michael 📈 👨💻
@mikedecr.computer
Low-latency quant finance R&D.
Former political scientist, sometimes bike rider.
Long form mikedecr.computer
Former political scientist, sometimes bike rider.
Long form mikedecr.computer
Pinned
michael 📈 👨💻
@mikedecr.computer
· Mar 17
I don't see what the big deal is. I simply leverage the computational power of LLMs while breathing the ineffable and inimitable human spirit into my creation. It's trivial really
GAUUACA
Downgrade a movie:
The Vector
The Vector
Downgrade a movie
Annoyed Max
Annoyed Max
August 28, 2026 at 12:37 PM
GAUUACA
Reposted by michael 📈 👨💻
the logs refer to me as "the event."
August 27, 2026 at 2:19 AM
the logs refer to me as "the event."
Reposted by michael 📈 👨💻
become unpriceable
August 26, 2026 at 3:18 AM
become unpriceable
Now available as a blog post: mikedecr.computer/blog/sugar-f...
August 24, 2026 at 10:45 PM
Now available as a blog post: mikedecr.computer/blog/sugar-f...
plenty of weird and tough things about getting older but one thing I really do love is not knowing what the hell people are talking about. I love not being bothered w/ stuff. "Alex Cooper's beverage business is failing." I don't know who that is or whether that's good or bad. you kids have fun
August 14, 2026 at 4:34 PM
plenty of weird and tough things about getting older but one thing I really do love is not knowing what the hell people are talking about. I love not being bothered w/ stuff. "Alex Cooper's beverage business is failing." I don't know who that is or whether that's good or bad. you kids have fun
I read stuff like this and it's like, okay I just can't pretend to be nonchalant about it in situations like this:
I. AM. BAYESIAN.
I. AM. BAYESIAN.
Hey #stats nerds, a confidence interval question.
Say you have one person with 60 repeated measurements; you want an estimate of just that person's "true" score.
Could you calculate a confidence interval using traditional frequentist formulas for 60 different people and have it be meaningful?
Say you have one person with 60 repeated measurements; you want an estimate of just that person's "true" score.
Could you calculate a confidence interval using traditional frequentist formulas for 60 different people and have it be meaningful?
August 12, 2026 at 12:30 AM
I read stuff like this and it's like, okay I just can't pretend to be nonchalant about it in situations like this:
I. AM. BAYESIAN.
I. AM. BAYESIAN.
there's a lot of what I would just call "expertise" or "experience" that lead me to get very frustrated with LLMs sometimes. Today's example: generating basis spline fit for me (great use case bc I hate spline boilerplate). But I want to set knots that go outside the range of X. I get an error...
August 10, 2026 at 11:05 PM
there's a lot of what I would just call "expertise" or "experience" that lead me to get very frustrated with LLMs sometimes. Today's example: generating basis spline fit for me (great use case bc I hate spline boilerplate). But I want to set knots that go outside the range of X. I get an error...
do I know any data engineers or software devs who work on research tech? There will be some hires on my old team working with our research platform / financial market data pipelines / distributed compute etc. Crazy data and very cool tech where Python meets high-perf C++
August 2, 2026 at 12:26 AM
do I know any data engineers or software devs who work on research tech? There will be some hires on my old team working with our research platform / financial market data pipelines / distributed compute etc. Crazy data and very cool tech where Python meets high-perf C++
went to a skatepark today, did not die, no broken bones
August 1, 2026 at 1:09 AM
went to a skatepark today, did not die, no broken bones
![error[E0308]: mismatched types
--> /messages/10378318/80554717
|
1 | je suis aller à Paris
| ---- ^^^^^ expected past participle, got infinitive
| |
| because of this auxiliary verb
help: replace with: `allée`
|
1 | je suis allée à Paris
| -- ^^^^^ feminine past participle
| |
| this refers to a person with feminine pronouns
For more information about this error, try `frenchc --explain E0308`.
error: aborting due to 1 previous error](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:ym6n47hhl23xcx3hi7qncbh4/bafkreihntcysfj7vbvf5pnyb3uyayfekcfqfbg6wt2dvgoon5gzm4okpwi)
gf and I have been going to a nearby parking lot to skateboarding. she's learning fast. and I'm gonna hurt myself for sure
July 23, 2026 at 1:55 AM
gf and I have been going to a nearby parking lot to skateboarding. she's learning fast. and I'm gonna hurt myself for sure
It's perfectly fine to have a principled commitment to empiricism and scientific discipline, actually. And when political actors try to bully you into being quiet or changing your story, it shows your integrity and bravery to resist that pressure. Right or left.
holy moly @hausofdecline.bsky.social in @inthesetimes.com ???
July 21, 2026 at 12:56 PM
It's perfectly fine to have a principled commitment to empiricism and scientific discipline, actually. And when political actors try to bully you into being quiet or changing your story, it shows your integrity and bravery to resist that pressure. Right or left.
Reposted by michael 📈 👨💻
Someone just sent me this photo from Chicago, unbelieveable
July 16, 2026 at 2:20 PM
Someone just sent me this photo from Chicago, unbelieveable
Reposted by michael 📈 👨💻
RIP to everyone techdebted by the LLMs, but I'm different. And better.
July 3, 2026 at 5:26 PM
RIP to everyone techdebted by the LLMs, but I'm different. And better.
Reposted by michael 📈 👨💻
therapist: and... would you say this flips the story?
claude: you're right *sobs*
therapist: I'm right tooooo?
claude: you're right to push back on this. *wails*
claude: you're right *sobs*
therapist: I'm right tooooo?
claude: you're right to push back on this. *wails*
June 30, 2026 at 11:03 AM
therapist: and... would you say this flips the story?
claude: you're right *sobs*
therapist: I'm right tooooo?
claude: you're right to push back on this. *wails*
claude: you're right *sobs*
therapist: I'm right tooooo?
claude: you're right to push back on this. *wails*
Safari will exit full screen if you hit the Esc key. it's infantilizing
June 30, 2026 at 2:18 AM
Safari will exit full screen if you hit the Esc key. it's infantilizing
Reposted by michael 📈 👨💻
Reposted by michael 📈 👨💻
Kalshi now allows you to place Pascal’s Wager
June 10, 2026 at 5:37 PM
Kalshi now allows you to place Pascal’s Wager
Reposted by michael 📈 👨💻
Also, people now need to use as.factor() a lot more often, and they don't necessarily realise.
notstatschat.rbind.io/2024/03/22/f...
notstatschat.rbind.io/2024/03/22/f...
Factors as factors - Biased and Inefficient
notstatschat.rbind.io
June 9, 2026 at 9:54 PM
Also, people now need to use as.factor() a lot more often, and they don't necessarily realise.
notstatschat.rbind.io/2024/03/22/f...
notstatschat.rbind.io/2024/03/22/f...
i bought a skateboard so you know I'm early to hit m*d l*fe cris*s
June 7, 2026 at 1:54 AM
i bought a skateboard so you know I'm early to hit m*d l*fe cris*s
This blog post does not render in quarto without a Jupyter kernel for Hy. There is such a thing already but it's old and broken. LLM diagnosed and rolled its own kernel adapter for Hy in like 3 minutes. It is shipped with the blog post's src code so you too can feel the Hy of lisp programming
I like language exercises like these, so I implemented some of these functions in Hy, which is a Lisp that generates Python.
We like functions right? Would we still like functions if they were **evil**?
mikedecr.computer/blog/r-hy/
We like functions right? Would we still like functions if they were **evil**?
mikedecr.computer/blog/r-hy/
June 1, 2026 at 11:45 AM
This blog post does not render in quarto without a Jupyter kernel for Hy. There is such a thing already but it's old and broken. LLM diagnosed and rolled its own kernel adapter for Hy in like 3 minutes. It is shipped with the blog post's src code so you too can feel the Hy of lisp programming
I like language exercises like these, so I implemented some of these functions in Hy, which is a Lisp that generates Python.
We like functions right? Would we still like functions if they were **evil**?
mikedecr.computer/blog/r-hy/
We like functions right? Would we still like functions if they were **evil**?
mikedecr.computer/blog/r-hy/
![hy code and output:
(import toolz [first second])
(require hyrule [->])
(defn which [lst]
(-> lst
(enumerate)
((curry filter (compose bool second)))
((curry map first))
(list)))
(which [False False True False True])
[2, 4]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:qjxju2w5sdit42sjb4qpnatk/bafkreifnww5z4tnt3m7uuoryvws6yil3ezpfcta3bdhojeij455t7lpmoe)
June 1, 2026 at 3:00 AM
I like language exercises like these, so I implemented some of these functions in Hy, which is a Lisp that generates Python.
We like functions right? Would we still like functions if they were **evil**?
mikedecr.computer/blog/r-hy/
We like functions right? Would we still like functions if they were **evil**?
mikedecr.computer/blog/r-hy/
Reposted by michael 📈 👨💻
May 31, 2026 at 4:47 PM
Reposted by michael 📈 👨💻
I still believe it's a mistake to believe that the fact that LLM capability along certain dimensions seems to increase with scale implies that a sufficiently large LLM can do anything imaginable. Or that if an LLM can perform a task with difficulty D, it can do anything with difficulty ≤ D.
May 21, 2026 at 6:50 PM
I still believe it's a mistake to believe that the fact that LLM capability along certain dimensions seems to increase with scale implies that a sufficiently large LLM can do anything imaginable. Or that if an LLM can perform a task with difficulty D, it can do anything with difficulty ≤ D.