philomath_mn

I think that article is more about the public not understanding “real poetry”(understandably) rather than testament to AI capability.

But I agree with you: there is a selection bias going on. I have no idea how much good AI writing I’ve been reading without realizing it.

Exam was administered on December 7th 2024, so it is possible this is in the training data but unlikely.

The open source model as worked through the problem step by step (it was R1 so I could read its chain of thought). That could be an elaborate ruse, but also unlikely.

This isn't a handout, it is an investment with an expected return. Which is good, because it is less likely to be applied to bad ideas like forcing solar roadways and solar farmers.

> you still need people doing the systems designs that the AI can implement code for

Given the rate of improvement up to that point, how much longer do you think this would be the case?

I have a sinking feeling that this is correct but I'd love to understand the arguments for/against this position.

Do you have any podcast/blog/videos that you recommend which make your case clearly?

> The only thing AI can code is X

I am curious, do you think it will stay this way? AI couldn't code anything 2.5 years ago, why should it stop with glue code?

Yesterday I fed the first 3 questions from the 2024 Putnam Exam [0] into a new open source LLM and it got 3/3. The progress is incredible and I don't expect it to suddenly stop where it is.

[0] https://kskedlaya.org/putnam-archive/

Nice.

Reminds me of the discrete metric in my Real Analysis course: distance is 0 if it is the same point, 1 otherwise.

We used to joke that it was the perfect metric for finding weird counter examples

He took paid someone money (a concrete action) with the intent to have someone murdered. This isn’t rocket science

Not exactly what you're talking about, but `uv` lets you specify dependencies in the header of scripts: https://docs.astral.sh/uv/guides/scripts/#declaring-script-d...

I think what you describe really only makes sense for a single file script. I _do not_ want to manage dependency hell within my own source files.

Intent matters!! For all he knew, Ulbricht had killed those guys and he was fine with that

The arms race is already running, I think this showdown is inevitable so we should get our asses moving

Unless we air strike the data centers, there is no way to control China’s progress

But why drop $500B (or even $100B short term) if there is not something there? The numbers are too big

A quick search shows this item [0] from 7 years ago -- I remember this critique well before then.

[0] https://news.ycombinator.com/item?id=16934942

I don't think you can do anything serious locally.

IMO, the best option is hosting a VS Code server [0] on a linux box and using Tailscale to connect to it via the browser. At that point you have a real compute environment and a pretty usable IDE.

But the screen and keyboard were too small so I gave it up.

[0] https://github.com/coder/code-server

I mean we are all susceptible to confirmation bias, OP is just being explicit about it ;)

You can install it in several ways without pip, easiest is standalone installer (which can upgrade itself)

https://docs.astral.sh/uv/getting-started/installation/

Every startup is different. They differ in expected working hours, stability, compensation, equity, impact, mission, etc. Imo, most of these dimensions overlap with bigger corps. except impact: it is much easier to have a lot of impact at a startup.

Some of the other commenters have mentioned startups with low stability, low equity, low compensation, and high expected working hours. I wouldn't want to work there.

But I am at a startup with 40 hrs expected / week, good compensation, high impact, and medium stability -- that is a good tradeoff for me.

I wonder if performance would improve if you asked it to create csvs from the tables first, then fed the CSVs in to a new chat?

Do you set the screen in your lap?

If it works (and has worked for decades) why fix it?

He is paying for 5 employees and an office with 7k-10k a month? 25k / month revenue seems pretty low for that overhead.

Agreed. I am rarely tracking my what I am saying vs other people. I also focus on paying attention to what people are saying and trying to engage and ask good questions. I rarely volunteer my own stories / thoughts unless they are directly asked for or I have a very strong sense that people want to hear it.

Is there a commercially available version of this? Or a very simple mechanism / kit with a similar approach?

> delves

Dead giveaway :D

Do you charge for the web app version? I'd expect the module version to be somewhat close to that?

Also make sure to expect a lot of pain working with IT at the banks. I worked with some people at US Bank and their restrictions were absolutely draconian.

For example, I had to make a vanilla version of the python tool I was working on for them because they had to review every one of my dependencies! I also couldn't send any kind of zip file to them so I ended up making a plain text encoding/decoding scheme so I could attach code distributions to emails.

September 1st was a Sunday, plus the 2nd was a US holiday. We're closer to a ratio of 1.0 now, it is probably best to compare ratios after a week or two to see how things settle out.

Sounds like you found a great employer who actually invests in their people.

2.5x in 10 years is about a 10% annual raise. I've always had great performance reviews but never broke 10%. I'm at 3x salary over 8 years, but almost all of that gain came from switching jobs.

So a lot of people end up job hopping because internal raises are rarely competitive. That and a lot of companies are poorly managed and it gets old after a few years.

Yeah I pretty much always think through the unit circle to get the value of sin(0) and cos(0) (unless I had been doing a lot of trig lately)

I remember drawing everything out on trig exams though and making myself an on-demand cheatsheet.

This is interesting, I'm not sure my alarm actually wakes me up all that often. I'm usually mostly awake and the alarm is more so an indicator of what time it is.

This is the case even when I set extremely early alarms like my 2:30 wake up time the other day.

But being a parent has also trained me to be an extremely light sleeper

They caught Ross Ulbricht (Silk Road founder) at a public library.

To be fair, he was kind of an idiot when it came to security. But there are ways of knowing who is using public wifi at different times.