess this claim in itself is a bit of a narrative about narratives)
I agree with you though that narratives may be required to actually do things in the world and pure empiricism will be insufficient.

silentbob 2 Feb 2026 18:36 UTC
3 points
0
on: The limiting factor in AI programming is the synchronization overhead between two minds
I read your title and thought “exactly!”. I then read your post and it was pretty much exactly what I expected after reading the title. So, ironically, it seems like you perfectly compressed the state of your mind into a few words. :) But to be fair, that’s probably mostly because we’ve made very similar experiences and doesn’t translate to human<->LLM communication.
When vibe-coding, many things work really fast, but I often end up in these cases where the thing I want changed is very nuanced and I can see that just blurting it out would cause the LLM to do something different from what I have in mind. So I sometimes have to write like 5 paragraphs to describe one relatively small change. Then the LLM comes up with a plan which I have to read, which again takes time, and sometimes there are 1-2 more details to clear up, so it’s a whole process, and all of this would kind of work naturally without me even noticing if I were to write the code.
A year ago I wrote a post in a somewhat similar direction, but the recent months of vibe coding with Opus 4.5 really gave me a new appreciation for all the different bottlenecks that remain. Once “writing code” is automated—which is basically now—it’s not like programmers are instantly replaced (evidently), we just hop on to the next bottleneck below. So, the average programmer will maybe be sped up by some percentage, with only extreme outliers getting a multiple-fold increase in output, and the rest merely shifts to focus on different things in their work. It’s still kind of mindblowing to me that that’s how it is. Perhaps it gets “solved” once the entire stack, from CEO to PM to testers to programmers, is AIs—but then I guess they would also have to communicate via not-flawlessly-efficient means with each other (and sometimes themselves, until continual learning is solved), and would still run into these coordination overhead issues? But I guess all that overhead is less notable when the systems themselves run at 100x our speed and work 24h/day.

silentbob 26 Jan 2026 7:22 UTC
9 points
3
in reply to: NoUsernameSelected’s comment on: To be well-calibrated is to be punctual
Even with a car, there are cases where traffic and/or finding a parking spot can cause huge variance. It really depends on the type of meeting / circumstances of the other people whether it’s worth completely minimizing the risk of being late at the expense of potentially wasting a lot of your own time.
E.g., when I visit somebody at their home, then it will likely be bearable for them to welcome me 10 minutes later. Whereas if we meet at some public space, it may be very annoying for the person to stand around on their own (particularly if the person has social anxiety and gets serious disutility from the experience).
That all being said, probably the majority of minutes that people are late to things are self-inflicted, and I agree with OP it makes sense in general to reduce that part (and more generally striving to be a reliable person).

silentbob 25 Jan 2026 20:21 UTC
5 points
2
on: Reinventing the wheel
I can relate to a lot of this. But I think in my case the motivation for reinventing the wheel also comes down to fundamentally not enjoying activities like “reading documentation” or generally “understanding what another person has done”. But implementing my own library is usually fun. And I can often justify it to myself (and sometimes others) because it will then match the given use case perfectly and will be exactly as big/complex as needed, rather than being some huge highly general universal solution full of bells and whistles we won’t even need. Which can be a real advantage—but it’s also just one side of a trade-off, and I tend to weigh that side more highly than others, for probably rather self-serving reasons.
I once heard from a developer friend that he sometimes just reads things like the Docker documentation for fun in his spare time. It gave me great appreciation for how different people can be and how difficult it really is to overcome the typical mind fallacy… :) I never would have thought people can enjoy that. And now I’m interested in somehow finding that same enjoyment in myself, because I think it would make many things much easier if I could overcome that aversion that keeps pushing me in the direction of reinventing all the wheels.

silentbob 19 Jan 2026 23:47 UTC
2 points
0
in reply to: Amalthea’s comment on: leogao’s Shortform
I’m not sure what you’re hinting at, but in 99.9% of cases when I’m out of the house, I do carry a smartphone around. If you mean that it’s annoying when the display gets confused by water, then I agree that’s a real disadvantage (but I doubt people’s attitude towards being exposed to rain changed that much between 2006 and today, so there certainly is some severe general dislike of rain independent from smartphones). If this is not what you mean, then please elaborate. :)

silentbob 19 Jan 2026 23:44 UTC
2 points
0
in reply to: cubefox’s comment on: leogao’s Shortform
Agreed, that’s one of the exceptions I was thinking of—if you’re getting soaked and have no way to get into dry clothes anytime soon, there’s little way around finding that rather unpleasant. But I’d say 95% of my rain encounters are way less severe than that, and in these cases, my (previous) attitude towards the rain really was the main issue about the whole situation.

silentbob 17 Jan 2026 10:33 UTC
4 points
0
on: Comparing yourself to other people
People compare things that are close together in some way. You compare yourself to your neighbors or family, or to your colleagues at work, or to people that do similar work as you do in other companies.
Isn’t one pervasive problem today that many people compare themselves to those they see on social media, often including influencers with a very different lifestyle? So it seems to me that comparisons that are not so local are in fact often made, it primarily depends on what you’re exposed to—which to some degree is indeed the people around you, but nowadays more and more also includes the skewed images people on the internet, who often don’t even know you exist, broadcast to the world.
But maybe this is also partially your point. Maybe it would theoretically help to expose people a lot to “the reality of the 90s” or something, but I guess it’s a bit of an anti-meme and hence hard to do.
I agree that telling people how well off they are on certain scales is probably not super effective, but I’m still sometimes glad these perspectives exist and I can take them into consideration during tough times.

silentbob 17 Jan 2026 10:15 UTC
8 points
8
in reply to: leogao’s comment on: leogao’s Shortform
Relatedly, at some point as a teenager I realized that being exposed to rain is actually usually not that terrible, and I had just kind of been accidentally conditioned to dislike it because it’s a normal thing to dislike and I never met anyone who appeared to enjoy the experience. But turns out, once you stop actively maintaining that resistance and welcome the rain, it can be pretty nice to walk around in rain while everyone around you tries to escape it. (Some exceptions apply, of course)

silentbob 9 Jan 2026 12:05 UTC
2 points
0
in reply to: gjm’s comment on: silentbob’s Shortform
Yeah, fair enough. My impression has been that some people feel guilty about caring about themselves more than about others, or that it’s seen as not very virtuous. But maybe such views are less common (or less pronounced) than the vibes I’ve often picked up imply. :)

silentbob 8 Jan 2026 10:23 UTC
15 points
5
on: silentbob’s Shortform
Egoism has a bad reputation, but I think that doesn’t do it justice. Some degree of egoism is likely very helpful, as it’s a form of ensuring available local knowledge is taken into account. If people were not at least mildly egoistic, a great deal of local knowledge would be ignored, leading to everyone supposedly helping others in not-actually-helpful ways.
What I think is much more harmful overall is the distinction between valuing public goods^[1] at least somewhat vs not valuing them at all in their decision-making. This is something I’ve in particular seen in some B2C companies I’ve been involved with: when things are going well for them, they’re proud of the value they produce for the public (as in, typically, their paying + non-paying users). But when the market gets tough and their growth/existence is at risk, they often very quickly stop caring about public goods entirely and start making very one-sided trade-offs that are (supposedly) beneficial for them while often being incredibly annoying to many users, when other, similarly-beneficial-to-them solutions might exist that don’t come at the expense of users. Two examples:
1. A software company I once worked at noticed that there was some account sharing between paying users. Their solution was to aggressively limit active logins to a single device / browser, so whenever you log in somewhere, you are logged out everywhere else. My best guess was that this may have increased sales by perhaps 0.3% at best, while being annoying to a large fraction of users. (Ultimately, I think it was even net negative for the company itself, as annoyed users are more likely to churn, but that’s a topic for another time)
2. Staying a bit vague here, but in another instance, a feature that was frequently used by hundreds of thousands of non-paying users, and which caused negligible maintenance effort, was entirely removed because it didn’t measurably drive conversions
So, the problem in such cases is not so much that the company cares about their own growth, the problem is when they completely disregard the (potential for positive) externalities^[1], compared to only mostly disregarding it, but at least having it represented in their model with a non-zero weight.
There are surely different reasons for why such absolute disregard for public goods can occur. Some speculation:
- zero-sum thinking where one implicitly assumes that the only variable in question is “benefits me” vs “benefits the public” and not seeing that these can move independently
- valuing something a bit vs not at all can increase the complexity of your decision and add to mental exhaustion^[2]
- diffusion of responsibility, where individuals inside the company may care about providing value to the public, yet nobody feels entirely responsible to defend this perspective in big decisions
- and, of course, Goodhart’s law and optimizing what you can measure, which often does not include the real value you provide, but instead mostly easy-to-hack superficial metrics and costs/benefits to yourself
I don’t pretend to have any solution for this. But my impression is that some of the decision-making, at least in the companies I’ve seen, tends to be highly path-dependent, and a good argument or suggestion made to the right people at the right point in time can make a huge difference. So I guess, even if this approach doesn’t scale all that well, having well-meaning individuals within companies occasionally speak up and make productive proposals could move some needles.
1. ^
  I can imagine that I’m not using “public goods” and “externalities” in precisely the ways they’re usually used. I hope the post makes some sense anyway. If you know of any simple ways to phrase things more precisely, please let me know.
2. ^
  I suspect this is why even many people who care about animals and dislike factory farming prefer to not think about the topic at all rather than making decisions case by case and trading off their comfort vs how much harm is caused. E.g., when you eat at a restaurant with a lot of veggy offers, it would (for most people) be very easy to eat something without meat. Whereas when friends invite you over and cook something with meat, it would be much more costly/unpleasant to refuse eating it. Still, I know only few people who are “vegetarian when it’s easy”, yet I know many people who dislike factory farming, but give it practically 0 weight in their decisions nonetheless.

silentbob 6 Jan 2026 7:44 UTC
2 points
0
on: Claude Wrote Me a 400-Commit RSS Reader App
Thanks for the write-up! Which Claude subscription did you use for this project, and what was your experience with rate limits?

silentbob 5 Jan 2026 21:42 UTC
3 points
0
in reply to: TristanTrim’s comment on: Don’t Trust Your Brain
Thanks for the comment! I like your granola bar example. Maybe survival-relevant properties like food are just more convincing to the mind than merely “feeling great afterwards”. :)
Regarding 3 / conflict, I also realize now it’s less general than I made it sound. I mostly react that way when explicitly blamed in the “you did something wrong and you really should have known better” sense, when I have the feeling that that wasn’t obvious for me at all. But even then, it’s always interesting to observe how my impression of reality can quickly shift when the circumstances change.

silentbob 5 Jan 2026 15:47 UTC
3 points
1
on: Semantic Topological Spaces
If, on the other hand, an activation function is non-monotonic, as in ReLU, then some unrelated parts of the input space will get folded together into the same parts of the output space
Just a small technicality, but you probably mean “strictly monotonic” instead of monotonic, because ReLU actually is monotonic, right? (Or perhaps “injective” would be even closer, although I suppose in continuous spaces that’s practically the same as strict monotonicity)
Of course, your actual point here still holds.