tech, math, physics, automation, quant trading, foodie, BKK/TH

In the cloud
Joined November 2011
Nice tale of an 1875 expedition to Siam by the Royal Society…
Not my most viewed videos but I love these so much. The latest Objectivity about “re-discovered” glass plates capturing a long ago eclipse. youtu.be/cW7P3LVV-YA
3
76
I couldn't not ...
69
352
49
1,308
64,139
Let me correct the image so it better fits in with my experience 😉
🤖 Made with AI
1
5
180
Walmart is selling real diamond earrings for $12. Not a typo. Half-carat lab-grown diamonds, same carbon, graded by the same labs that grade the stone in your engagement ring. A jeweler can't tell them apart without a machine. And that $12 price tag is why the company that invented the engagement ring is about to get sold for scrap. De Beers ran the diamond trade for over a century. At its peak it controlled up to 90% of the world's rough supply, and when demand dipped it just locked diamonds in London vaults until prices came back. The slogan "A Diamond Is Forever" came from their ad agency in 1947. So did the rule that a ring should cost two months of your salary. Before that campaign, most American brides never got a diamond at all. The whole thing held for 80 years because there were only so many mines. Then China started growing diamonds in reactors, and there's no vault on earth big enough to hide a factory's output. Lab-grown wholesale prices have crashed roughly 96% from their peak, down to around $100 a carat. The Knot says 61% of engagement rings sold in 2025 had lab-grown center stones, up 239% since 2020. Even De Beers' own average selling price fell 32% in a year, to $105 a carat. So Anglo American, which owns De Beers, has written it down by $6.8 billion in two years and is now in talks to sell the whole thing for about $1 billion. This is a company that was valued over $18 billion in 2001 and $13 billion as recently as 2011. The stones coming out of the ground are the same as ever. The only scarce thing De Beers ever owned was the marketing, and marketing doesn't survive a $12 price tag at Walmart.
Diamonds are about to be absolutely worthless, nature is finally healing…
Readers added context they thought people might want to know
Walmart $12 0.5ct "Created White Diamond" studs (Paris Jewelry) often fail diamond testers per reviews and are not genuine lab-grown diamonds. Certified lab-grown 0.5ct pairs list around $79-89. walmart.com/ip/14k-White-G… walmart.com/ip/JeenMata-Ce…
330
1,286
194
8,072
946,497
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel. - Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all. - Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract. - Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code. - And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while. The funnel is driven by a combination of factors: 1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel. 2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel. 3. Access. Free-tier, paid-tier, internal. So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.
2,198
5,866
2,257
43,793
6,790,794
My dream had always been to prove beautiful theorems as the AI agents unloaded my dishwasher. Well, it went the wrong way... 😂
40
311
38
2,871
136,084
TIL 👇🏻🤷‍♂️
I was having dinner last night with a friend working in the chicken industry and I learned something that blew my mind. Where do chicken farmers get their chickens? Most of us would say "from their existing chickens obviously, they breed and lay eggs." That's what I thought too, but turns out I was dead wrong. The reality is that the immense majority of meat chicken farmers in the world get their chickens from one of 2 companies: Cobb-Vantress, owned by Tyson Foods, or Aviagen, controlled by Germany's EW Group (the Wesjohann family). They have a concept specific to the industry called grandparent stock: what you buy from these 2 companies is - essentially - a bunch of chickens with planned obsolescence. The way it works is that each company keeps "pure" bloodlines on its own farms. Let's label them A, B, C and D. Those are the crown jewels - these companies' IP - and they never leave the building. What they sell you are the "grandparents": chicks from those four lines, but each line in only one sex. You get A roosters but no A hens, B hens but no B roosters, C roosters but no C hens, D hens but no D roosters. Which means you can never breed more A, B, C or D birds yourself. What you can do is cross them. A roosters with B hens give you the fathers, C roosters with D hens give you the mothers - these are the "parents". The parents then give you the chickens that actually end up on our plates, slaughtered at around 6 weeks. And that's the end of the road. If you tried breeding those chickens, their chicks would be a complete genetic lottery: still chickens, sure, but without the genetic properties that make them commercially viable. They'd grow slower and less evenly, you'd have to sell them for way more than your competitors, and your business would be dead. So you have to go back to Cobb or Aviagen for new grandparents. Forever. Why buy grandparents at all, and raise 2 generations just to sell the third? The answer is scale. If you buy say 100 grandparents, each of them gives you around 50 parents, and each parent hen then gives you around 150 chicks: from those 100 grandparents, you have yourself some 600,000 chickens whose meat you can sell. It makes no sense for these 2 companies to sell the final chickens directly. They'd have to hatch tens of billions of chicks a year and ship them all over the world. Much easier to simply sell the grandparents and let farmers do the multiplying and pay for the farms, the feed and the labor. And, you knew this was coming, there is also a China angle here: as you can imagine, China isn't too pleased to be dependent on foreign companies for such a basic food item as chicken. So China did what China does best: long-term planning. The Ministry of Agriculture and Rural Affairs launched a plan for "national broiler genetic improvement" (2021-2035), which aims for domestically bred breeds to make up at least 60% of the domestic market by 2035 (globaltimes.cn/page/202307/1…). And, of course, they're ahead of the plan. As of 2025, homegrown breeds held more than 30% of the market: halfway to the 2035 goal in just four years (agripost.cn/2026/08/17/china…). Can you escape all this as a consumer? Sure, you can buy organic or Label Rouge chicken, from slower-growing breeds raised for at least 81 days instead of 6 weeks, for two to three times the price. Except, plot twist: the genetics behind most of those slow-growing birds come from a French company called Hubbard... which has belonged to Aviagen since 2018 (euromeatnews.com/aviagen-ann…)😅 And even if you could truly escape, you wouldn't want everyone to: we'd need vast new swathes of farmland, both for the chickens and their feed. Ironically, advocating for organic free-range chicken for everyone is one of the worst positions one can take for the planet's resources. If you want to minimize your environmental footprint, the counterintuitive answer is to buy the cheapest supermarket chicken - or eat less meat, of course. As my friend was saying last night, you may not like the way this industry works but, at the end of the day, it's what puts affordable protein on the plates of billions of people, using proportionally fewer resources than any other way of raising chicken.
1
68
New premium seating option lets you sit next to Israelis for extra security
1,576
3,860
574
41,691
2,120,251
.david retweeted
My god this is such a good speech that every SWE needs to hear. You know what? Every person should hear it Keep the happy memories, eyes on the reality, be excited about the future. That’s the best that anyone can do
377
2,070
656
15,855
3,697,740
Sukhumvit Fridays // rainy season edition
6
185
holy shit i asked claude to make a video on western civiization
2,743
10,565
2,911
61,460
15,188,066
'About your cat, Mr. Schrödinger — I have good news and bad news.' (Benjamin Schwartz, New Yorker)
6
20
63
1,602
Want 👇🏻🤩
Lmao ok fine I’ll buy it
Readers added context they thought people might want to know
This video is an AI-generated/CGI promotional concept for a fictional F-14 toaster prototype, not real footage of an available product. Its creator describes it as such. youtube.com/watch?v=z7OYMN…
1
7
1,319
I just had my Grok Bot pull together the best clips of Starship sitting in the ocean after recovery. And I’ve watched this part of @SpaceX's “Holy Grail of Rocketry” film over and over. Bc when you isolate these shots, the scale of what SpaceX is doing really hits you. A full Starship sitting in the middle of the ocean. SpaceX employees standing on it and next to it looking absolutely tiny. Bro... that ship just went to space. And it came screaming back through Earth’s atmosphere faster than a bullet. And it survived reentry. And it splashed down beautifully in the ocean. And then humans went out there to recover this giant stainless-steel spaceship, looking almost like an alien spaceship that just landed. You see, for most of the history of rocketry, rockets were basically one-shot machines. We would launch them and then lose them. SpaceX is fighting tooth and nail to completely change that with a goal to build a rocket system where the booster comes back, the ship comes back, you turn them around, and you fly them again. And again. And again. That is the "holy grail" of space exploration. We’re watching reusable spaceflight slowly turn from something that sounds impossible into something people will one day look at as completely normal. SpaceX is making this a reality... and this is the stuff that inspires me so much.
143
671
46
4,899
133,243
A nice explainer 🧵👇. One thing I’ve been coming back to in recent days is that the observation that the human process for solving these lighthouse problems also generates many more interesting question/problems as we gain insight. The current approach of using AI to solve these problems autonomously generally does not populate the next frontier of interesting questions. It occurred to me that the field of mathematics has something like a “joy reserve” of problems that humans are capable of/poise to solve, and that automated approaches are essentially burning through that joy reserve as a non-renewable resource. In contrast, in the traditional human approach the joy reserve is more like a renewable resource. That’s not fundamental to using AI to assist in doing research, but a property of the current approaches to auto formalization. The use of AI as a thought partner or creative assistant is quite the opposite— It helps to generate a lot of new ideas.
Navier-Stokes Millennium saga continues. Here are my slides and a summary from a recent talk. Disclaimer: this is a physicists view of the problem & solution! 🧵
2
8
19
3,309
.david retweeted
“Relativity Park”: a very fun interactive JavaScript game/demo which lets you explore walking around a toy universe in a world where the speed of light is 5km/h and see the strange visual effects that come with going at relativistic speeds. rivendell.dmitrybrant.com/re…
2
19
1
79
3,189
.david retweeted
Gen X defined: Ages 45-61 Looks 35-40 Will 100% throw hands Doesn't care about your feelings Cusses a lot Will make you cry with words Great work ethic Pragmatic Zero tolerance for stupidity Skeptical of authority Doesn't trust anyone at 1st Extremely loyal friend Figures shit out Most likely your boss Works hard, plays hard No fear Sarcastic as fuck Still listens to the same music Knows lyrics to Rock, Rap & Country Proficient in analog & digital tech Self reliant Adaptable to any environment Just wants to be left alone
3,082
9,484
2,099
57,178
2,528,130
1/ Since a lot of people are waking up to see their perps positions closed and wondering what the hell “Auto-Deleveraging” means, here’s a quick and dirty primer. What is ADL? How does it work? And why does it exist?
583
1,743
582
8,048
1,893,007
Bonds waiting for “the most important CPI print”
63
283
48
2,811
410,926