Joonbeom Kwon

316 posts

Joonbeom Kwon banner
Joonbeom Kwon

Joonbeom Kwon

@kwonai

Prev. Undergrad @UCBerkeley

Berkeley, CA Katılım Haziran 2020
1.5K Takip Edilen232 Takipçiler
Joonbeom Kwon retweetledi
Joonbeom Kwon
Joonbeom Kwon@kwonai·
@mitchellh Could you share your CLAUDE.md and other useful system prompts? I wonder the difference between yours (a top-class engineer I respect) versus mine (a novice/nobody). And, whether using yours increases code output quality.
English
0
0
0
226
Mitchell Hashimoto
Mitchell Hashimoto@mitchellh·
Btw, for Ghostling (libghostty demo), I didn't write a single line of... anything. Agents wrote 100% of everything you see incl. Nix flakes, CI jobs, etc. I reviewed every line of code manually and constantly nudged the agents in the right direction. I used a mix of Opus+Codex. Note: libghostty itself is of course heavily hand written (with agent assistance all over too, though). I'm just talking about Ghostling itself. Even for CI setup (GitHub actions), I had the agent sit in a `gh` CLI loop watching failures, fixing, pushing, fixing, pushing, etc. Doing GHA with agents are the only way I stay sane, honestly. I just re-read the full main.c from top to bottom and I'm very satisfied. I would've done some things differently but if an engineer I worked with PRed all this, I would've accepted it. Its good enough. Some people have pointed out the commenting is over the top and indicative of AI. This is actually my personal comment style, and I told the agent to comment heavily (its even in the AGENTS.md, look for yourself!). Anyone who has worked with me professionally or in OSS knows that I comment everything all the time.
Mitchell Hashimoto tweet media
English
76
45
771
108.4K
Joonbeom Kwon retweetledi
Joonbeom Kwon
Joonbeom Kwon@kwonai·
However, time will tell whether this was truly helpful for me. While I found it necessary/helpful in the present, it depends what one is optimizing their life for, and if one is optimizing their life to be more like Elon, then my disagreement with him here is a clearly negative indication, and my introspection will have been a mistake.
English
0
0
1
12
Joonbeom Kwon
Joonbeom Kwon@kwonai·
Introspection indeed amplified/disinhibited my emotional signals, but that helped me realize/distinguish suppressed emotions causing mental anguish, which helped me identify real causes for those emotions, which allowed me to take real action against those causes, effectively decreasing mental anguish and significantly helping treat my depression.
Elon Musk@elonmusk

Reinforcing negative neural pathways via therapy or introspection is a recipe for misery. Don’t cut a rut in the road.

English
1
0
2
32
Joonbeom Kwon retweetledi
vitalik.eth
vitalik.eth@VitalikButerin·
Programming languages, a domain that does not have deep established cultural history, quickly adopted syntax formatting and highlighting, because those things are better for readability. Math notation has a much deeper cultural history, and also did a similar thing: math notation is inherently a very 2D language, because that's better for readability. I think it's reasonable to believe that the underlying advantages of these styles are likely also at least sometimes true for written text. Prose is one-dimensional because we historically have such a strong tradition of wanting spoken text and written text to correspond to each other. Now, we get an opportunity to re-evaluate that convention, and we discover that patterns other than one-dimensional linear prose are more reader-friendly, of which bullet points are a major one. I think this is all a healthy development.
English
56
37
894
127.1K
Joonbeom Kwon retweetledi
Paul Graham
Paul Graham@paulg·
Another advantage of focusing on growth rate rather than absolute numbers is that it makes it easier to switch to a new variant of the product if you discover one. It makes it easier to see tails that will eventually wag the dog.
English
86
86
1.2K
89.2K
Joonbeom Kwon retweetledi
Gaurab Chakrabarti
ASML's EUV scanners will be the last machines on Earth to lose helium. Party balloons will be the first to go. Helium is not manufactured. It is a byproduct of uranium and thorium decaying deep underground over billions of years. Vent it and it escapes to space. Permanently. A third of the global supply went offline when Qatar's Ras Laffan plant was hit on March 2. The rationing has already started. Here's what happens: Day 1. Party balloons. Distributors cut retail supply immediately. Day 7. Industrial welding and pressurization. National allocation kicks in. Switch to argon where possible. Day 14. Routine fab leak detection switches to hydrogen. Ultra-sensitive qualification still needs helium. Day 21. MRI machines. Older systems that vent helium cannot get refills. Elective scans delayed. Day 45. Global buffer depletes. Fabs enter conservation mode. Non-critical depositions switch to nitrogen. Day 60. Backside wafer cooling on older etch tools. Nitrogen conducts heat six times slower. Throughput drops. Day 90. High-power etch. Advanced memory and logic nodes cannot run without helium-grade cooling. Wafer production drops. Day 120. ASML's EUV lithography tools. $200 million scanners making the highest-value wafers on Earth. Leading-edge chip production stops. Day 240. $700 billion in data centers are being built this year. Higher GPU prices, delayed cluster expansions, slower scaling. Four months from birthday balloons to AI chip shortage.
Balaji@balajis

I'm going to make some obvious points. (1) Blowing up all the oil infrastructure in the Middle East is an insane idea, and may well result in a global economic crash and humanitarian crisis unrivaled in the lives of those now living. We're talking about the price of everything everywhere rising, from food to gas, at a moment when inflation was already high. All of that will be laid at the feet of the authors of this war. (2) The antebellum status quo of Feb 27, 2026 was just not that bad, but we're unlikely to return to it. Expect indefinite, long-term, ongoing disruptions to everything out of the Middle East. (3) Also assume tech financing crashes for the indefinite future. The genius plan to get the Gulf states caught in the crossfire has incinerated much of the funding for LPs, for datacenters, and for IPOs. Anyone in tech who supported this war may soon learn the meaning of "force majeure" as funding gets yanked. (4) Many capital allocators will instead be allocating much further down Maslow's hierarchy of needs, towards useful basic things like food and energy. (5) It's fortunate that all those progressives yelled about the "climate crisis." Yes, their reasoning about timelines was wrong, and much of the money was wasted in graft, but the result was right: we all need energy independence from the Middle East, pronto. It's also fortunate that Elon and China autistically took climate seriously. Now they're going to need to ship a billion solar panels, electric vehicles, batteries, nuclear power plants, and the like to get everyone off oil, immediately. (6) It's not just an oil and gas problem, of course. It's also a fertilizer problem, and a chemical precursor problem. Maybe some new sources will come online at the new prices, but it takes time to dial stuff up, particularly at this scale, so shortages are almost a certainty. That said, China has actually scaled up coal-to-chemicals[a,c] (C2C), and there's also something more sci-fi called Power-to-X[b] which turns arbitrary power + water + air into hydrocarbons. But all of that will need to get accelerated. I have a background in chemical engineering so may start funding things in this area. (7) Ultimately, this war is going to result in tremendous blame for anyone associated with it. It's a no-win scenario to blow up this much infrastructure for so many people. Simply not worth it for whatever objective they thought they were going to attain. But unless you're actually in a position to stop the madness, the pragmatic thing to do is: scramble to mitigate the fallout to yourself, your business, and your people. [a]: reuters.com/business/energ… [b]: alfalaval.com/industries/ene… [c]: reuters.com/sustainability…

English
42
195
1.2K
120.7K
Joonbeom Kwon retweetledi
Andrej Karpathy
Andrej Karpathy@karpathy·
@ChristosTzamos Wait this is so awesome!! Both 1) the C compiler to LLM weights and 2) the logarithmic complexity hard-max attention and its potential generalizations. Inspiring!
English
28
43
1.4K
35.2K
eran shir
eran shir@eranshir·
Most AI assistants agree with you. I built one that doesn't. Challenger is a Claude Code plugin that tears apart your thesis, steelmans the opposition, researches evidence, then tells you what it actually thinks — with verdicts, assumption chains, and falsifiable predictions. It's open source, installation link the first comment. Tested it today on the Iran war's oil impact. It challenged 5 claims across military analysis, supply chain economics, and commodity markets. Ended with 20 specific predictions and review dates (check my feed for the full article on the predicted oil prices dynamics).
English
4
4
75
11.2K