9/18/26
8/24/26
I have a bunch of links open that I read a while back and didn’t take notes on. I want to close them out, so I’m just going to dump them here without notes.
- https://geohot.github.io/blog/jekyll/update/2026/05/24/the-eternal-sloptember.html
- https://hbr.org/2026/08/how-much-time-do-your-employees-spend-botsitting
- https://www.honeycomb.io/blog/spend-more-time-talking-to-humans
- https://blog.florianherrengt.com/ai-removing-middle-class-software-engineering.html
- https://brettcodes.com/im-done-using-ai/
- https://daringfireball.net/2026/08/anthropics_watermark_text_adulteration_in_claude_is_a_perversion_of_writing
- https://declaude.org/watermarking/
- https://daringfireball.net/linked/2026/08/11/economist-ai-writing
8/14/26
- I didn’t realize what kind of threat “you’ll be replaced by someone who uses AI” was
- Tabs, Spaces, Hand tools, and Seat belts
7/20/26
5/19/26
5/18/26
4/29/26
4/24/26
4/14/26
3/26/26
3/24/26
- Programming as Theory Building
- Generative AI as compilers doesn’t make sense to me
- I think prompting LLMs short circuits my thinking
3/12/26
3/6/26
3/4/36
3/2/26
2/24/26
2/19/26
2/18/26
2/17/26
2/16/26
2/14/26
2/10/26
2/7/26
- Generative AI as compilers doesn’t make sense to me
- Don’t tell me how many lines of code your LLM wrote
2/2/26
1/19/26
1/12/26
1/4/26
1/3/26
- LLMentalist
- Hallucinations
- AI summaries are almost certainly unreliable
- ChatGPT Survey - Performance on NLP datasets
- Modern software quality, or why I think using language models for programming is a bad idea
- No Silver Bullet
10/18/25
10/14/25
10/13/25
9/29/25
9/27/25
I should make a text expansion for use with AI prompts like ChatGPT:
- do not speculate if you do not know. keep answers concise. don’t use platitudes. don’t pretend to be a person.
9/12/25
- The CoPilot productivity paradox
- Thoughts on AI and AI-veganism
- Book recommended to me from somewhere…
- Where’s the shovelware?
9/10/25
9/5/25
9/1/25
- LLMs are Slot Machines
- The Futzing Fraction
- AI software assistants make the hardest kinds of bugs to spot
- Stack Overflow data reveals the hidden productivity tax of ‘almost right’ AI code