michaelmarkell

6 ms·

1.
by michaelmarkell 8mo ago
You're 100% right, thanks! Updated my comment.
2.
by michaelmarkell 8mo ago
That's approximately ~15.3M USD per person in the top 1% assuming 340 million americans
3.
by michaelmarkell 11mo ago
Not really. Take for example: item, date, price abc, 01/01/2023, $30 cde, 02/01/2023, $40 ... 100k rows ... subtotal. $1000 def, 03/01,2023, $20 "Hey Claude, what's the total from this file? > gr
4.
by michaelmarkell 11mo ago
IMO, a real solution here has to be hybrid, not full LLM, because these sheets can be massive and have very complicated structures. You want to be able to use the LLM to identify / map column headers, while using non-LLM tool calling t
5.
by michaelmarkell 11mo ago
Danya was like the Mr. Rogers of chess. He had a way of making you feel accepted into the chess community even if you were a beginner, and was such a clear thinker. I strive to be more like him, and am devastated by this loss.
6.
Fast UPDATEs for the ClickHouse column store(clickhouse.com)
4 points by michaelmarkell 1y ago|0 comments
7.
by michaelmarkell 1y ago
The way my company uses Clickhouse is basically that we have one giant flat table, and have written our own abstraction layer on top of it based around "entities" which are functions of data in the underlying table, potentially ad
8.
A New Postgres Block Storage Layout for Full Text Search(paradedb.com)
21 points by michaelmarkell 1y ago|2 comments
9.
Anthropic's Circuit Tracer(github.com)
2 points by michaelmarkell 1y ago|1 comments
10.
by michaelmarkell 1y ago
From the Readme: Given a model with pre-trained transcoders, it finds the circuit / attribution graph; i.e., it computes the direct effect that each non-zero transcoder feature, transcoder error node, and input token has on each other
11.
by michaelmarkell 1y ago
Syncopate | NYC (Hybrid ~3d/week) | Full-time | Senior Full Stack Engineers / Focus on AI + Finance Syncopate builds tools to help automate financial diligence and management of long-tail financial assets. We've found product
12.
by michaelmarkell 1y ago
In our use-case we have many gigabytes of PDFs that contain some qualitative data but also many pages of inline pdf tables. In an ideal world we’d be “compressing” those embedded tables into some text that says “there’s a table here with th
13.
by michaelmarkell 1y ago
It seems to me like chunking (or some higher order version of it like chunking into knowledge graphs) is the highest leverage thing someone can work on right now if trying to improve intelligence of AI systems like code completion, PDF unde
14.
by michaelmarkell 1y ago
If I were to guess, most (adult) humans could not add two 3 digit numbers together with 100% accuracy. Maybe 99%? Computers can already do 100%, so we should probably be trying to figure out how to use language to extract the numbers from s
15.
by michaelmarkell 2y ago
Can someone with more expertise help me understand what I'm looking at here? https://crt.sh/?id=10106356492 It looks like Deepseek had a subdomain called "openai-us1.deepseek.com". What is a legitimate use-ca
16.
by michaelmarkell 2y ago
Archive link: https://web.archive.org/web/20240330143422/https://www.nytim...
17.
Dinner and Deception (2015)(nytimes.com)
1 points by michaelmarkell 2y ago|1 comments
18.
Parsebox | Parser Combinators in the TypeScript Type System(github.com)
1 points by michaelmarkell 2y ago|0 comments
19.
by michaelmarkell 2y ago
Timing of moves
20.
by michaelmarkell 2y ago
Artists do not get paid per stream on Spotify + many other DSPs. The platform sums up all of the ad revenue and divides it pro rata among all of the streamed artists. So the fraudulent streams dilute the pie for legitimate streams.
21.
by michaelmarkell 2y ago
The Monty hall problem is a great example of something I’ve been educated into believing, rationalizing, whatever you want to call it…but I would still never claim I “understand it.” I think that’s maybe the source of disagreement here, the
22.
by michaelmarkell 2y ago
can you run the whole task as a postgres transaction? like if i want to make an idempotent job by only updating some status to "complete" once the job finishes.
23.
by michaelmarkell 2y ago
From the article it seems like the content wasn’t entirely AI generated, it “digitally superimposed the faces of child actors onto nude bodies”, which makes this a lot worse than purely AI generated content
24.
by michaelmarkell 3y ago
> which implies what casual mechanisms you're controlling for. Without thinking terribly deeply about it, it seems like this could be modified to "what correlated mechanisms we're controlling for", in which case, yes
25.
by michaelmarkell 3y ago
This reads like it was written by ChatGPT
26.
by michaelmarkell 4y ago
Treatments like exercise also have high variance - depending on the circumstance and how diligently you implement it, it may have no effect, a big effect, or even cause injury in the worst case. Insurance likes low variance outcomes because
27.
by michaelmarkell 4y ago
I'm building https://royal.io , which helps turn music royalties into an asset class people can invest in. When investors (fans, mostly) "buy" a portion of a song on our platform, they become entitled to a percent
28.
by michaelmarkell 4y ago
I right a lot less now then I did as a kid, so maybe it’s about just staying sharp
29.
by michaelmarkell 4y ago
But saying “I don’t want to be wildly unkind” implies that there is something about the author and their claim that is deserving of meanness. I think that is an unkind thing to say.
30.
by michaelmarkell 4y ago
As a dev, you build on mainnet because that’s where the liquidity is. Your project will get less trade volume on any other chain (even compared to an EVM compatible chain like polygon, where transaction times are faster and gas fees are low

More ›