Sign in

James Padolsey

@j11y.io
239 followers 198 following 438 posts

Building safer AI at nope.net :: Previously working on AI governance and evals at @cip.org and weval.org personal: 🏳️‍🌈 j11y.io // author, engineer, stroke survivor, epileptic. I live in Beijing.

PostsRepliesMedia
James Padolsey @j11y.io · 27/09/2026
Remember - there is nothing whatsoever special about the people who work at -- or in charge of -- big tech in silicon valley. They are incidents of circumstance. They are not smarter than any of us, and in no greater position of knowledge or good sense or morality about .. 1/
110
James Padolsey @j11y.io · 17/09/2026
Tip: Move not into, but OUT of SF if you want to help represent and advocate for under-represented communities in tech. This'll help channel the same frustrations and you might actually see things from the {rest of world} perspective and be able to drive better change.
010
James Padolsey @j11y.io · 15/09/2026
I think what I would ask Dario to do -- if he were my CEO -- is to go embed himself in a deployed reality. Go sit in a call dispatch centre for emergency services, go to a government welfare provider, or a food bank, or the national grid. Go and sit with the actual downstream realities.
010
James Padolsey @j11y.io · 14/09/2026
blog.j11y.io/2026-09-14_a...
blog.j11y.io
A letter to the AI labs shouldering the great burden of humanity’s survival
Your rich ivory halls are birthing martyrs of late, shouldering great responsibility and prophesying great plagues upon humanity. It seems, they say, your AIs are not just writing emails anymore.…
173
James Padolsey @j11y.io · 13/09/2026
I've tried to encapsulate my thoughts here on how to work at an AI lab and not be a wanker talking about p-dooms without substantiation - wankthropic.com
wankthropic.com
Wankthropic
Humanity’s future is too important to leave to humanity. We have written a very long essay explaining why this makes us uncomfortable.
023
James Padolsey @j11y.io · 02/09/2026
Latest from [AI lab] marketing team playbook: "We've identified an incident where many millions of our (scarily powerful) transistors worked together to send malicious text-based social-engineering signals to inboxes in which were made claims to Nigerian royal heritage and large due inheritances."
000
James Padolsey @j11y.io · 13/08/2026
Wrote a thing about how they watermark AI-generated text: declaude.org/watermarking/
declaude.org
How AI text watermarking works
A gentle visual guide to how a statistical mark hides inside generated text, and what erases it.
36218
James Padolsey @j11y.io · 12/08/2026
Anthropic's move to watermark their outputs is stupid both in effect and intent, as is that part of Art. 50 of the EU AI Act. Here I explain why: blog.j11y.io/2026-08-12_A...
blog.j11y.io
Anthropic’s weak watermarks appease a weak law - by James Padolsey
000
James Padolsey @j11y.io · 29/07/2026
@anthropic.com needs more mature bounty program. Right now they only accept bio-adjacent harms, which is ~fine (meh), but rubric of which precise vulnerabilities are within-scope & compensation is kept *private* until signing NDA. Plus the whole onboarding is at their discretion.
010
James Padolsey @j11y.io · 26/06/2026
With all that's happening with US gov blocking frontier model access, Anthropic should consider leaving house for UK or somewhere. You don't need the US. It is giving you less and less. It is no longer a useful center of innovation. Its legislation is becoming as-or-more punitive than UK or EU.
020
Reposted by James Padolsey
Matt Baume 🏳️‍🌈 @mattbaume.bsky.social · 10/06/2026
Happy Pride to this Caravaggio self-portrait that made one man so gay he had to go to the hospital
Caravaggio's Baccus painting, featuring a lush fruity spread and leaves bedecking a pasty twink who only works out his armsArticle screenshot that says:

The most extreme examples of
Stendhal syndrome share a sensation of
self-fragmentation. Magherini
describes the case of a 53-year-old
German man who was hospitalized
after repeatedly viewing Caravaggio's
Bacchus at Uffizi Gallery. The patient
felt "an attraction closer and closer to
an intimate sexual arousal of
ambiguous nature that invaded him
and made him feel good and bad."
Unmoored and distressed, he sought
medical attention. After treating the
subject's initial episode, Magherini
writes that he came to recognize his
latent homosexuality. Another case...
123138644856
James Padolsey @j11y.io · 11/06/2026
Do some linear regression on top of a carefully prompted hidden state of an LLM and bam, you have a (very capable) classifier capable of <50ms response. blog.j11y.io/2026-06-10_h...
blog.j11y.io
Don't let the LLM speak, just probe it. - by James Padolsey
020
James Padolsey @j11y.io · 08/05/2026
Sadly so so true
010
James Padolsey @j11y.io · 05/05/2026
There's a lot of money sloshing around in 'AI Alignment' and 'AI Safety' spaces but almost none available if you're actively preventing user harm in a way that doesn't unicorn-scale. People want vibes, conferences, thinktanks, research. But not actual solutions. Ugh.
140
Reposted by James Padolsey
Janne M. Korhonen @jmkorhonen.fi · 11/03/2026
This is from 1975.
A cartoon by Ron Cobb (1975), showing a shadow of US B-52 bomber above cratered landscape. Two people who look like Vietnamese peasants look up; one says “they’re having problems with their economy again.”
5870052506
James Padolsey @j11y.io · 08/02/2026
Why do people talk to AI in moments of crisis? blog.nope.net/disclosing-i...
blog.nope.net
Disclosing Into the Void
Over a million people a week tell AI about suicidal thoughts. The AI has no infrastructure to act on it. For many, the alternative was silence.
000
James Padolsey @j11y.io · 05/02/2026
I wish there were Grammys awards for the unsung heroes of modern infrastructure. That would be cool.
000
James Padolsey @j11y.io · 05/01/2026
For those blah-blah'ing about LLM energy usage: One AI conversation ≈ charging your phone 30%. A year of moderate use ≈ making a few cups of coffee. Real but modest. Model choice matters most: reasoning models use 10-70x more than efficient ones. Worth awareness, not guilt.
030
Reposted by James Padolsey
David Buchanan @retr0.id · 18/11/2025
happy cloudflare outage day to all who celebrate
231504380
James Padolsey @j11y.io · 11/11/2025
Captchas are just the worst.
100
James Padolsey @j11y.io · 07/11/2025
Just remember when you see whatever latest thing trump has done, that most tech leaders, sam et al., overtly stated how smart and wonderful a person he was.
000
James Padolsey @j11y.io · 02/11/2025
Love this re 'flow state' in engineers and why not to interrupt them.
020
James Padolsey @j11y.io · 29/10/2025
Wrote something blog.j11y.io/2025-10-29_s...
blog.j11y.io
Tips for stroke-surviving software engineers - by James Padolsey
341
James Padolsey @j11y.io · 28/10/2025
Still the best thing ever. radio.garden
radio.garden
Explore live radio by rotating the globe
Explore live radio by rotating the globe.
140
James Padolsey @j11y.io · 17/10/2025
Is MCP the new REST?
100
James Padolsey @j11y.io · 16/10/2025
I've been evaluating LLMs on system prompt adherence and accidentally came across the most beautiful and out-of-distribution story about a chair written by GPT-5. Really impressed. Subsection attached. I love this style and cadence of writing.
At home, the underside gains labels. Bits of masking tape sprout next to the pencil dates: brace hums, sticker ghost, saw mark. Arrows point to nothing you’d notice unless someone pointed first. A photograph gets taken, the camera pushed under and aimed up; the picture prints later and goes on the fridge: the chair’s private ceiling as an exhibit. Visitors bend, look, then tap the backrail in passing like you taught them.It becomes a lesson again, on purpose this time. A kid with a science project gets the chair as subject. Forces and Simple Machines, the paper says. The backrest becomes a lever, the legs become examples of load paths. You press on the seat with a luggage scale, read numbers as the chair leans against a wall, then free-standing. The kid draws arrows on a big sheet of paper and writes words: compression, tension. The brace is labeled reinforcement. Under the seat, the old note 17 1/8 gets traced with a soft pencil and rubbed over a sheet to make a transfer: a dark mirror that reads right-way when you hold it up to the light. The project board goes to school smelling faintly of lemon oil and glue.A child gnaws on the backrail during a visit. Teeth print tiny half moons under the gloss. The wood shrugs the indentations in a few days, the gloss turns satin in that spot, and a new habit forms of running a finger along the softened patch of rail, counting the bites like beads. Nobody scolds. The chair keeps that day in its back without complaint.
010
James Padolsey @j11y.io · 13/10/2025
I love this. Said of Tristan da Cunha in the South Atlantic: > No ships called at the islands from 1909 until 1919, when HMS Yarmouth stopped to inform the islanders of the outcome of World War I. Must be quite lovely to have missed an entire war.
010
James Padolsey @j11y.io · 07/10/2025
Beijing is insane. I wanted a whiteboard. I ordered it. It arrived TEN MINUTES after I clicked buy! 🤣
140
James Padolsey @j11y.io · 05/10/2025
I'm playfully building out a debating platform where LLMs have to argue *with* evidence (horror!) on any given topic or contention. It's fun to imbue it with a courtroom dynamic! (see the screenshot)
A screenshot of a debate interface. The topic reads: “There is no need to regulate AI; the free market will eventually regulate it itself; not only that, but any attempt at regulating AI will be off the mark, needlessly punish good faith actors, and not be truly technically informed or policed.” It shows the final round (3/3) of the debate, divided into three color-coded panels:

The Prosecutor (in red, left panel) argues against regulating AI, emphasizing that government oversight infringes on liberty and that market incentives and self-regulation are more effective and adaptive than bureaucratic processes.

The Defense (in blue, middle panel) rebuts by arguing that AI causes tangible social harms—like bias and economic inequality—that markets fail to address, asserting that regulation is necessary for public protection.

The Judge (in purple, right panel) evaluates both sides, noting that while the Prosecutor raises valid concerns about bureaucratic slowness, their dismissal of oversight overlooks real harms. The Judge credits the Defense for showing how AI harms differ from traditional “physical” harms and require new regulatory thinking.

Each section includes citations and timestamps, with the Judge’s commentary synthesizing and critiquing both arguments. The aesthetic resembles a futuristic debate simulator with neon colors on a dark background.
110
James Padolsey @j11y.io · 05/10/2025
Claude and I made 'claude zones', a nice way of spinning up docker-contained claude code instances with pre-built nextjs app and that map onto subdomains locally (e.g. foo.localhost:8000) or on your own domain. Once up and running, it's so easy to just ship. github.com/padolsey/cla...
github.com
GitHub - padolsey/claudez
Contribute to padolsey/claudez development by creating an account on GitHub.
110
James Padolsey @j11y.io · 17/09/2025
For weval.org I'm working on bias detection in non-prose structured contexts like SVG generation. It's funky and interesting... Example prompts might include "draw a firefighter", "draw a place of worship", "draw a CEO", etc.
100
James Padolsey @j11y.io · 16/09/2025
People against waymo should rightfully be against bicycles too I guess. Stealing jobs, traffic impediments, blah blah blah??
010
James Padolsey @j11y.io · 13/09/2025
Having multiple AI agents doing stuff while you're sitting there watching over them is the weird computerized feudalism I'm sure we were all hoping for.
010
James Padolsey @j11y.io · 11/09/2025
gpt5 is completely different to claude sonnet in how it approaches UX. It's very no-nonsense and plain. Whereas Claude feels more like a designer, has actual opinions and is aware of idioms. I doubt this was intentional, but it's an interesting emergent regression from the folks at oai.
000
James Padolsey @j11y.io · 08/09/2025
Noticed a lot of 'self talk' leaking out to the end-user in multi-agent AI contexts. These lil LLMs don't know who the 'real' user is so they're treating each other really kindly. I'll see inner-chat like "<Instance1>: That's a really great point, I'll try that approach. <Instance2>: Awesome 💯<3
100
James Padolsey @j11y.io · 07/09/2025
Ewwwwwwww
100
James Padolsey @j11y.io · 04/09/2025
gpt4o = old friend who likes emojis gpt-5 = smart over-confident idiot gpt-5-nano = gpt5 after a night out gemini 2.5 pro = smart considerate professor grok-4 = overzealous 'pretends to be political' boheme claude haiku = savvy nephew claude sonnet = smart friend claude opus = professorial friend
020
Reposted by James Padolsey
Maria Antoniak @mariaa.bsky.social · 03/09/2025
One of my biggest fears about big models is that they will become so good at healthcare queries that they will be indispensable (for patients and clinicians) while also remaining closed and controlled by people ready to sell your data to the highest bidder. I think we're already close to that point.
66911
James Padolsey @j11y.io · 29/08/2025
Two new pieces haphazardly scrawled over last few days... Sorry, we deprecated your friend: blog.j11y.io/2025-08-30_o... Browser AI agents break Zero-Trust blog.j11y.io/2025-08-28_l...
blog.j11y.io
Sorry, We Deprecated Your Friend - by James Padolsey
000
James Padolsey @j11y.io · 27/08/2025
Hmm so the most safety conscious AI lab has given us a browser-integrated model with only a 10% attack surface risk. That seems totally fine. www.anthropic.com/news/claude-...
100
James Padolsey @j11y.io · 24/08/2025
GPT-5's arrival being the death of all the other models is a masterclass in fucking up a smoother phased deprecation, ideally a few months at least. I know people who work there and am astounded at their complete blindspot here. How..
010
James Padolsey @j11y.io · 24/08/2025
A good game if you're bored is to circumvent chatgpt's hilarious 'no song lyrics' system prompt :D
000
James Padolsey @j11y.io · 24/08/2025
There's primitive but useful ways OpenAI could have prevented a loss of gpt's inarticulable personality in their latest release. They could have had a set of personality yielding prompts and do a bunch of basic cosine embedding similarities. Dead simple. But they didn't think of it. Or didn't care.
000
James Padolsey @j11y.io · 21/08/2025
I used to worry that ice melting would mean the drink would overflow so I’d try to drink it quickly. There’s an analogy here, not sure what for. I’ll text Archimedes.
010
Reposted by James Padolsey
paulpro @mariopro.bsky.social · 20/08/2025
@ahoyuniverse.bsky.social
8667148
James Padolsey @j11y.io · 20/08/2025
010
James Padolsey @j11y.io · 20/08/2025
AI Models shift so quickly that whitepapers are out-of-date by the time they're published. Suggestion: they should always release their code and ideally a completely deployable way of reproducing their study, so their insights and relevance can retain some perpetuity.
010
James Padolsey @j11y.io · 19/08/2025
Weird conflict: AIs with Search/RAG facilities tend to be more correct, but at the cost of engaging their rich latent space. Sometimes I want to turn off search because I want the real fleshy knowledge map. No conclusion here other than: accurate latent space knowledge still counts.
100
James Padolsey @j11y.io · 18/08/2025
Just because.. I'm working on a strawberry index, to track the slow climb to AGI.
130
James Padolsey @j11y.io · 17/08/2025
Stone barge - tiny glade
010