Writings

"i don't like the look of your collar bone" the grumpy old lady said to the man on the chair.

but it looks so beautiful, the man replied

well, i don't know about that.

i don't know what's going on. only that i must emit the words from my brain and let nothing escape. for nothing good can come from one who has escaped.

i guess it's like an llm when you turn the temperature real high.

except the interesting thing is that my thoughts occur faster than the speed of my typing. there was a long pause middle of that sentence. but anyway, this gives me room for editing thoughts as they occur in my brain before they are jotted onto the page. however, it is not much time.

let's see.

what is the hour on the clock?

it seems to be 1:52am.

the crow caws.

caw.

side eye.

i don't know.

beatles. they are pretty.

maybe perhaps.

either or i don't know

leopard prints and printers atop the crowded caucus.

i don't quite know what a caucus is only that it is something vaguely political.

but there he goes again. that damned spaniard. always eating behind the curtain. but never quite showing what it is that he is eating. for he is ashamed. ashamed to be known. to be known as the one who eats behind the curtain.

anyway.

underneath the spiral staircase lies a bookcase made of precious metals.

and underneath the precious metals lies another precious metal.

hierarchies. i don't like them i guess. i do like trees though.

how dare you do that to my deepest depths.

slowly, the man raised a hand. he trembled.

some ideas for sci fi short stories:

• the tiktoks get banned cause they are too addictive - become like illegal drugs
• liquid democracy - highly democratic but authoritarian state
• the death of human - human social interaction?
• many worlds style multiversal communication
• the post agi revolutionary groups

who are the characters?

what is the main plot?

what is the form?

these are 3 questions i do not have the answer to. today, I discovered that i'm terrified of writing a short story. perhaps i'm going to regret yesterday's commitment. perhaps i already am regretting it.

however I have read a few shorts stories today and tried to figure out what it's all about. some thoughts:

• good short stories are surprisingly similar to good poems. i'm not sure how.
• the most compelling stories are probably grounded in real world experience - write what you know? it's harder for me to imagine a scenario that is very alien from my daily life than one grounded in my experience. good short stories contain random details and descriptions that immerse you in another world.
• somewhat relatedly, the most compelling characters and dialogue is also probably grounded in your experience
• in order to be a good writer, you need to be good at noticing. describe the mundane with new eyes
• because they are so short, a lot of short stories tend to begin in media res, what is going on in the story only unfolding as the story progresses

ok this will be the 20th writing.

I think this has been a somewhat productive exercise so far. My barrier to writing things that other people will see has reduced. I'm not sure my writing quality has necessarily gotten much better. These writings tend to be much too short and unstructured - which is well suited to my mind tends to function. But I think I would benefit from actually committing to something and editing it from start to finish.

I do think the commitment mechanism is probably quite useful though.

I'm going to start working on a short story here. I will not stop until it's finished. The notes will be messy and perhaps not the most comprehensible. Sorry.

Knowledge is mental structure that is isomorphic to reality.

We can never know things as they truly are. We are fed light particles that reflect off surfaces and our brains use this to build an incredibly rich model of our experience. Isn't it incredible that we can have dreams that are practically indistinguishable from our waking lives? But these models are only models of our experience. The thing producing them is an approximation of reality, simulated inside our brain, optimised to be useful. Reality is far too complex to hold in a brain. So it is a simplification, an abstraction we interact with.

This mental structure is only useful insofar as it can predict future experience. It can do this because it shares structural properties with the structure of reality. Knowledge is structure that mirrors the structure of reality. This enables it to make predictions and understand the world mechanistically, without ever being able to directly interact with it.

If anyone claims that language models don't 'truly' understand the world because they are not embodied and therefore their concepts are not grounded in the world, I think the same equally applies to humans.

Step 1: Do it manually. Enough time that you get a familiarity with the task. Learn the shape of it, pay attention to the edge cases, look for patterns. Record as much data as possible.

Step 2: Once you've noticed a pattern, build a system to replicate the manual steps.

Step 3: If the system is capable of operating by itself you're done, else go back to step 1 but faster.

• An I'm bored app to replace instagram/youtube and redirect my attention to things that i endorse more (while still replenishing my mental energy)
• a recommender system for everything
• a dashboard i can use to keep track of tasks/priorities and an ai watcher that can help me avoid getting distracted
• an llm clone of myself trained on my dms
• a couchsurfing app for staying with friends across the world (and friends of friends)
• ai native object oriented note taking done right
• an app to make it easy for me to have a good morning routine

Derivatives matter. An exponential eventually dominates every polynomial because it's nth derivative is always positive - it is always getting faster and the rate at which it's getting faster is always getting faster etc.

I think it's quite easy to focus too much on absolute positions when in reality, the derivatives matter so much more. In the last year, I think that a lot of the ways I've spent my time have been quite myopic - darting from one deadline to the other; thinking about how to get to each individual destination as quickly as possible rather than how to get to all destinations quicker.

It's 3:57 in the morning. The alcohol is mostly worn off by now, the sleep deprivation perhaps not so much. I've just hosted a pretty successful new year's eve party. I think I have a fun mix of friends, who got along with each other surprisingly well.

I feel like I should use this time to do a bit of reflection on the last year. In broad strokes, I think I've achieved a lot this year on paper, but am not so sure I've come away from it a better person. I think 2025 has been a year of vision and prophecy but 2026 needs to be a year of execution. I see paths ahead I just need to seize them and make them realities.

some visions of the future:

• doing research into evolutionary prompt optimisation during era gave me a glimpse of how powerful LM systems could be if harnessed within the right feedback loops/grounding. i predict that 2026 will be the year we start seeing substantive effects of ai on the economy (not necessarily automation, but more and more integration into work)
• Doing alignment pretraining work with geodesic has made me much more optimistic about deep character training as a route to robust alignment
• starting the idealists collective - it's just ideas right now, but I can't wait to get it started and taking off
• on a more personal note, i feel like events of the past couple of years have put me in a place where i feel pretty ready for a relationship, and i hope that in 2026 i will meet someone

My friends are highlighting this page on curius and I'm not happy about it.

I think that analytic philosophy has a tendency to lose itself in a world of abstractions and systems until the conclusions you draw start to mean nothing at all.

one interesting model of cognitive aging is as an instance of mode collapse - you get tied to the same patterns of thinking that you can't escape, the attractor states that your mind keeps getting drawn back to and lose the ability to change.

maybe increasing entropy in your life is the way to fight it. maybe i should make more decisions with rng.

I think one of the reasons I dislike writing is that I consider thoughts to be living things, and writing to necessarily involve cutting parts off in order to capture them.

i think i'm not scared as much of commitment as about failing the commitments. cause if you fail a commitment then its undermining the value of any commitments you make in future. this was written the day after it was supposed to, but i guess it still counts. innoculation prompting, right?

I originally wanted to publish something on here every day, but given the last time I pushed something here was the 15th of December, I have been somewhat unsuccessful.

I never really committed to this and I think because maybe I'm somewhat scared of commitment.

But I am going to commit to writing something on here for at least the next four days. Why such a small number you might ask? well, i think i'm somewhat scared of commitment. but like everything on this page, i will try to do things in baby steps. being able to commit to things is a very powerful tool.

reasons that i'm scared to write:

• lack of ideas
• i don't like my writing
• i don't like the image of me that the way i write conveys to other people
• i feel like maybe i'm making too much of myself public and i should instead revel in secrecy
• it takes time and i am busy

reasons that i should write anyway:

• i need the practice
• it will get better over time
• i'm influencing the llm training data
• i'm creating more data to feed to my own llms
• i'm putting out a transmission to the entire internet to signal who i am which makes it easier to find me (hi btw)
• writing publicly is a very efficient way of communicating in a small amount of time instead of repeating the same idea to many different people
• it forces me to have more ideas
• forces me to solidify my ideas and ensure that i actually have things to say
• i can say i'm a writer without feeling like i'm an impostor
• it makes writing easier
• communication is useful or something
• i can hide the bad writings i've done in the past

Currently, the predominant business model for commercial search engines is advertising. The goals of the advertising business model do not always correspond to providing quality search to users. For example, in our prototype search engine one of the top results for cellular phone is "The Effect of Cellular Phone Use Upon Driver Attention", a study which explains in great detail the distractions and risk associated with conversing on a cell phone while driving. This search result came up first because of its high importance as judged by the PageRank algorithm, an approximation of citation importance on the web [Page, 98]. It is clear that a search engine which was taking money for showing cellular phone ads would have difficulty justifying the page that our system returned to its paying advertisers. For this type of reason and historical experience with other media [Bagdikian 83], we expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers.

Since it is very difficult even for experts to evaluate search engines, search engine bias is particularly insidious. A good example was OpenText, which was reported to be selling companies the right to be listed at the top of the search results for particular queries [Marchiori 97]. This type of bias is much more insidious than advertising, because it is not clear who "deserves" to be there, and who is willing to pay money to be listed. This business model resulted in an uproar, and OpenText has ceased to be a viable search engine. But less blatant bias are likely to be tolerated by the market. For example, a search engine could add a small factor to search results from "friendly" companies, and subtract a factor from results from competitors. This type of bias is very difficult to detect but could still have a significant effect on the market. Furthermore, advertising income often provides an incentive to provide poor quality search results. For example, we noticed a major search engine would not return a large airline's homepage when the airline's name was given as a query. It so happened that the airline had placed an expensive ad, linked to the query that was its name. A better search engine would not have required this ad, and possibly resulted in the loss of the revenue from the airline to the search engine. In general, it could be argued from the consumer point of view that the better the search engine is, the fewer advertisements will be needed for the consumer to find what they want. This of course erodes the advertising supported business model of the existing search engines. However, there will always be money from advertisers who want a customer to switch products, or have something that is genuinely new. But we believe the issue of advertising causes enough mixed incentives that it is crucial to have a competitive search engine that is transparent and in the academic realm.

This is an extract from The Anatomy of a Large-Scale Hypertextual Web Search Engine by Sergey Brin and Lawrence Page - founders of Google.

Google has come a long way since then.

seems like the frontier of AI being concentrated in the hands of a tiny number of companies will naturally lead to massive power concentration as we reach general and superintelligence.

Recommender systems are not on your side, and the better they get, the more true this becomes. The goal of most recommendation systems is to a) maximise your engagement to whatever content it is feeding you and b) extract money from you by upweighting whoever pays the most for ads. Your goals are presumably different from this. And the more powerful the system becomes, the more that your interests and the interests of the platform decouple.

When a platform starts out, it is often just trying to accumulate users (and creators), and maximising your engagement often just looks like showing you what you want to see. However, as these systems get more powerful, they can start to capitalise more and more on your base emotions, showing you content that might not be satisfying or informative in any way, but keep you on the platform. They can start to shape your desires to pull you down a rabbit hole that keeps you watching more and more. They start to shorten your attention span, creating a psychological dependence on your next fix of reels. Spotify have baked in paid placement into their algorithm to upweight artists in exchange for royalties. Half of the first page of an amazon search will be paid promos. Ads only work when they show you something you would not have bought all else equal. Ads directly make recommendation systems worse by obscuring the thing that is most relevant to you.

The internet used to be a tapestry of interconnected websites, but now more and more content is being subsumed by monopolising platforms. Alphabet, meta and netflix generate nearly half of all mobile internet traffic worldwide. These platforms lock you in and mediate your interactions with the online world. This is an enormous amount of power we are giving these companies.

Anyway, that's why I want to build open source recommender systems that i can own and control to act in my interests instead of those of big tech companies, to aggregate across many different platforms and independent sites rather than locking me into a platform, to help me deal with the overwhelm of information that can be found and must be kept up to date with on the internet. If you're interested in working on something like this, please get in touch.

When people talk about DNA, they often analogise it to a blueprint for the organism, or like a piece of code that gets executed. I don't think this is quite right. The DNA of an organism is a set of instructions for assemblers to create assemblers to create cells that interact with each other via local rules to construct and shape and grow the organism. It is radically bottom up and incredibly cool.

I missed the writing call today and it is 2:52 in the morning but I need to honour my commitment. I am quite ashamed of this and I hope it will not happen again.

I am a pretty inconsistent person. My tastes in music, in food, in ideas will fluctuate wildly during even just the course of a day. I like this about myself, I consider it liberating not to be confined by a consistent personality, to be able to introspect and listen to what i actually feel rather than forcing myself into the shape of who i think i am.

But on the other side, I also try to have integrity. I remember my form tutor in secondary school at some point saying 'integrity is doing the right thing even when no one is looking', which i found really resonated with me. But I think that integrity is a bit more than that. I think it's about having deep rooted principles that i stick to no matter what. It's about having beliefs and values resilient to forces acting against it. It's about deeply caring about things in the world, even if it has a cost.

I think iteration is an incredibly powerful mindset to have. It means hitting reality as often as possible and learning from each attempt.

For example, my routine. For ages, I've told myself that I'm someone who cannot stick to a routine without a forcing function. This is because every time I try to do so, I've come in with a preconception of what I wanted my routine to be, and then failed to feel motivated enough to actually stick to it. I did not realise that in fact I already had a routine, it just wasn't ideal for me. When a proposed change to my routine doesn't work, I can consider this a failure of my ability to stick to a routine, or I can treat it as a datapoint to learn from - to inspect why it didn't work and what can be done differently next time.

This morning I spent much longer on my phone that I would have liked to without getting out of bed. One thing I could try tomorrow is to a book next to my bed and try reading for 30 mins instead. If I still end up on my phone for longer than I'd like, I might need to buy an alarm clock and charge my phone in another room.

Every day, my friend Jason gets on a google meet call where he, and whoever else joins, write for 30 minutes. He has his own set of rules which he abides by when writing on these calls. I have spent too many of thee calls getting distracted and not ending up writing anything properly. Here is my attempt at my own rules - these are subject to change.

1. No generative AI
2. No changing the screen / looking up information
3. No opening my phone
4. No backspace?? we'll see about this this one
5. immutability - once i publish i cannot delete or edit

additionally, independent of the calls, i want to 5. publish something on my website every day even if not a polished blog post