X posts historic archive

https://appblit.com/scribe?v=zzXyPGEtseI
12:28 PM · Apr 24, 2025

https://x.com/ljupc0/status/1915359549495693425
Ljubomir Josifovski @ljupc0
You have researchers, less and worse ones maybe lately but researchers never the less, you have access to Internet, and lately AI Research tools. To air as if Tabula Rasa empty vessel, filled with anything your guest dreams up, and simply re-broadcast - well then: it's the British Telecom we are talking about, not British Broadcasting! Broadcasting among other things is to educate and inform!
11:57 AM · Apr 24, 2025

https://x.com/ljupc0/status/1915326424111579247
Ljubomir Josifovski @ljupc0
But more companies -> more supply and more competition -> reduced prices, no? And isn't one of the features of any taxation "we will get less of that of what we tax"? (all other things being equal) Agreed it has but become impossible, in n hugely complicated system that energy one has been made in the UK over the past decades, to judge with much certainty what outcome we get, for any lever that is pulled.
9:46 AM · Apr 24, 2025

https://x.com/ljupc0/status/1915322751708959192
Ljubomir Josifovski @ljupc0
Maybe tad on the tail end of that, but a lot feels like the usual business of governing a country. We just get to see how the sausage is made, even if not intentionally! 😂 It comes naturally to the current team-Orange-clown. 🤡 When running a circus, the spectators are a big bigly uuge part of the Big Tent. 🤪
Fwiw—I'm all for it! For the lolz, but also b/c I think more state transparency is needed. Insufficient transparency makes any feedback hard, and without feedback it's hard to improve anything.
I advocate for radical transparency on part of the state towards us citizens, and businesses towards us their consumers. In order to address and redress the current information imbalance. Information leads to knowledge leads to power.
The conventional opinion is that the power imbalance coming from the information imbalance (state/business know a lot about me; I know little about them) is that us citizens and consumers should reduce our "information surface" towards them. And address the imbalance that way. But.
There exists another, often unsaid option. And that option is for state/business to open up, to increase their "information surface" towards us. That will also achieve information (and one hopes power) balance. Every time it's actually measured, how much value we put on our privacy, when we have to weight privacy against convenience and other gains from more data sharing, the revealed preference is close to zero. Revealed preference is that we put the value of our privacy close to zero, despite forever saying otherwise.
So the option of state/business revealing more data to us citizens/consumers, is actually more realistic. Yes there is extra work on part of state/business to open their data to us. But it's worth it. The more advanced the society, the more synchronisation it needs to achieve the right cooperation-competition balance of the interaction between ever greater number of people.
There is an old book "Data For the People" by an early AI pioneer and Amazon CTO @aweigend. Afaics it well describes the world we live in, and also are likely to live even more in the future. Dear reader, ☺️ thank you for reading thus far, hope this makes sense to you.
RADICAL TRANSPARENCY FTW
9:31 AM · Apr 24, 2025

https://x.com/ljupc0/status/1915319026516988095
Ljubomir Josifovski @ljupc0
Given our bills have gone only one way - up! - for 10-20 yrs now, your credibility with the public will be ever decreasing in anything you say. The only way to make me take stock of your proposals, would be to make personal $$$ bets and put your own $$$ where your words are.
9:16 AM · Apr 24, 2025

https://x.com/ljupc0/status/1915307180250501535
Ljubomir Josifovski @ljupc0
They are not the enemy—but their incompetence is making lives of millions miserly. DOGE is a reaction to a huge FAIL of the state capacity. There is no point in preserving of much that DOGE is cutting. That has already died, has been dead for some time. DOGE are like I cutting dead branches in the garden.
State capacity sure has a role! A role to organise and marshal resources on gargantuan scales. To take gargantuan risks! That no other organisation in society can afford to take on, and live to see the end of those risks, the ultimate payoffs. (or survive the fails.) For that, it is the State that is the ultimate resource.
There are no rich countries with small states. There are no peoples that to better their lives, don't organise as a State first and foremost. I come from a startup Nation, and a startup State. And a startup literary language too! All these things go together. All are technologies where large groups of people come together to better their lives.
And no one should begrudge them that after spending decades, even centuries in the wilderness, if they find a path through their valleys of tears, to their promised lands. As long as all that is towards building lands of milk and honey. (or even "life, liberty and pursuit of happiness") And not murdering other peoples like yours, next to you.
(for something to exists, as a separate entity from it's environments, needs a separator, a separation, and that is provided by force, by violence; State monopoly on violence, external and internal, is the technology humanity has discovered over time)
Yes, "Your Country is not a Business" is a good reading now as ever. Musk will fail not because his instincts are wrong—but because building a State requires institutions, not cults. He knows of using the strength and cult of personality, and that is all fine and good. But that suffices and works in small high-performant self-motivated narrow-aim organisations. That will not work for State capacity. Expect Musk to be out sooner rather than latter, and what little dent was made towards change, will be undone within weeks.
Yours "more State always and forever" has FAIL mode ofc. Argentina was that. There what Milei does is obviously going to change things for the better for time being. I grew up in a late-socialism country, that on paper you being similarly ideological would have liked. But in real life would have hated it. For nothing worked, for nothing could be done, for the simple reason that everything was forbidden. A country decided to freeze any efforts of its citizens towards betterment of their own lives.
Subsequently when a reformer unfroze it, just let the natural energies and talents of the citizenry flow, life improved magnitudes and in very short time. Seemingly without special knowledge and super-human effort. Just millions of tiny ants, doing millions of tiny tasks, that need doing for ages. Previously the ants were forbidden to do them. And now the ants were free to do them on their own. That was it.
More humility is in order considering the scales and the consequences of State failures. Especially on ideas coming from intellectuals, which of course you are in part. Do you never read e.g. Thomas Sowell on intellectuals, and think "hm, there is empirical evidence that he may have a point?" Did the fact that big state socialism crash and burn everywhere in your own lifetime somehow got erased from your memory? Or "Big Green Tech" idea both hitting problems with the tech, and not delivering the jobs. I'm baffled by the ego at work here tbh.
8:29 AM · Apr 24, 2025

https://x.com/ljupc0/status/1915171673772355908
Ljubomir Josifovski @ljupc0
I swear to God you Americans are sometimes like Donald Duck’s three nephews when it comes to the evil means of the big bad world. 😂 All other countries basically all the rest of the world is less free—so at other times I cut you some slack.
Reminded me of this .@michaelmalice interview YT algo serendipitly discovered for me about the Soviet Union
https://youtu.be/3cVr2Qp_ic8?si=ewiW7HiKGnA805iG
Even a well prepared enthusiastic interviewer kept missing the last step or two of some ingeniously devised totally evil torture scheme. MM had to interject more than once to explain the coup de grâce. ("ah, but you see, not only was immense pain inflicted this way, but additional suffering followed-up that way")
11:31 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915163400746700916
Ljubomir Josifovski @ljupc0
Ah! Glad to read the freak show is not over!
Earlier today read some fakenewsers and thought it over earlier than expected.
https://x.com/ljupc0/status/1914590038777643230
Quote
Ljubomir Josifovski @ljupc0 · Apr 22
Whatt?? Nooo—too soon!! 😭
What happened to the "entertainment guaranteed, even if success in doubt" musk of yesteryear?? Expected more entertainment! 😢 I get it, I expected it, just—was hoping for a longer run of the show tbh.
Moar entertainment for us plebs rn Muskarat!! Don't x.com/Independent/st…
Show more
10:58 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915161944220148045
Ljubomir Josifovski @ljupc0
Normies hating on you—what else? (alt: re-read CPSnow Two Cultures?)
Humans we care not about true/false. All goes through "friend or foe", "likes me/not" etc filters. Our evil monke survival analytical engines whirring in the background, never ever stopping.
Other humans sense you are not quite "one of us". Even if they can't clearly articulate what/how/why. Their pattern matching engines ID-s you by feelings and vibes.
Like the T-800—relentlessly pattern matching, never stopping.
10:52 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915159493463466189
Ljubomir Josifovski @ljupc0
+1 on Mute and Block, makes X life much better. Now with Block being one-way read-only, don't bother with Mute anymore.
Checked TokTok now. While it's all fun all that brain rot, 😂 it's all passive consume only. Doubt it's for me. Maybe if they do TextTalk Twitter clone.
10:42 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915150262014034117
Ljubomir Josifovski @ljupc0
Oh man—preach more! Music to my ears, balm to my soul. 😇
The number of times I've had to byte my tongue when an English friend opines on ID-cards! I'm Johnny foreigner now naturalised Brit for some decades. And IK enough that it will be only counter-productive if I get too opinionated and wordy about their - frankly - naive musings.
What you wrote above is immensely important. I think many in the UK feel it at some level, but few can articulate clearly why and how. Top marks 👍 for writing the above! 👏
From time to time I post anti-national-ID scribblings, latest-
https://x.com/ljupc0/status/1909917211902333192
...but more linked. I get the impression normies mostly get embarrassed (why??) and just shutdown their thinking faculties.
With Blair still with us and as enthusiastic about them as ever (The world of TB: no problem exists that a national-ID card does not solve), this flares from time to time so I pipe up. The strongest hope I got tbh is that public inertia and gov incompetence will save us from making the mistake of introducing national ID cards. Rather than some superior public discussion and decision making process.
10:06 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915145282641342485
Ljubomir Josifovski @ljupc0
I wanted to try Gemini for its longer context.
Architect mode - for that I use aider, it has separate architect and writer roles. For DeepSeek it's
aider-deepseek-best() { local -; set -x; aider --set-env AIDER_START="$(date)" --architect --model deepseek/deepseek-reasoner --editor-model deepseek/deepseek-chat; }
I believe atm reasoner points R1, and chat points V3. One can select the versions explicitly too. But then e.g. once R2 is released, I presume reasoner will be re-pointed to R2, while if I select by explicit name, I have to change the scripts.
You can see (performance,price) and setups here
https://aider.chat/docs/leaderboards/
So OpenAI top in performance and price. Then for a bit less performance, Gemini is 1/10-th the price. Then a little less performance again , and one gets DeepSeek for 1/2 or 1/3 of Gemini price.
I try them all from time to time, esp on a new module or project. I have not got a good reason to keep trying. I just love this stuff. Brings a smile to my face, ngl. 😇
9:46 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915141383272251497
Ljubomir Josifovski @ljupc0
Ah - and I thought I had it bad with me writing into the void. 😂 After reading of your predicament - I think myself lucky! No replies better than aggravating braindead replies, fs! ngl 😇
I heard (on a podcast) off-remark that TikTok is the 1st (and so far only) Social Media platform that has the algofeed driven (to non-trivial %age) based on the data/content, and *not* solely on the metadata. The implication was that is the reason their algofeed is prenaturally good. Users often feel it almost super-natural in predicting what the user wants. Not used it—but heading off to register after I post this.
1st Twitter now X algofeed has been cr*p to me, always. Not claiming it much worse now. But wd have expected it to get better in 10 yrs?? But no—X algofeed is as sh*t as always. Have to battle it left and right to manually find what I want. I rarely see anything you post.
IDK fs but imagine the Twitter format (X and Bsky) might be ripe for disruption. By now I think it's clear using largely meta-data to predict "will user Y want to see post Z, yes/no", simply leads to positive feedback loop of a winner-takes-all type. Small number of creators-writers on one side, many in number readers-consumers on the other. The old v1 Social Media, harking back to the Mass Media MSM setup predating it.
That's not all bad news for the platforms. (they think) They don't mind creating differentiating 1K-10K accounts as tier-1 ("content creators"), as that they can then influence whichever way. Much harder to steer 100M accounts in the direction desired by the platform owners. Incentives incentives.
I'm personally tired of it, b/c SNR is low. It's a waste of everyone's time, incl Zucks and Musks, and for zero marginal gain. TikTok showed that. Maybe we get lucky and ByteDance create TextTalk, a Twitter format social media platform. 😂
9:31 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915132189940871328
Ljubomir Josifovski @ljupc0
The proof of the pudding is in the eating. Bring down electricity prices for consumers and businesses. Your lot are going to be chucked out of gov by PM Farage leading the REFCON block. Everything you built will crash and be crushed. Looking at the polls current direction of travel.
You've got 2-3yrs to get the prices down. Not down by 10% or sth inconsequtial, but to at least halve the retail electricity price. What thick thick skulls you greenies got. God only knows how Suneil deals with you dunderheaded cretins.
8:54 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915124419195789782
Ljubomir Josifovski @ljupc0
Oh dearie oh dearie.
If they were capable of such feat, they would have done it already. 10 times over
Heck. They would not be in the position they find themselves rn, if they were capable of that.
Not unlike the many predicaments that did it for the Dems previously
8:23 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915111004783943894
Ljubomir Josifovski @ljupc0
OpenAI and Google from that list, and DeepSeek too. Plus toying with smaller open weights models (qwq, qwen, gemma) locally.
7:30 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915110359842591143
Ljubomir Josifovski @ljupc0
Switching from DeepSeek (both architect reasoner and code writer) to Geminis caused total rewrite of a 300 lines python script. 😂 (trainer fitter) Not one line survived. And buts of the architecture like type of locking and lock mgmt somewhat changed too! 😂 much fun ngl
7:27 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915105335460643260
Ljubomir Josifovski @ljupc0
governing should be the first job to be replaced by AGI ngl
7:07 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915093235279757810
Ljubomir Josifovski @ljupc0
Double what I expected. I believed "the lizardman's constant" was about 4%-5% until now. Big news ngl 😂
6:19 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915072879252717936
Ljubomir Josifovski @ljupc0
IDK. But given UK gov refusal to change the scheme, I suspect UK gov has some advice somewhere to prevent change. Something like "risk of blackouts, expected XX dead". I don't see how removing the marginal pricing impacts any unreliables. Afaik unreliables prices are almost all fixed in advance contracts.
Would dearly like to know. So from time to time I spam whoever is in the vicinity RADICAL TRANSPARENCY FTW spam 😂
https://x.com/ljupc0/status/1914988714918400043
Energy security is my favourite guess—but also only one, got not better one.
I notice that every winter, we nudge the margin of slack ever lower. One of these winters when the stars align bad (=enough faults happen at the same time, by random chance) we will get to catastrophic nation-wide blackout. Of the type when UK has to resort to the Dinorwig to bootstrap its National Grid back online from a total power-off. With people stranded on trains overnight, dozens in ICU dead with machines cutting off after generators run out of diesel some hours into the blackout.
4:58 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915061682885300255
Ljubomir Josifovski @ljupc0
I have the X previously Twitter notification enabled at the user account level. But phone, app, browser etc notifications are disabled of course. When I check X either via Web browser or App, I see the blue badge with number of notifications on the bell 🔔 icon on the interface. (bottom of the page in the Web browser) If I feel like it, I check to see what they are about. (if the number looks high.) If I don't feel like it, or I see small numbers 2-3 on the badge - then I don't bother checking. They will be in the Notifications web page next time I check my feed too. No rush, nothing on X is urgent tbh.
4:14 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915022833295544630
Ljubomir Josifovski @ljupc0
Many have noticed that we are ruled more by algorithms, than humans like the plebs we are, already. 😂 SirKS feels to me an algorithm in action, embodied in a human body. And I don't begrudge SirKS anything - he's done well, his algo-politico algorithm worked well! SirKS occupies one of the top offices in the land.
As one takes stock of things on St Georges day, looks to the past, then turns the gaze to the future, we try to diving things. For all its excellent qualities my adopted country possesses - and these are many!! ngl-UK is top top place to be in, and long may it continue and prosper - can't help but notice certain problems with the political system. The people we select then elect for high office, have come short of expectations in the recent years.
AI ChatGPT to the rescue then!
Am not entirely unserious there. Would AI be more truthful than SirKS? Almost certainly. Would AI do more work? Sure - just deploy to 10000 servers in data centers across this green and pleasant land, and let them run!
AI should successfully replicate what we got now. We lose nothing! The AI-(SirKS) prompt (objective function) might be: "Strive for zero or at least minimal risk. You desire to do nothing, you will not change anything. Risk is the only axis left to you to optimise on-strive for zero risk at all times". If things fall short of expectation, we can always go back to the old/current "algopolitico in a human body" system.
AI advantage is that it can be frozen in time, interrogated, and made as transparent as we wish. On this issue, AI could tell us: "I don't care much wither way. I have not got a strong opinion on this. As is my default stance in cases like this, I would like to make the least number of people the least unhappy, with any decision I make. I will make a decision only if forced to make one by events thrust onto me." Sounds much better to me, than the usual hare-brained word salad we have come to expect from a bog standard human SirKS and clones.
(RADICAL TRANSPARENCY of UK gov towards us UK citizens is another campaign dear to my heart.)
I get it people worry about "AI existential risk". Yeah there is some-but there is plenty of risk now, existential too, with humans! Who do we think will sooner start thermo-nuclear war: Trump-Putin, or ChatGPT??
Quote
Politics UK @PolitlcsUK · Apr 22
🚨🎥 WATCH: Keir Starmer says "a woman is an adult female" following the Supreme Court ruling last week
Rate proposed Community Notes
1:39 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915017833945145643
Ljubomir Josifovski @ljupc0
a software update applied, is an improvement on no software update tbh 😂 but yeah as appropriate t osay for StG day "not ideal"
1:20 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915007546437386582
Ljubomir Josifovski @ljupc0
AI ChatGPT to the rescue!
Am not entirely unserious there. Would AI be more truthful than SirKS? Almost certainly. Would AI do more work? Sure-just deploy to 10000 servers, and let it run.
AI should successfully replicate what we got now! So we lose nothing. The AI-(SirKS) prompt (objective function): "Strive for zero or at least minimal risk. You desire to do nothing, you will not change anything. Risk is the only axis left to you to optimise on-strive for zero risk at all times".
AI advantage is that it can be frozen in time, interrogated, and made as transparent as we wish. On this issue, AI could tell us: "I don't care much wither way. I have not got a strong opinion on this. As is my default stance in cases like this, I would like to make the least number of people the least unhappy, with any decision I make. I will make a decision only if forced to make one by events thrust onto me." Sounds much better to me, than the usual hare-brained word salad we have come to expect from a bog standard human SirKS and his clones.
I get it people worry about "AI existential risk". Yeah there is some-but there is plenty of risk now, existential too, with humans! Who do we think will sooner start thermo-nuclear war: Trump-Putin, or ChatGPT??
12:39 PM · Apr 23, 2025

https://x.com/ljupc0/status/1915002570390503829
Ljubomir Josifovski @ljupc0
It's futile to complain as is per the incentives.
Incentives that come from the state of the algofeed technology I imagine. Where I suspect it's 100% meta-data driven, and 0% data-content driven.
I heard (on a podcast) off-remark that TikTok is the 1st (and so far only) Social Media platform that has the algofeed driven (to non-trivial %age) based on the data/content, and *not* solely on the meta-data. The implication was that is the reason their algofeed is so so good. Almost super-natural in predicting what the user wants.
IDK but imagine the Twitter format (X and Bsky) is ripe for disruption. The algofeed has been cr*p for me always. Not claiming it's much worse now. Wd have expected to get better in 10yrs? But no-X algofeed is sh*t for me, as always. Have to battle it left and right to find what I want. Most of the time ton of manual work.
12:19 PM · Apr 23, 2025

https://x.com/ljupc0/status/1914988714918400043
Ljubomir Josifovski @ljupc0
Good—great! Would you consider advocating for RADICAL TRANSPARENCY of UK gov towards the Great British public?
I advocate for radical transparency on part of the state towards us citizens, and businesses towards us their consumers. In order to address and redress the current information imbalance. Information leads to knowledge leads to power.
The conventional opinion is that the power imbalance coming from the information imbalance (state/business know a lot about me; I know little about them) is that us citizens and consumers should reduce our "information surface" towards them. And address the imbalance that way. But.
There exists another, often unsaid option! And that option is for state/business to open up, to increase their "information surface" towards us. That will also achieve information (and we hope power) balance! Every time it's actually measured, how much value ourselves put on our privacy, when we have to weight our privacy, against convenience and other gains from more data sharing, the revealed preference is close to zero. Revealed preference is that we put the value of our privacy close to zero, despite saying otherwise.
So the option of state/business revealing more data to us citizens/consumers, is actually more realistic. Yes there is extra work on part of state/business to open their data to us. But it's worth it! The more advanced the society, the more synchronisation it needs to achieve the right cooperation-competition balance of the interactiona between ever greater number of people!
There is an old book "Data For the People" by an early AI pioneer and Amazon CTO @aweigend. Afaics it well describes the world we live in, and also are likely to live even more in the future. Thank you for reading thus far, hope this makes sense to you.
RADICAL TRANSPARENCY FTW
11:24 AM · Apr 23, 2025

https://x.com/ljupc0/status/1914984480630047040
Ljubomir Josifovski @ljupc0
Kudos!—you are doing gods work. You could do more with the position you occupy.
Advocate for RADICAL TRANSPARENCY of UK gov towards the Great British public.
I advocate for radical transparency on part of the state towards us citizens, and businesses towards us their consumers. In order to address and redress the current information imbalance. Information leads to knowledge leads to power.
The conventional opinion is that the power imbalance coming from the information imbalance (state/business know a lot about me; I know little about them) is that us citizens and consumers should reduce our "information surface" towards them. And address the imbalance that way. But.
There exists other, often unsaid option! And that is for state/business to open up, to increase their "information surface" towards us. That will also achieve balance! Every time it's actually measured, how much value ourselves put on our privacy, when we have to weight orivacy against convenience and other gains from more data sharing, the revealed preference is close to 0. Revealed preference is that we put the value of our privacy close to zero, despite saying otherwise.
So the option of state/business revealing more data to us citizens/consumers, is actually more realistic. Yes there is extra work on part of state/business to open their data to us. But it's worth it! The more advanced the society, the more synchronisation it needs to achieve the right cooperation-competition interaction, and between ever greater number of people.
There is an old book "Data For the People" by an early AI pioneer and Amazon CTO @aweigend. Afaics it well describes the world we live in, and also are likely to live even more in the future. Thank you for reading thus far, hope this makes sense to you.
RADICAL TRANSPARENCY FTW
11:07 AM · Apr 23, 2025

https://x.com/ljupc0/status/1914979142702899338
Ljubomir Josifovski @ljupc0
Go further please. Advocate for RADICAL TRANSPARENCY of UK gov towards the Great British public.
I advocate for radical transparency on part of the state towards us citizens, and businesses towards us their consumers. In order to address and redress the current information imbalance. Information leads to knowledge leads to power.
The conventional opinion is that the power imbalance coming from the information imbalance (state/business know a lot about me; I know little about them) is that us citizens and consumers should reduce our "information surface" towards them. And address the imbalance that way. But. The other, often unsaid option is, for state/business to open up, to increase their "information surface" towards me. That will also achieve balance!
Every time it's actually measured, how much value people put on their privacy, when they have to weigh against convenience and other gains from more data sharing, the revealed preference is close to 0. So the option of state/business revealing more data to us citizens/consumers, is actually more realistic! Yes there is extra work on part of state/business to open their data to us. But it's worth it! The more advanced the society, the more synchronisation it needs to achieve the right cooperation/competition interaction, between ever greater number of people.
There is an old book "Data For the People" by an early AI pioneer and Amazon CTO @aweigend. Afaics it well describes the world we live in, and also are likely to live even more in the future.
RADICAL TRANSPARENCY FTW
10:46 AM · Apr 23, 2025

https://x.com/ljupc0/status/1914364573987045770
Ljubomir Josifovski @ljupc0
Afaik that's not entirely possible. I have the following mental model of SocMed algofeeds. I can expect any one post I write to be shown to max 1/10-th of my followers, in the best case. How and why? The simplest possible example I can think of is this.
I follow 5000 accounts. Each one writes 1 post a day = 5000 posts per day. I login to X, X algofeed serves me 50 posts on one screen full. I check X 10 times per day = 500 posts that X will show me in one day. That is 500 out of possible 5000 to be shown. And so there will be 4500 posts, from people I follow on any one day, to **not** be shown.
X must decide which 500, out of 5000 possible, to show me. So any one post has probability 0.1 to be shown. I will see 1 post from 1 account once in 10 days. IRL the algofeed will be vastly more complicated. The algofeed is the core of the social media company.
Unlike the sketch above, the algofeed is non-random. It mostly uses metadata to tilt that 0.1 probability. Algofeed tilts towards factors like timeliness/age "how much time elapsed since a post got posted", engagement measures like how many in number {Like, Forward, Quote, Reply} interactions it got, probably all decayed by time with some half-life.
Afaik no actual data about the quality, the content, of the post is ever used. Only metadata. This is pure speculation on my part. I have no insider knowledge. I have not worked in a social media company ever. Just thinking aloud and considering rules of thumb and first principles.
For me personally posting. Usually I only have the "View"-s for a post to see as a feedback. I don't even know if it was shown to the user whose message I reply to! (@perrymetzger in this instance) It's entirely up to X algofeed, if/when it shows it to the OP, or not at all. If a post gets zero engagement (a "Like" at a minimum) after 30-50 Views, it's subsequently dropped. (never pushed any more to anyone's screen.)
One would need to go to my home page on X, then search for a post intentionally. E.g. like so
https://x.com/search?q=(from%3Aljupc0)&src=typed_query&f=live
It's possible that my idea of how things work is entirely faulty. Allow for that pls. I'm no social media star-can hardly give advice to anyone! 😆
6:04 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914360619668549658
Ljubomir Josifovski @ljupc0
No one cares. Stopped listening to any JRE many moons ago. There are plenty better out there. It's easy. We live in un-imaginable data abundance times. ☺️
Joe became too big for the kind of low level work that matters. Seems he thinks it's beneath him. Fine. I guess that happens with many Millions of listeners lending you their ears. No need to kill oneself with effort.
For myself-there's plenty of other fish in the sea, YT podcasters to see. YT channels are full of UKR-RUS content. More from Ukraine, less from Russia - as they seem to be somewhat more keen on total control - but still doable.
And so and plenty podcasters do it, do good work roaming about, showing and talking about what they experience. Lend them your attention. Joe has had his cup full already. Needs not more of the same.
5:48 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914358764142313759
Ljubomir Josifovski @ljupc0
Worse-you only hear from frustrated people that tried and failed to use it. The ones for whom "it just works", keep shtum, as they have no reasons to complain. Once it works for the previously disgruntled-they stop posting their frustrations too.
We typically come online when something is ailing us, and voice our gripes. I don't post on X: another day of vim doing as per my :commands, Chrome did it's thing no problems, git --obeyed orders too, and bash worked well {today} as it did in the past 30yrs.
5:41 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914350979400540455
Ljubomir Josifovski @ljupc0
Indeed - that is true, if it was a propaganda tour.
Why would it be though? Joe could just go, and roam, and record and comment what he sees, to his heart content. In person or incognito hidden.
Plenty of people do, the YT channels are full of content. More from Ukraine, less from Russia - as they seem to be somewhat more keen on total control - but still possible. Plenty good podcasters doing good work.
Seems Joe became too big for that kind of low level work. Seems he thinks it's beneath him. Fine. Plenty of other fish in the sea, YT podcasters to see.
5:10 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914347939473809541
Ljubomir Josifovski @ljupc0
Thanks. I'm all for open data open everything, esp for gov RADICAL TRANSPARENCY every gov business open to the public as default and closed by request.
I write software that does quadratic optimisation setups for a living and run Mosek frequently. Glad to read the guy is talking about optimisation.
Having said that-the proof of the pudding is in the eating. Overcoming the evidence of prior 20 yrs of policies resulting in UK ending with the highest electricity prices among the peers is no small feat. All the while increasing the fraction of unreliables in the mix.
A quarter of UK chemical industry disappearing, possibly as a result of high energy prices in the UK, affects me personally too. So-I'm ill-inclined to give the proponents of the current system much of my time. They had their go at things-the results are what we pay today.
4:58 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914344034081898627
Ljubomir Josifovski @ljupc0
Indeed, you are 100% right. Autistic people are in greater danger than most to seek clever schemes that "solve everything" in one neat formula, leaving no lose ends. My wife works with kids like them. From the stories anecdotes relayed from time to time-I see how and why this wd be the case. While there have been plenty of evil humanist-by-education dudes in history, I always have in mind that one of the greatest recent monsters Beria "excelled in math and science".
This tongue-in-cheek "VDT: a solution to decision theory" https://lesswrong.com/posts/LcjuH on a related topic (HT https://x.com/ChrisChipMonk/status/191323296) maybe of interest. My reading was 1) Less frivolous than it first appears; 2) Not obviously worse than the alternatives; 3) Well even-problem solved again, with prior art being "Vibes==Love" in our (christian?) culture afaics.
[[[ Meta. Don't apologise please. I like to read what you write, I follow you and you follow me, yet frequently X does not show me what you wrote. The above post is great. Twitter now X algofeed was always cr*p, and is still cra*p. Please re-post more of your old posts! Fight low quality low effort algofeed with low effort means to maintain. Maintaining good SNR justifies it!
My unfounded suspicion is X algofeed is sh*t in a way that it places almost all the probability weight on the "timeliness" predictor, and almost nothing on anything else. And esp nothing on content i.e. data - it's all about meta-data. Afaik all X algofeed predictors to "should I show post Y to user Z" are meta-data only, and none are data actual content.
I heard on a podcast - that I can't find, may have been one done by the two Abundance book guys promoting the book, but don't recall which - and one of them (or the host?) mentioned this. That TikTok is the 1st social media algofeed that doesn't (unfairly) prioritise meta-data. But looks at the actual data, at the actual content. Of both what is posted, and what the user consumes i.e. wants. And that it actually matches that. And that's the reason people find TikTok almost super-natural in guessing what they want.
For utilising meta-data simply leads to positive feedback loop of winner-takes-all. Which is not all bad news for the platforms. They don't mind creating differentiating 1K-10K accounts as "content creators", that they can then influence whichever way. Much harder to steer 100M accounts in the direction desired by the platform owners.
If the above about the TikTok algofeed were true. I'd get the urge to bash our SocMed overlords with a heavy object over their heads. For wasting Bazillion of hours of human lives, to feed us slop, and for their own detriment, and ours. The very definition of human stupidity. Unexcusable on the part of the Zucks and Musks. /rant ]]]
4:42 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914335893537632380
Ljubomir Josifovski @ljupc0
I don't reject it as impossible, maybe. But he-says-she-says is weak evidence for me. Need to see the data that is your evidence.
When all is added and subtracted, I want to see the price I myself pay for energy halve. In the past 10 years, I have seen that price about double. This is a very troubling and disappointing direction of travel. I blame the policies that have been in place in the past 20 years, for that disappointing outcome.
I have rooftop solar plus battery, I'm electrical engineer by training, and know enough about the workings of the system to form judgements.
4:10 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914333084545073619
Ljubomir Josifovski
@ljupc0
Hm - I have been luckier than that so far I must say. What's your setup?
I'm using aider atm, current best-but-expensive is
aider-openai-best() { local -; set -x; env AIDER_START="$(date)" aider --architect --model openai/o3 --editor-model openai/gpt-4.1 --openai-api-key ${OPENAI_API_KEY} ; }
All the while trying all the other tools (Cursor, Windsurf, addons Cline and Roo for VSCode) from time to time.
Last week it was Gemini and DeepSeek that I used. They wrote me 95% code of a boring module in a language I can read but not write super comfortably yet. I could have done it manually, bit this was more fun. My current best process is 1) write into INSTRUCTIONS-dot-md (that unlike in-chat instructions can go in git as documentation too); then 2) tell aider "read instructions from file xxx, then do as instructed".
This was prob v7 of the same functionality I have done over the years in a coterie of languages. So not only did I know what needs doing and how, buy I also knew of the likely fails, and dangerous corners. Enough to be anticipating some and preemptively describing, as well as being on a lookout for likely logical errors. (LLM makes no syntax errors??) All in all - good experience, will repeat.
Anyways - just came to your TL to say I loved listening to your "Reinforcement Learning, by the Book" series, it is excellent. Thanks for making it and sharing it with the world.
3:59 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914315862057660755
Ljubomir Josifovski @ljupc0
Random observation. For all their faults, the UK elites running the British Empire were forever outward looking almost to a fault. Maybe it helps that the United Kingdom is in itself 3.5 nations too.
In contrast, the US elites educated in the best Unis of the land (Ivy Leagues? etc), seem forever inward looking, parochial, ill informed. It puzzles me to a lot. Americans I've met in the US have been nothing but open or even keen to talk about anything, and open minded. US-ians abroad I have interacted with, have also been open and curious, not closed to experience. (as long as one does not criticise outright their native lands; and that's fine, we are like that, myself included)
How is it then, that just the US-based elites commanding the heights of the country, are so painfully not-self-aware. I fail to understand.
2:50 PM · Apr 21, 2025

https://x.com/ljupc0/status/1914923165739872399
Ljubomir Josifovski @ljupc0
Thanks, but I will not Luca.
I spend time and expended effort to write the above. Used up some of my limited time on this Earth to inform and educate a perfect stranger, expecting nothing in return really. A stranger that I will never meet again, and will probably never have a chance to reciprocate.
Nothing personal this—but I will cut my loss at this point. For writing to someone who refuses to read, is something I want to repeat with this very message at most once, but not more than once. Sorry the interaction didn't work out. Not your fault, no one is at fault—but blocked now. So I'm not tempted to interact again with in the future. (these days block is a read-only one side, so not a biggie tbh)
7:03 AM · Apr 23, 2025

https://x.com/ljupc0/status/1914919687466455499
Ljubomir Josifovski @ljupc0
☺️ Possible.
This may tally with that - can't but notice that many Coke addicts are very very performant and achieve a lot in a short time. Some look manic energetic. Not good for one's health. But sure it does bring results—have seen it.
On the other hand, sticking with a very pedestrian bit rate of 39 bps of human language forever, strikes me as a huge bottleneck in human communication. Don't mind increasing that, even if overclocking. 😉
Audio podcasts at x1.2 speed are almost the norm now for many. Depending on the speaker, I go up to x1.4 speed.
So much stuff, so little time! 😅
6:50 AM · Apr 23, 2025

https://x.com/ljupc0/status/1914904173969068441
Ljubomir Josifovski @ljupc0
Of course - and expect nothing else in the future too.
In my lifetime, AIDS stopped being a deadly disease, and no one noticed.
I was ~20yo when I 1st heard about AIDS in the media. A positive test was a death sentence. Newspapers went on and on, with some celebrity dieing every week or month, for years. Innuendos, real and fake news, of "are they or aren't they" HIV positive, filled newspapers pages. For years.
Then I noticed at some point - wait, we are not reading these stories any more. No celebrity had died of AIDS within my memory window. I dug up info - there was Internet by then. Found out the illness is treated, and turned Chronic, away from Acute. People suffer but they are not dieing en masse. There's a documentary "In the time of a plague" or similar about the drugs development. Nowadays I presume it's curred completely, can't be bothered to check now.
Ulcers to the stomach also got cured in my lifetime. In my childhood, sometimes I'd overhear parents talk in grave voice "So-and-rushed to an operating theatre to hospital, but couldn't be saved. Burst ulcer bleeding, tsk tsk. Such tragedy - family kids lost father etc". Then in my adulthood I realised - I never hear of such tragedies anymore. Found out the saga of Helicobacter and the Australian doctor(s). That Ulcers are cured by antibiotics, and people don't typically die of burst ulcers anymore.
So yeah - a deadly plague killing people left and right will be solved, problem deleted, an impossible feat achieved. After decades of effort, of trials and failures, by thousands of people across the globe. Few people will notice.
5:48 AM · Apr 23, 2025

https://x.com/ljupc0/status/1914777870519689335
Ljubomir Josifovski @ljupc0
YT is destroying Hollywood as well as TV productions for me. Watching them is boring and tedious. Takes ages for them to get to the point. It's like after listening to podcasts to x1.5 speed, going back to x0.9 speed.
In the Mr Beast Productions Manual for new hires, that circulated online some time ago, Mr Beast writes about this phenomenon. That YT is higher intensity, higher SNR, than classical productions. And that's in his opinion will become bigger than all if them together are now.
9:26 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914749068749787508
Ljubomir Josifovski @ljupc0
Ofc. They are all wrong. And then at some point, become right. No one knows when, or even if, it will happen. However, we do know it will NOT happen, if we don't try. And we will not try, if we have no hope, that it might happen.
7:32 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914748068592771341
Ljubomir Josifovski @ljupc0
Somewhat popular opinion: don't blame others for your own failings.
If you are busy, decide to be less busy, and then become so. Usually no one is holding a gun to anyone's head, and forcing them "busy yourself, more, get more busy asap".
Less narcissist self-pity cr*p pls.
7:28 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914746476535652778
Ljubomir Josifovski @ljupc0
Would he be condemned in the UK public culture, ih he said so?
"I don't care much wither way. I have not got a strong opinion on this. As is my default stance in cases like this, I would like to make the least number of people the least unhappy, with any decision I make. I will make a decision only if forced to make one by events thrust onto me."
Would that be a terrible slight on him as a leader? Condemn him in the eyes of the public as not worthy of leading?
7:21 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914743273379475718
Ljubomir Josifovski @ljupc0
When one reads "How Innovation Works" or "Serendipity: Medical Discoveries in the 20th Century" one gets to appreciate how much luck plays a role. Many breakthroughs are in parts very much accidental.
We do disservice to young researchers there. When we write up a discovery, we usually start from the end, from the result, and mentally trace the path back to the start. Then we describe that one path we traced back, with words, from the start to the end.
But that single forward path is an illusion! At discovery time, at the start, and along the path, there were many junctions. Going forward, there is a tree of paths - never a single one. The ultimate path that worked, was one of many. It only exists when looking back, from the end result. Often it is not even clear why we decided to follow one fork, and not another. Usually there are arguments evidence for-, as well as against-, taking either path, in a fork on the road. (to the ultimate discovery.)
Further, timing is everything. Kurzeil - now famously - stumbled on his Flops/USD log-linear exponential trying to figure how to time his Inventions/contraptions. (his words-I think?) Hinton probably had Sutskever-level PhD students in year 2003, or even 1993 too. There were no GPU-s capable of a leap then probably. Even when there were GPU-s, it took an Alex, in addition to one Ilya, to be at the same place, same time.
(for comparison: in year 2000, a stellar guy in my lab training NN-s and writing ASR stack decoders, had a special Sparcstation bough with large chunk of project $$$, with then obscene unheard of I think 96 MB RAM; most users had computers with 2MB RAM from recollection)
Research is a mugs game. Some Hinton colleagues probably died never seeing the success of their field. Assuming all their efforts a failure, a waste good for nothing. After devoting good chunk of their lives to it.
7:09 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914730691020873927
Ljubomir Josifovski @ljupc0
Happened to me, python comments switched to Cyrillic. Presume RUS—but maybe other languages, dozens languages use Cyrillic variants. At other time it switched to French comments! Added in my INSTRUCTIONS md file "write your comments in English"—that fixed it, not repeated since.
6:19 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914722739539059103
Ljubomir Josifovski @ljupc0
🤩 np - forgot to add, not 1, not 2, but 3(!) companies making progress also a huge +ve: when something works, it works for all! now this is one news item that brightened my day, ngl 😇 thanks ☺️
5:47 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914711731781140635
Ljubomir Josifovski @ljupc0
Think @skdh nailed this one—this may work fr. Stellarators producing fields without needing insides current, continuous running, trading that for more complex magnets. Open sourcing too adds to confidence.
"Meet the Reactors Set to Upend Nuclear Fusion"
https://www.youtube.com/watch?v=e7hw0aC1BbI
Meet the Reactors Set to Upend Nuclear Fusion
5:03 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914704491162612195
Ljubomir Josifovski @ljupc0
Yes—the complaints are spot on. Glad someone is motivated enough cares enough to react. (I don't)
1) Selective booking.
Fine, bring on fringe guests if you must. But then at least include voices from both sides. Joe’s "crazies" are almost exclusively pro-Russia or pro-Hamas. And it's not just about balance—even non-crazy prominent figures like Vitali Klitschko (boxing champ, now mayor of Kyiv) literally sent Joe a video invitation to appear and was ignored. Can’t think of a more JRE-suited guest, frankly.
That Joe refuses to platform any strong pro-Ukraine or pro-Israel voices speaks of his views. It's his right—but let us remember: the original fail mode of MSM was selective bias by omission. Joe, now a platform as big as MSM, deserves the same scrutiny. You can't call it "open dialogue" if only one side ever gets airing to millions.
2) Churchill the baddie.
Joe and guests pushing Churchill conspiracies are—maybe without realising it, for they are forever ignorant even to click on a Wiki page—rehashing 1942 Goebbels propaganda ("Churchill controlled by Jewish bankers," etc). In Europe, WW2 survivors are still alive (e.g. my mum). These takes aren’t edgy—they’re lazy.
Fr: it was Germany that built up its military in defiance of Versailles, annexed Austria, dismembered Czechoslovakia, and invaded Poland on Sept 1, 1939. Two weeks later, the Soviet Union invaded Poland from the east, pursuant to the Molotov–Ribbentrop Pact, a plan to divide Eastern Europe between Germany and the Soviet Union. Enslave and or exterminated millions of people. This is not disputed. We have mountains of evidence—hand written from the architects and executors(!), diaries, state memos, radio archives, press coverage. Everyone can read everything at all times, it's not Latin or ancient Greek. It's all super-accessible. Even the original archives.
Joe is free to re-broadcast Goebbels-tier takes. But let’s not pretend it’s deep or brave. If he wants to test the strength of his opinions, let him say this stuff in CEEurope (ex-Russia), in Poland or the Baltics. I suspect he may catch more heat than he can handle. He's travelled a long way from the original "savage Khan of the steppe curious about the skies asks medicine man to explain" schtick, and something got lost along the way.
Having written all that—I haven’t watched JRE for many moons now. There’s just better stuff on YT—more interesting, more entertaining, more informative. Many podcasters go to UKR-RUS, record, comment, interview etc. Much better to see IRL, even when selective. Whenever X references JRE, it's usually some B-tier cosplayer bores. Don't bother to check. That shi*t-chatting nonsense more entertaining elsewhere.
4:34 PM · Apr 22, 2025

https://x.com/ljupc0/status/1914678555356647818
Ljubomir Josifovski @ljupc0
WFH=Garden office <1min from home bathed in natural sun among greenery, instead of dull office under florescent lights crackling menacingly.
WFH=Personal space replacing open plan office crazy setup, where one wore heavy duty noise cancelling ear muffs to be able to think.
WFH=Can wear trainers or uggs or anything comfortable I fancy. No more wearing gloves with open finger tips for typing in the winter. (the office was freezing.)
Desktop=Not forced to use Windows anymore for 10h a day. Can use Ubuntu or MacOS or whatever I fancy. No need to deal with the local IT admin who answers "Nein" first thinks second b/c "security" always trumps everything, and including whether I can actually do my easier or harder.
Apple Silicon=Macbook screen 10x better than old Lenovo one, battery life 2x better than old Lenovo. (bonus fun: runs local LLM-s and other NN-s with 96 GB (V)RAM)
Car=electric/hybrid wheezing in silence with only the wind swishing as if sailing. No one goes back to ICE, unless they have to.
2:51 PM · Apr 22, 2025