Indeed, it kills me how much perfectly hardware is constantly thrown out because Windows refuses to run on it.

I think the trick has to be that somebody who has a bit of technical skill sets the laptop up initially. I did this for my mom a while back, and once I set it up once, it just worked from there on. Non technical users tend to have a fairly small set of things they need to do like check email, browser the web, and play media. Once that’s working, they never need to change anything. In fact, they don’t want to change anything because they get used to the workflow, and they’re comfortable.

It would be great if people set up community centres where people can bring their old laptops, and somebody switches them over to Linux for them.

I find what you generally want is a service bus. For example, if you have common tasks like sending emails, making PDFs, etc. you can create generic services that handle the action, and then call them providing the context such as the document the service should operate on. Meanwhile, actual business logic and the state should absolutely live in a single place.

welcome to the world that capitalism built

:)

Because they have a predatory business model.

Usually what happens is people create lambda functions on AWS which are basically managed functions that get invoked when a request comes in (think a modern version of CGI), and then if there are a lot of requests then the user can easily end up with a huge bill.

I find the most relevant aspect of an estimate is how familiar the person making the estimate with the problem. People who have the best understanding will inevitably give the best estimates.

Genetic algorithms use a fitness function to converge on a solution, so it’s not just random iterations hoping one will work. It’s basically a gradient descent algorithm applied to output from LLMs.

I’ve never touched ChatGPT out of principle, but everything I hear about it makes it sound like hot garbage.

My experience is that you want to keep changes as focused as possible. The smaller the context it has to work on the better the results tend to be, and the easier it is to check that it did what you actually wanted it to.

here’s what DeepSeek gave me for the following prompt: write a small bash script to check if a CSV exists and if it does rename to the filename_date.CSV and upload to hdfs, keep it simple

#!/bin/bash

# Check if filename argument is provided
if [ $# -eq 0 ]; then
    echo "Usage: $0 <csv_filename>"
    exit 1
fi

CSV_FILE="$1"

# Check if file exists
if [ ! -f "$CSV_FILE" ]; then
    echo "Error: File $CSV_FILE not found!"
    exit 1
fi

# Generate new filename with date (YYYYMMDD format)
DATE=$(date +%Y%m%d)
NEW_FILENAME="${CSV_FILE%.*}_${DATE}.csv"

# Rename the file
mv "$CSV_FILE" "$NEW_FILENAME"

# Upload to HDFS
hdfs dfs -put "$NEW_FILENAME" /path/to/hdfs/directory/

echo "Successfully processed: $NEW_FILENAME"

Not just OpenEvolve, I’m talking about how agents can use MCP to run code now. Cursor can now have agents iterate on a solution completely on their own.

Nowadays, the agents can iterate. It basically works like a genetic algorithm where it converges on a solution.

I’ve had pretty good success with DeepSeek making decent bash scripts. What was the script out of curiosity, I can try and see what DeepSeek spits out.

Exactly, unless it’s a cooperative all I care about is my compensation.

yup

yup

program tends to be the more general term while application tends to refer to a tool with a GUI