simonw

17 ms·

331.
by simonw 2mo ago
US golf courses hit 531 billion gallons in 2024 (1.63 million acre-feet) https://www.gcsaa.org/who-we-are/media/news-release/2025-new...
332.
by simonw 2mo ago
Drawing things with p5.js.
333.
by simonw 2mo ago
You can use their AdSense token to see other sites by the same operator. This one fits their pattern: https://builtwith.com/relationships/tag/CA-PUB-1851068468056... Sites like: chatgpt4.org picai.com cha
334.
by simonw 2mo ago
The FAQ says: > We are not the model's creators — we're fans of it, like everyone else trying to guess who built it. Seeing as you can access the model for free in a number of different ways right now - including via OpenRouter
335.
by simonw 2mo ago
What advantage does blockchain ticketing hold over a centralized database of tickets?
336.
by simonw 2mo ago
The "until at least Nov 21st" thing presumably mainly affects teams that pin to GPT-5.6 Sol (maybe after extensive testing) such that they won't be switching to GPT-5.7 or GPT-6 or whatever new model is released between now a
337.
by simonw 2mo ago
> The bytes pass through a small Cloudflare Worker on the way, because the free r2.dev URL is rate-limited. For a 40MB file I suggest hosting it directly on GitHub Pages - that's effectively a free CORS-enabled CDN and supports HTTP
338.
by simonw 2mo ago
"static datasets with range payloads small enough to fit into a web response" fits a lot of workloads. I expect that if your overall data is less than a GB this trick will work really well for you.
339.
by simonw 2mo ago
The latest ChatGPT voice mode is really good at being interrupted - I'll often say "no, no, no, that's too much information" while it's talking to stop and redirect it.
340.
by simonw 2mo ago
Because it's the most clear indication Anthropic have given of policies around AI writing with respect to their own company. Things like: > Not allowed: Prompt: "Write my answers to the application questions for an AI safety r
341.
by simonw 2mo ago
I expect the energy usage of a thousand people all running their own local models is significantly less efficient than those same people sharing a model hosted in a data center.
342.
by simonw 2mo ago
Earliest capture on the Internet Archive is July 15th so I think this is recent: https://web.archive.org/web/20260715131001/https://corkmac.a...
343.
by simonw 2mo ago
The app is Cork: https://github.com/buresdv/Cork/network/dependencies It depends on Factory: https://github.com/hmlongco/Factory/blob/main/CLAUDE.md Sticking to a strident
344.
by simonw 2mo ago
Anthropic's guide for job applicants on how you should use Claude when applying for a job there is relevant here: https://www.anthropic.com/candidate-ai-guidance
345.
by simonw 2mo ago
The word "Your" was dropped from the start of the title for some reason.
346.
by simonw 2mo ago
I generally let them run riot.
347.
by simonw 2mo ago
Any model with a shell tool or a Python tool or a JavaScript tool can access the current time.
348.
by simonw 2mo ago
We have a Samsung Frame TV. I told Codex to scan our network to find it and then build a custom tool for updating the image gallery that it uses when it's in "art" mode. It did that, and now I can tell a Codex session control
349.
by simonw 2mo ago
DeepSeek/deepseek-v4-pro and z-ai/glm-4.7 are the two models that hallucinated answers - notably, these are text input only models. It looks like the PDF was being treated as a sequence of images. Testing this via OpenRouter seems
350.
by simonw 2mo ago
5.3 isn't available as open weights yet, and only became available via API three days ago. Prior to that the only way to access it was via a Z.ai subscription.
351.
by simonw 2mo ago
I just went looking for a TikTok I'd seen about this and discovered that UK supermarket class distinctions are an entire genre on TikTok. Here's the one I liked: https://www.tiktok.com/@willdhallam/video&#x2
352.
by simonw 2mo ago
Sometimes you don't even have to travel to visit a "foreign" supermarket. If you live in the SF Bay Area there are dozens of great opportunities to visit international supermarkets. I'm particularly fond of the Korean su
353.
by simonw 2mo ago
It's that I asked for an SVG of a circle and it created a beautiful gradient filled animated technical diagram. I wanted something closer to <svg><circle cx="50" cy="50" r="40"/></
354.
by simonw 2mo ago
I enjoy it. I have a small army of weird little golems that mostly do what I tell them to do. What's not fun about that? The other day I asked one of them (running directly on my laptop) to "draw an svg of a circle" and it &q
355.
Ask HN: If you're burning tokens at scale, what are you using them for?
5 points by simonw 2mo ago|4 comments
356.
by simonw 2mo ago
"Combining them into one heroic speedup would make a better headline and a worse benchmark." "The machine immediately taught me that capacity estimates are just admission tickets." "Useful in production, poison in a
357.
by simonw 2mo ago
Scaling ChatGPT and scaling GitHub are very different problems. For LLMs, prompt handling is effectively stateless. They do try to route follow-up prompts to the same cluster to benefit from prompt caching, but those can be effectively shar
358.
by simonw 2mo ago
I tried "render an svg of five intersecting squares. don't overthink this" and... it overthought it: https://static.simonwillison.net/static/2026/cors-chat.webp
359.
by simonw 2mo ago
I've been passing "reasoning": { "effort": "none" } to LM Studio's Responses API and the resulting response doesn't include any reasoning chunks, so that seems to work.
360.
by simonw 2mo ago
Gemma 4 E2B, E4B and 12B unified accept audio - here's a recipe using MLX that can use it for transcription: https://simonwillison.net/2026/Apr/12/mlx-audio/ Only up to 30s though, and the larger 26

More ›