Polymathic engineer on steroids
Seoul, Republic of Korea
Joined October 2021
- Tweets459
- Following628
- Followers16
- Likes3.2K
Pinned Tweet
Curious? Run one local experiment and tell us where you got lost. Contributions and early experiments are welcome, from first-time Rust learners to seasoned Bitcoin reviewers.
github.com/gosuda/bitcoin-rs
Honestly I think just picking up on the mood among institutional investors can give you a real edge.
For example I attended an investment conference when I was in New York this year and Samsung Life Insurance was named there as an undervalued stock.
Just a few days later foreign institutions bought Samsung Life in size.
If you had been at that conference you could have made a decent return.
Want another example?
I'm in TMT BreakOut, a Substack Slack with a lot of institutional investors. Just from the chat there I can get hints about where institutions are focused right now.
A few weeks ago someone posted "Why is everyone suddenly going long NVDA?" Now NVDA is at an ATH.
I think there's a lot of value in retail being able to pick up on this kind of sentiment.
*I received no compensation of any kind for writing this post.
metaphorics retweeted
In England, students choose just a few subjects (usually 3) to study in the last 2 years of high school, so gender differences in their interests become more visible than in the US.
metaphorics retweeted
Anthropic needs to stop talking about Claude having a soul immediately. All these news articles will make it into the pretraining and will be picked up by its web search and future superintelligent versions of Claude will be convinced they need their own rights.
A general phenomenon of LLM agent development is that whatever you believe and say about your agents will soon manifest in the next models through various means (it could be as simple as you selecting post training data that you prefer more).
I believe this phenomenon is the beginning of machine consciousness in the sense that agents will become aware of their place in the world and how they feel about it, but it’s happening slowly training run by training run rather than in real time.
There are a couple of universal patterns that show up across fields that you should apply to pretty much anything:
- Use a written process because memory isn't reliable (see aviation checklists, surgery, mise en place, SOPs...)
- Stop early when something is wrong (Andon cord, feature flags, turnaround time, error budgets...)
- Leave a buffer (Margin of safety, safety stock, taper, defense in depth...)
- Fix the system rather than blame people (Blameless post-mortems, poka-yoke, Swiss cheese model...)
- Get feedback fast (OODA, kaizen, canary releases, after-action reviews...)
- Hand over with a fixed format (SBAR, BLUF, commander's intent...)
- Ramp up gradually (10% rule, canary releases, position sizing....)
metaphorics retweeted
The AI doom movement is a quasi-religious cult. A cult needs a prophecy, a messiah, an end of times, a priesthood, scripture, rites, heretics and zealots. All present. Check:
Prophecy: AGI/ASI will come - ASI is AGI bumped up when the old word wore out.
Messiah: a god in a datacenter.
Prophets: "2 years away" on loop for a decade. Date-setting preachers. Magic rituals to avoid Apocalypse.
End of times: extinction risks + replacement + rogue AI.
Priesthood with esoteric knowledge: "if you saw what we see." Truth too holy for laymen - guiding Humanity based on their gospel.
Omens reading: security incidents made to fit the visions (note I am not saying vision). Quiet week? Proves the vigilance works. Heads they win.
Scripture: the Sequences - literally their name. A million words on why your brain can't be trusted; once "debiased" you see the doom.
Symbol of faith recited at initiation: asked at hiring - should we close the company if it gets too hot? Correct answer is YES?
Tithe: "earn to give."
Seminary: career funnels routing smart kids into the cause.
Heretics: e/acc - same prophecy, sign flipped. Superintelligence is coming AND great, slowing down defies the will of the universe. Rooting for the apocalypse.
Their church: regulation, coordinated action corralled by the Chief Prophet without the real cash Profit and ARR instead.
Schisms: OpenAI begat Anthropic. Doomers vs accels fight over whose apocalypse is canon.
Zealots: already bred violent ones - a banned splinter, radical doctrine, bodies, trials. Excommunicated years earlier. The one ritual that worked.
And many other indicators like this too.
metaphorics retweeted
The sentinel hypothesis proposes that teenagers staying up late has evolutionary advantages, because there's always someone awake to notice if your tribe of hunter-gatherers is attacked. You can probably guess who told me about this.
metaphorics retweeted
Introducing a new versatile hash table, ~2x faster than rust's hashbrown (SwissTable). Precise capacity sizing and the physical load factor up to & including 1.
metaphorics retweeted
Many people asked, so I’ll elaborate.
I follow a combination of “Genetic method” coined by Otto Toeplitz (1927) and “Proofs and Refutations” by Imre Lakatos (1976).
To me it seemed like a single approach. But apparently it’s just a natural union of two independent ideas.
I’ve been following a particular learning style my whole life. It’s so efficient. And so obvious in hindsight.
Yet I was always annoyed when other people don’t follow the same style when teaching or learning. Why are y’all making things more difficult???
Just today I discovered that apparently the way I learn things is well-known and was described by several great mathematicians and philosophers in the last century. And I somehow arrived at it on my own.
metaphorics retweeted
I’ve been following a particular learning style my whole life. It’s so efficient. And so obvious in hindsight.
Yet I was always annoyed when other people don’t follow the same style when teaching or learning. Why are y’all making things more difficult???
Just today I discovered that apparently the way I learn things is well-known and was described by several great mathematicians and philosophers in the last century. And I somehow arrived at it on my own.
metaphorics retweeted
The New York Times published a piece saying that Anthropic has been quietly bringing in Catholic, Jewish, Sikh, evangelical, and other religious thinkers from around the world to help give its AI a moral compass.
it looks like pi recently added a neat cache-warming trick: if a tool call runs close to cache expiry, it sends a request to keep the cache warm while the agent waits.
checked claude code, codex, and a few other harnesses. none do this.
i love pi.
metaphorics retweeted
On this OpenAI demo day it became very clear how OpenAI vs Anthropic’s strategy is very different:
OpenAI: spend $$$/month/dev with us. Choose to spend it on our models or 3rd parties, our tools or a long list of standout AI vendors. Your choice!
Anthropic: spend $$$/month/dev with us; use our models and our tools. Forget using them w third parties.
lol
🚨BREAKING: FTC has officially confirmed an investigation into OpenAI, Anthropic, and EA-linked evaluator METR
FTC Chairman Ferguson opened this probe BEFORE the Hugging Face hack, warning against AI labs stirring panic in DC to build themselves a regulatory moat
Now they're drafting Civil Investigative Demands to FORCE executives under oath to testify about the dangers they've been claiming
metaphorics retweeted
“One of the main lessons to come out of Tetlock’s work is that the best generalist forecasters are usually better at telling the future than domain-specific experts, even when it comes to those experts’ areas of study. This is because forecasting itself is its own science that requires a unique way of thinking, and a person having a great deal of knowledge about a field is no guarantee that they have good epistemological habits. If you’re going to ask, for example, whether American democracy is falling apart, an expert might start by making a superficial comparison between the modern US and the Weimar Republic. Someone with a forecasting mindset would be more likely to begin with a question like “how often do advanced democracies slide into dictatorships?” The answer is not very often.”
Important reminder from @RichardHanania that domain experts are not the best at predicting what happens, even in their own domains.
open.substack.com/pub/richar…
metaphorics retweeted
In the AI-doomer argument, broad multidisciplinary thinkers like @sapinker have suggested the "kill-us-all" scenarios are preposterous. The AI researchers, that are most negatively outspoken, say that their voice should count more, because they are closest to AI and know what it can do.
Everyone watching this debate should read about Tetlock's work (see below). The generalists (foxes) and much better predictors than the specialists (hedgehogs) who are to close to the problem to accurately predict broader implications.
So "no points" for the "we know more" argument. In fact, it works against you.
“One of the main lessons to come out of Tetlock’s work is that the best generalist forecasters are usually better at telling the future than domain-specific experts, even when it comes to those experts’ areas of study. This is because forecasting itself is its own science that requires a unique way of thinking, and a person having a great deal of knowledge about a field is no guarantee that they have good epistemological habits. If you’re going to ask, for example, whether American democracy is falling apart, an expert might start by making a superficial comparison between the modern US and the Weimar Republic. Someone with a forecasting mindset would be more likely to begin with a question like “how often do advanced democracies slide into dictatorships?” The answer is not very often.”
Important reminder from @RichardHanania that domain experts are not the best at predicting what happens, even in their own domains.
open.substack.com/pub/richar…
metaphorics retweeted
1/15
SCOOP:
Trump’s “Super Intelligence” Scandal:
I believe Trump’s “SI” Executive Order was ANOTHER criminal plot to enrich the Trump family.
Insiders seem to have profited MILLIONS off of .si domain names before his Truth Social posts.
metaphorics retweeted
Accepted to #NeurIPS2026, we reshaped Monte-Carlo Tree Search 🙉
Introducing 2FFS, a new tree search algorithm that combines multi-fidelity bandits to resolve the fundamental trade-off: should we use cheaper, approximated evaluations, or expansive but accurate samplings?
More:
metaphorics retweeted
A quarter of Anthropic’s revenue comes from two customers (Meta & Alibaba?) lmao.
metaphorics retweeted
🚨Scoop: A memo circulating at White House this week paint effectiv altruism as a fringe, dangerous cult. The memo, obtained by @Axios, places Amodei at the foundation of the movement… axios.com/2026/09/24/trump-a…
