Live data from Hacker News

Ask HN: What are you working on (September 2024)?

news.ycombinator.com

331–340 of 1001 posts

Re: Ask HN: What are you working on (September 2024)?

#331

After a long break, I'm working on writing about Japan things I find interesting: https://onefromnippon.com/ I'm hoping Japanese vending machines will be an interesting topic for the next post.

I'd love to bring a Japanese beverage vending machine and drop it off in the middle of a reasonably populated US city and see how quickly it gets used (or, unfortunately, destroyed). Would only accept Yen (and ideally Suica/Pasmo, though unsure if those would work properly?).

Re: Ask HN: What are you working on (September 2024)?

#332
post #190

My primary project the past couple weeks has been applying the AI interpretability technique from Anthropic's famous paper this past summer (you know, the Golden Gate Claude one) to single-cell RNA-seq data. What works on one huge, inscrutible vector space ought to work on another, right? In either case, a fun way to keep in practice, both for comp bio and deep learning. My side-project, in essence, the bash '&' oper…

I may be misunderstanding what you're trying to achieve, but you can simulate the bash '&' operator for a running process by pressing ctrl-Z to suspend the process and send it to the background, then running 'bg' to continue the process in the background.

Yeah I was unclear there - I meant the other use of the & operator, 'foo & bar' where bar starts after foo completes. AFAIK, if you ran foo by itself, there's no way to make bar automatically run after foo completes.

Re: Ask HN: What are you working on (September 2024)?

#335
post #180

My primary project the past couple weeks has been applying the AI interpretability technique from Anthropic's famous paper this past summer (you know, the Golden Gate Claude one) to single-cell RNA-seq data. What works on one huge, inscrutible vector space ought to work on another, right? In either case, a fun way to keep in practice, both for comp bio and deep learning. My side-project, in essence, the bash '&' oper…

The few times I built TUIs with ncurses I wondered: why do I have to program so much by myself? ncurses is so basic, I didn't have too much fun building UIs with it (more than once). Is there a wrapper or a more modern alternative out there that provides containers and widgets like most GUI frameworks do?

Sadly, I think the modern alternative for ncurses-based interfaces is javascript.

Re: Ask HN: What are you working on (September 2024)?

#336
post #139

I learned how to weld (MIG) and built a giant mushroom to house a mannequin I dubbed "the mushroom man" over about 100 hours in the last 4 weeks. I covered the outside with thick foam panels cut to size, cementing them in place with copious amounts of spray foam. I shaved the outside to a nice shape with a sawzall and the inside I covered in chicken-fenced, then attached a painters tarp to that (so it could be painte…

Where/how did you learn welding? Have been considering doing a community college welding course (as they have all the equipment, instruction, etc.).

Re: Ask HN: What are you working on (September 2024)?

#337
Im currently working on AI agents that control my chrome browser, I have a desktop app with a chat where it does various tasks for me like research/posting on sm/finding free ebooks for me. All using open source LLMs and custom code for the auto web scraping / interactions.

Re: Ask HN: What are you working on (September 2024)?

#339

Building algorithmic trading models. So far results continue to be good with every model outperforming the market on both absolute and risk-adjusted basis since going live. Since launching https://grizzlybulls.com in January 2022: Model | Return | Max drawdown ------------------- S&P 500 (benchmark) | 21.51% | -27.56% VIX TA Macro MP Extreme | 64.21% | -16.48% VIX TA Macro Advanced| 59.13% | -19.12% VIX TA Advanced |…

If so good why are you selling it as a service ?

I expect the long term CAGR for the top models to be in the 20-40% annual range. That's certainly high enough to get wealthy over a couple decades, or sooner if you are already starting with 8 figures, but it's not overnight Roaring Kitty style fast money. Grizzly Bulls' growing revenue helps even out my overall income, and I could definitely see it growing to $10M+ ARR over the long term, very significant even with a 9 figure net worth.

The models are not HFT. Swing-trading the most liquid instrument in the world (ES futures) has extremely high strategy capacity, well into the billions or perhaps 10s of billions, so selling signals does not (currently) in any way negatively impact my own returns.

The alternative would be to start a hedge fund, but that's an expensive and highly regulated endeavor that appeals to a different audience.

Re: Ask HN: What are you working on (September 2024)?

#340

Building algorithmic trading models. So far results continue to be good with every model outperforming the market on both absolute and risk-adjusted basis since going live. Since launching https://grizzlybulls.com in January 2022: Model | Return | Max drawdown ------------------- S&P 500 (benchmark) | 21.51% | -27.56% VIX TA Macro MP Extreme | 64.21% | -16.48% VIX TA Macro Advanced| 59.13% | -19.12% VIX TA Advanced |…

I've been curious about doing algotrading for both the data engineering aspect and the quant. Do you have suggestions about books or others sources to get inspiration from ? Is this a one man venture or do you have a group discussing edges ?

For inspiration, I highly recommend "The Man Who Solved the Market" about James Simons and Renaissance Technologies. Some of Ernie Chan's books are great for learning about the basics, but ultimately finding an edge is the most difficult part. Books can teach you some of the best practices for researching edges, how to avoid common pitfalls in backtesting, etc, but no book will ever lay out the details of any strategy that contains alpha of course.

Grizzly Bulls is currently a one man (and wife) venture :)

Post reply on HN