Also, what are your main use cases for artifacts? Do SPAs and HTML pages work best for them?

As a hobby I run https://smalldocs.org / https://github.com/espressoplease/smalldocs an agent agnositc artifact layer, so I'm looking for feedback on artifact use in general.

Here are some sample scientific SmallDoc artifacts:

Palmer penguins: body mass, flipper length and species - https://smalldocs.org/s/YrJDCfCD8-6j_uRE0pNCbI#k=odJRxWOIySV...

AlphaFold 3: capabilities, evidence and operational limits - https://smalldocs.org/s/rRRuByQhmhZNBH_4RNCx1t#k=S0Jhyf72Q6M...

1. Being okay with being "less" productive. I'm probably still 10x more productive than I was in the pre-AI era, but in the early days of AI I would have eight Claude sessions running. Now it's two or three and I take more advantage of the free time I have between Claude assignments. My bosses are still happy with my output and I'm not being burnt out.

2. Very personal, but building my own tooling really helped. I built https://smalldocs.org / https://github.com/espressoplease/smalldocs (a document creator for AI agents) which allows me to take a lot of work outside of a chat history. E.g. I might make 10 MRs and then say "smalldoc me a list of the MRs" and I have that in my library, or if I'm considering various architectures I'll ask "smalldoc me a simple overview of the three architectures with mermaid diagrams" and then I can review the agent's thinking in a cleaner way.

Creating something visually interesting Invisible-Cities-ish is basically a task with a billion or more satisfactory outcomes. You don't have a precise view on what makes an invisible city a good one as long as it has all the fancy looking bells and whistles. This means the AI can generate a bunch of tokens which as long as they technically work (which it seems the max-thinking did not fully) produce a "good" result. This makes you think - wow AI is impressive. However if many things in the system really mattered, it might have been much less successful as the AI may have made poor assumptions. This was basically a task that given a large volume of tokens and technical competence it's fairly easy for the AI to do a good job at because we don't evaluate it with any depth/precision. I think the same is true for music, where there are basically an infinite number of combinations of sounds, beats, melodies, etc. which are pleasant to hear - and songs aren't mission critical - so alignment is not a big issue. Generative AI is good at these tasks.

The result is: https://tradernews.fyi

Trader News is the Hacker News interface you are familiar with, with a difference. Above the article area is a live feed (up to a ~30 min delay) of market data. I show the major indexes and every company within them. E.g click on Nasdaq-100 to see how the index performing, and then scroll the list of companies and click on META to get more detail on its specific performance (and links to Yahoo Finance + it’s investor relations). This gives you a single place to read finance related articles and stay in touch with the markets.

The CSS for Trader News is linked to global market performance. On days when stocks are down, Trader News is red. On days where stocks are up, Trader News is green.

I have an agent running a 24 hour cron job to research articles and post them to the site to keep it a stimulating place to read. I will tone this down if user submissions pick up. If there are sources you think I should include, please leave a comment.

I also added a basic “chat” panel with polling every 30 seconds for casual market related chat.

It’s open source and I’m very interested in feature pull requests. The code is available here:

https://github.com/espressoplease/trader-news

Public anonymised analytics are available with a 15 min delay here:

https://tradernews.fyi/analytics

[1] https://github.com/shawwn/sharc

Comments URL: https://news.ycombinator.com/item?id=49828019

Points: 26

# Comments: 18

I think this is a bad take... if you are going that long without noticing, you are (hopefully) delivering end user value all that time. That is the main driver of code. You can normally dig/hack/rearchitect your way out of an ugly code situation. If you've been building value for the last year based on the hacky code, that's a win.