PeakoQuant

@PtrPomorski Getting hired is all marketing, so you can ultimately help others do more marketing. Capability is just one sadly tiny component of that marketing.

Degree is one thing, but additional projects, github, even Uni (harsh truth) is second. If you’re only capable of what AI can do then obviously you’ll get rejected.

We're close now.

@investingidiocy It's not bizzare at all. People add indicators and lines to charts, feel out the trades, then call it a systematic model. That's the norm. What's bizzare is doing anything rigorous.

This is a bizzare mischaracterisation of how systematic models work. They don't consist of rules like "If price = 4000 and price rise last 7 days = 1000 then do X".

@f4stsc0pe Hard to keep up w/ models and tools, def need to give sonnet a try. I'm certain ChatGPT is the absolute worst for coding.

@PeakoQuant Now you know you should ask it to insert this in the header right away in your prompt. Insult it if you have to 😂 You should try Sonnet or Opus if you're rich though. They hit different

Repeatedly criticizing Gemini's code reliably causes it to insert "The 'God Mode' Solution" headers into all subsequent code attempts. Am I really asking that much ✝️😂

Adding `PASS ✅` and `FAIL ❌` icons in console log prints has been my biggest UI commit of the last 6 months.

@__paleologo When you are an expert at something even slightly complicated, nearly every pop movie about it seems embarrassing.

It's just so hard to make a good show about finance, because the insider's and the outsider's views are so mismatched. Both "Billions" (buy-side!) and "Industry" (sell-side!!) start well and then decline very rapidly.

@investingidiocy Finally a bit of sanity in the world 🙏

I think we all need some good news with what else is going on in the world. Here is some: as of today, you can pre-order my new book harriman-house.com/authors/robert…

@__paleologo Got to buy extra insurance for RAM/SSD's these days.

Fun fact. I paid $280 for a 4TB SSD external drive, eight months ago. Same model selling now for $490. It's real.

@MarkBestForex @PtrPomorski She used `from hmmlearn import hmm` instead.

@PtrPomorski did she end up using kalman filters?

Ok, I think we’re done

@systematicls Ravenpack is out of my league at $50k+ per year. Selfishly happy to hear it's usefulness has declined a bit 😁 Always wanted to peek at it though.

Wrote about ravenpack, an old and established dataset with amazing coverage of news providers. It's a good dataset to create news sentiment signals, but you'll have to watch out a few pitfalls embedded in the data! One common one is not filtering for entities that have low relevance in an article! (e.g. news article releases today with glancing presence of AAPL, but actually has no material news about AAPL). If you try to create an AAPL signal from this - it will be noise!

@HangukQuant You must have just found some new alpha 👀😁

the things that make money can be deceptively simple, if not pedantic. it is the path to finding this simplicity that is complex and laborious

@josedonato__ @didier_lopes That deep level of understand will always have great value though. I remember coding everything from scratch before "libraries" or auto memory management were even all that common 😅

I feel the same but with building OpenBB Workspace from scratch truly impressive how we used to build stuff before LLMs were a thing OpenBB has always encouraged innovation and experiments, which is exactly what helped shape and create such a unique product @didier_lopes 🫶

@quantbeckman Don't forget market selection, signal selection (portfolio), and portfolio selection (pf of pf). Models all the way down!

A #trading system is a model of models.

@5000e12 AI、ソーシャルメディア、政治。これらすべてが、私たちにとって、人々が本当は何者であるかを理解することを容易にしていますが、その真実は不快なものです。

AI活用できてる人が一次ソース確認せずに適当に共有している事例をなんども見てなんとも言えない気分になる。

@jakubpolec Does the flap of a butterfly's wings in Brazil set off a tornado in Texas?

Just run 5 different strategies over QuantJourney, Zipline and VectorBT - tried all similar from slippage/costs/rebalance/signals - modified calendar, etc. Still some differences. Any hints?

@kruzfx You've already lost if you think your only options are (a) chat with a mentor/grifter, or (b) watch grifter youtube videos. And no, don't mention podcasts and courses 😂

The failure rate in trading is higher than in medicine And some guys think they’ll become profitable With no mentor Just YouTube Crazy

@systematicls You're describing attributes of a skilled novel problem solver. Always been the ideal in R&D, before or after AI. But quite difficult to assess w/ interviews. AI is currently terrible at it.

One interesting thing about hiring practices from just 3-5 years ago is that they are built on the foundational belief that your ability to contribute is bottle-necked by "what you know". The idea then is that it is costly to "understand" knowledge, even if access to knowledge was cheap. So you looked for people that could demonstrate "understanding" deeply AND broadly. The T-shaped (deep and broad) knowledge pool was especially sought after - someone with great expertise in one area, and generalist understandings of various other connected and complimentary areas. We looked for this via technical tests, coding tests, and all forms of brain teasers. Knowledge used to be a moat, and if you were a T-shaped human you could find a job any where, any time. I'm not sure it will continue to be one. Prompting agents give you access to a pool of knowledge that is extremely deep and extremely broad. It is not (yet) as deep as the humans working at the frontier of said knowledge, but deep enough that for a vast majority of humans, they wouldn't be able to tell the difference. Prompting an agent is so (attention & effort) cheap that knowledge can no longer be a bottle-neck. In that case, what should you hire for? I'm still reasoning through this - but my initial sentiments are that you should assume that you can provide someone with any knowledge they need about how to do anything. In that case, you need someone that can put that knowledge into good, practical use. What does that mean? I think it means you are looking for people with: 1) High drive for action that they overcome all inertia in using this knowledge bank. 2) Have really high upload throughput to process a large amount of information / knowledge. 3) Synthesize disparate knowledge banks into something coherent AND applicable for the current problem. 4) That they can take something abstract and academic and put it into practice and face all the implementation details that come their way. TLDR: As knowledge become commoditized, it appears that agency (willingness to access that knowledge), throughput (literal upload speed of knowledge), reasoning (haha) becomes increasingly valuable.

@f4stsc0pe Per capita you'd need ~1 million followers (or less) today to beat J. Or even just a million views. Let's pray 1mil+ influencers don't set the precedent for the next 2000 years 😶. In Hawk Tuah's name we pray.

@PeakoQuant I'm not sure any of these accounts with 500 followers will be the foundation of civilization for the next 2000 years though

If you have ~500 followers you have more influence than Jesus did 🙏. Some top accounts reach almost the same number of people as existed at that time.