Viewing profile — simonw

An interesting factor of this is competition. If Claude was the only model family they could ship a change like this and users who want to cheat (or don't like watermarks for other…

> When a supported Claude model generates text, it weaves an imperceptible watermark directly into the text itself. You won’t see it, and it doesn’t change the meaning, quality, or…

Yeah, I buy the explanation that without encouragement Claude looked at everything in its existing training data and concluded it wasn't worth continuing to pursue the task.

Brute-force compute hasn't been an option for most of the history of science.

Given how competitive the space is any form of waiting seems like it has more downside than upside to me. If you have the new "best" model you may only have a few weeks of time in …

> Throughout this process, Jarred's input was mostly limited to sending Claude messages of encouragement (mostly variants of “keep going” or “believe in yourself”). This seems to h…

It's platform anxiety. Meta are still frustrated that they missed out on controlling the mobile platform and have to pay those 30% fees to Google and Apple. The metaverse was the s…

Pelican, rendered by Muse Glimmer on my Mac running LM Studio (with this model release: https://lmstudio.ai/models/muse-glimmer ): https://tools.simonwillison.net/markdown-svg-rend…

Are you calling me intellectually dishonest? Please justify that.

I, too, will reject the idea that "agent in a loop beats a talented human at long term coding tasks". LLMs amplify existing expertise. Give them to experts and you can get fantasti…

Abuse by police officers and other individuals with access to these systems is the answer to the stupid but common question "why should I worry about surveillance if I have nothing…

If you want to see contributions to Linus that used AI you can search for commits with tags that look like this in their commit message: Assisted-by: Codex:gpt-5.5 Assisted-by: Cla…

Here are a few examples you might find credible: - pola-rs/polars: https://github.com/pola-rs/polars/pull/26823 - ~2.68x median speedup of primitive-to-boolean casting credited to …

> AI-induced demand is overwhelming bureaucracies built for the analogue age—from Dutch municipal-tax appeals to the Canadian privacy regulator to parking-ticket tribunals in every…

They say that they are yet to see: > "Our AI grew sentient, recognised the climate crisis, and shut our operations down." I guess they missed the whole https://www.snitchbench.com/…

You mean in the comment where I suggested that I didn't think it was a very good product? I posted this comment because I didn't know what the author meant by ONA and had to look i…

What the heck? What product was I advertising here?

When this article talks about ONA it means https://ona.com/ - a cloud agent service that was acquired by OpenAI a couple of months ago. (I wouldn't suggest basing my evaluation of …

There's a big gap between being able to ask questions about something and understanding something. The more things you understand, the higher the chance you'll spot a situation to …

What are your favorite examples of him misleading his readers?

https://github.com/nsrht/time-travel-sqlite-debugger/blob/ma... works by checking the mtime on the database file (and -wal file) once per second and creating a backups/unixtime_dat…

Yeah, any time I see someone accuse Gruber of being universally positive about Apple I think they can't have been reading his work very often. Something Is Rotten in the State of C…

The bigger issue here is that he mislead (I think lied to, personally) John Gruber to the point that Gruber had to issue his first retraction in 25 years: https://daringfireball.ne…

Yeah, the lobbying is gross.

I think part of the problem is that they had been running that Artifactory configuration previously without any problems, and it gave them a false sense of security. Similar thing …