simonw

12 ms·

241.
by simonw 1mo ago
"This is a highly selective analysis of Zitron's writings that focuses on a couple of mistakes" While reading this I was thinking it would be interesting to see just a few examples of Zitron predictions that he got right. Sin
242.
by simonw 1mo ago
As public companies, the megascalers publish pretty detailed financial reports.
243.
by simonw 1mo ago
I'm not financially literate enough to trust my own analysis of the numbers. Ideally I'd like commentary from someone like Bloomberg's Matt Levine, a genuine expert in financial matters who is also extremely good at explainin
244.
by simonw 1mo ago
> For any of his posts that I read, while there are numbers thrown around, the numbers don't actually connect to a coherent argument. In many cases, as we saw above, the numbers don't even really support his argument (such as a
245.
by simonw 1mo ago
Whoa, where did that come from? Is it confirmed to be an SVG?
246.
by simonw 1mo ago
I didn't want to shell out for Max again, so I piped the SVG created by Max back into Fable 5.1 at its default thinking level (of high): llm logs -cx | llm -m claude-fable-5.1 -s 'animate this' Here's the result,
247.
by simonw 1mo ago
Pelicans for thinking effort low, medium, high and xhigh (that xhigh one is pretty good): https://tools.simonwillison.net/markdown-svg-renderer#url=ht... I'm still waiting for effort max to finish. EDIT: I fixed a bug
248.
by simonw 1mo ago
I've used Fable for so much stuff. My experience has been that it can pretty much one-shot most of my complex problems, if I describe them clearly and provide a solid way for it to verify its work. I get punted down to Opus 5 occasiona
249.
by simonw 1mo ago
Bit of a discount if you're using caching: > same input and output prices, with cache reads at a quarter of the cost This should impact any long-running agent since subsequent calls can benefit from cached reads for previous transcr
250.
by simonw 1mo ago
Assuming this is accurate, the diagram showing movement between different AI labs is pretty interesting: https://select.keenable.ai/r/frontier-ai-researcher-moves-si... Suggests that when Meta AI started throwing aroun
251.
by simonw 1mo ago
How do you make a feature film - budget ranging from low millions to hundreds of millions of dollars - if anyone can copy it and view it without paying for it?
252.
by simonw 1mo ago
I thought this was excellent. It pulled together a whole bunch of interesting evidence, was very careful not to make claims that weren't supported by that evidence, and every time I had a question it answered my question in the next pa
253.
by simonw 1mo ago
Thanks, I passed that on to ChatGPT Work, should be fixed now (I'm on mobile so haven't tried it on desktop yet.)
254.
by simonw 1mo ago
It does, but the ChatGPT Chat VM isn't allowed to talk to the internet, doesn't get a filesystem that persists between sessions, and doesn't have the ability to control a web browser.
255.
by simonw 1mo ago
Codex requires you to have an always-on computer to run it on. ChatGPT Work Cloud gives you that computer as part of your existing subscription.
256.
by simonw 1mo ago
ChatGPT Work in the ChatGPT desktop app is effectively Codex with a different skin. It can operate on your local files in the same way. ChatGPT Work in the ChatGPT web and mobile app experience is very different. If you like, you can think
257.
by simonw 1mo ago
I told it "Design aesthetic should be technical docs, minimal flare" - full prompt here: https://news.ycombinator.com/item?id=49504625#49505116
258.
by simonw 1mo ago
Seriously, what's your definition of spam? Or an easier question: how am I gaslighting here? You should be able to spell that out. (You do understand that the top linked site here was built by me, but has no authorship information on i
259.
by simonw 1mo ago
I just signed into https://chatgpt.com using my burner free account and I don't see a "Work" tab. Do you know if free tier gets ChatGPT desktop app access to Codex and Work? And if that Work access is local-only o
260.
by simonw 1mo ago
In this case they really are, and you can read the guidelines that helped create the site on the site itself in the skills: https://codex-tool-reference.simonw.chatgpt.site/skills/site... Scroll down to the "Estab
261.
by simonw 1mo ago
How is this comment "spam"? Spell it out for me. (I think you may be confusing "spam" with "prolific". I write a lot.)
262.
by simonw 1mo ago
I think the most interesting skill here is the control-browser one: https://codex-tool-reference.simonw.chatgpt.site/skills/cont... It tells ChatGPT Work how to launch a Playwright instance via its Node.js REPL, and te
263.
by simonw 1mo ago
What am I "doing" here?
264.
by simonw 1mo ago
My comment here is entirely justified in this case. Someone else submitted my vibe-coded slop site to the front page without any context, and it accumulated enough votes to show up at the top. I could have linked to my blog post about it, b
265.
by simonw 1mo ago
See this comment for background (and creation prompt): https://news.ycombinator.com/item?id=49504625#49505116
266.
by simonw 1mo ago
> What are the open-source alternatives to this? OpenClaw, Hermes, and a dozen other Claw-like tools. They aren't trivial for non-technical users to setup, hence the demand for systems like Claude Cowork and ChatGPT Work and Grok Bo
267.
by simonw 1mo ago
The moment we establish a standard for offering a "optimized for scrapers" version of a site, people who do not want to be scraped will weaponise that to serve junk to scrapers... and scrapers will subsequently refuse to use it.
268.
by simonw 1mo ago
I think of it more as a SaaS hosting product providing some extra features that aren't in ChatGPT Chat - most notably a VM with a persistent filesystem, internet access, and a headless browser, made available to the standard OpenAI UI
269.
by simonw 1mo ago
It saves you money. ChatGPT Work uses up your Codex allowance. Luna can also be quite a bit faster.
270.
by simonw 1mo ago
Allowed to do what?

More ›