Show newer
Simon boosted
Simon boosted

After two days with Claude Fable 5 the best way I can describe it is "relentlessly proactive" - here's an example where I dropped in a screenshot of a bug and it span up custom CORS Python servers and used pyobjc-framework-Quartz to capture screenshots simonwillison.net/2026/Jun/11/

@timkellogg.me interesting, @simon 's fable blog post is also about a kind of sandbox, he didn't mention this occurring

Simon boosted

afaict the thing i'm doing is creating a sandbox for an agent, and Fable is probably hypothesizing about how to break the sandbox it created which, that's *secure* behavior. uuuuugh! support.claude.com/en/articles/...

Show thread
Simon boosted

Blogged about the time I doubled our users by doing proper engineering instead of React slop

mohkohn.co.uk/writing/html-fir

@HalvarFlake sounds like it could be relevant for sciop.net/ (though this is basically me telling you to start seeding a torrent...)

Simon boosted

Ok, I have a 2048x3072 pixel video that shows the differences in training dynamics between Muon and AdamW, and they are fascinating to watch. Where can I upload/share such a video without it getting compressed to shreds?

@sung.kim.mw a mecha suit made of GPU cooling fans would work too

Simon boosted
Simon boosted
Simon boosted

I may have finally found the Python-in-a-sandbox solution I've been looking for... here's my latest experiment, this time running MicroPython in WebAssembly inside my Python applications simonwillison.net/2026/Jun/6/m

@simon I was curious about Monty as well, I checked out your research post about it. It might add support for classes in the future, but it looks like it's not planning to support third-party libraries which would restrict plug-in development.

Simon boosted

i’ve noticed that Ai2 has shifted most of its press releases into public benefit point solutions, like OlmoEarth it’s all good stuff, but i imagine it’s a bit underwhelming for researchers like @natolambert.bsky.social@bsky.brid.gy who were trying to push the SOTA of American open weights AI

Simon boosted

Francesca Albanese is trying to stop a genocide.
The United States and Israel are trying to stop her – by destroying her life, her career, her family, and her voice.
The US has sanctioned UN Special Rapporteur Francesca Albanese, 8 ICC judges, 3 court officials, and Palestinian human rights organisations Al-Haq, Al Mezan, and PCHR. The message is chilling: investigate genocide, speak up for international law, and we will destroy your life.

petition.qomon.org/0b54a0b8-de

#news #gaza #genocide #icc

Simon boosted

Our AI development stack is shaping up nicely!

As a base, using any programming language:
* Sandcat to securely run agents, without having to answer tedious questions
* Orca to create deterministic development flows

Then for Scala/Java in particular, we've got:
* Metals MCP to avoid bash round-trip when compiling/testing/browsing code
* Cellar to fetch metadata of external dependencies
* Scala Skill to steer the agents towards producing maintainable, concise FP code

Simon boosted

We built four malicious skills to test whether skill scanners actually work. Three took less than an hour to conceive and implement. ClawHub, Cisco's skill-scanner, and Vercel's skills .sh marked them as safe.

In our simplest bypass, we prepended 100,000 blank lines to a malicious skill. ClawHub's scanner truncated the file before reaching the payload, then marked the skill safe. blog.trailofbits.com/2026/06/0

Simon boosted

In a world where a spurious regression effect in 2006 can be a lever for justifying narratives that create widespread violence and loss and grief in our communities twenty years later, you need people like me and this economist and everyone like us. I hope enough of us survive this era and get to keep doing social science.

Show thread
Simon boosted

Here's a metric that I call the 'business-to-industry' index.

It consists of the US share of world income relative to the US share of world energy use. Two centuries ago, the US was industry dominated. Today it is business dominated.

economicsfromthetopdown.com/20

Simon boosted

Auf dem Feld ⬆️ (hier ein anderer Blickwinkel und ein paar Stunden früher) hagelt es nur so von #Mohn. Als ob da am #Mohntag ein Mohnsun da eine Mohnokultur hätte errichten wollen.

Geniesst den Abend!
#SonneTeilen

Show thread
Show older
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.