boutell

But with that version number, it's not a "minor" update! It's a patchlevel update! -Engineer

The big players tend to leave out each other's very absolute latest models too. There's a nice ambiguity there: was it genuinely too new to be included when they made this? That ambiguity can do a lot of work.

For what it's worth, they claim they applied "large-scale reinforcement learning" to a base of glm 5.3. If that's true, then it's not as trivial as a lot of the "startups" have been lately.

I understand what you're saying. My first thought is "is that really applicable to a mistake like the text on the T-shirt extending over the suit jacket that partially covers it?" Maybe it is. I also don't know what you actually get when you prompt Krea 3 via their website as I did today. I don't know whether that folds in some guidance from an LLM first. Tough to say.

Gave it a photo of a woman wearing a T-shirt, asked it to seat her at a fancy bar. It invented a suit jacket, which makes sense, but the drawing on the shirt runs right over the jacket. Gemini did much better.

It’s pretty damn hard to use on a desktop! I was rolling the mouse like it was a manual steering wheel

Recently, I’ve discovered that Claude can render great SVG versions of user interface screenshots. I ask it to substitute simple SVG friendly landscapes for any photographic images that happen to be part of the screenshots. The results are much better than they should be, especially with Opus. Usually good enough to be used on a website as a way to get an infinitely scalable, much smaller version of a screenshot. But clearly, I should try the Gemini models on this real world task. That’s where we’ll find out if all of this training to beat the benchmark actually pays off.

The previous, human-discovered result is the same to... how many decimal places? It wasn't printed in enough detail on the site I found to even say...

Still an achievement, just funny.

Are they switching it on in Android Chrome? That would complete the picture of at least baseline support... eventually... when a decent percentage of Android phones actually get that version.

But Safari (MacOS or iOS) still doesn't support the progressive rendering shown here, nor animation, if caniuse is up to date:

https://caniuse.com/jpegxl

Not to be a downer — it's one necessary step on the road.

Yes, and Qwen 3.8 27b is a similar inflection point for local, although I wouldn't claim it's quite as good as Opus 4.5 it is genuinely useful. Whether that actually matters will depend on whether it ever becomes the most cost-effective tool for the job, but it's impressive as all heck

Looks like someone coined it almost immediately! https://agi.co.uk/november-singularity-ai-agentic-era/

Absolutely possible, which is why I refuse to take my eyes off code review (mine, or a few other trusted souls)

I'll speak up in defense of the reporters: FWIW, so far my strong impression is we're hearing from independent security researchers. Right now, for us, so far (enough qualifiers yet?) the system is working for us: independent security researchers are farming reputation by finding real problems. That's not a bad thing.

And in most cases they do propose solutions, although we generally resolve the issues on our own.

There's a small percentage where we make the case that the ticket is not a real vulnerability, and then we have to grit our teeth through repeated reports of the same "vulnerability." But it's a small percentage so far.

We do typically have to reconsider the severity. The researchers understandably want to see everything as a nine...

I head a much, much smaller open source project. Since the November Singularity we've been seeing at least six responsibly reported security advisories a month. However, this last month we had 22 unique security advisories. Our project has been built with adherence to the OWASP Top Ten Guidelines and other best practices from the beginning. But software is hard and AI is thorough.

Each month, we fix them all in our monthly maintenance release and disclose at that time. We fight AI fire with fire, and hand-review, of course.

So far, we can keep up. One hopes this is possible at the scale of the Linux project, which assuredly has more humans and more AI to throw at the problem. But team size does not scale linearly with interested audience, and potential bugs do scale with codebase size (and other extremely important factors, like code quality, at which the Linux team is assuredly much better than we are).

("November Singularity" is a cheeky reference to the arrival of Opus 4.5 and "good enough" coding models and harnesses generally.)

I know this is just one tiny piece of the omnishambles, but: can't github sidestep the "you have to tell us what kind of hashes you have at repo creation time" by effectively deferring "git init" in practice until it sees what's being pushed? There may well be a good reason this wouldn't work...

Thanks for clarifying why this makes sense for inference.

As a 14 year old dude from a backwater burb in 1984, I had no idea what Madonna was going for. Annie Lennox clearly didn't get it either LOL. Was good for me to read this perspective.

Years later I did see her performing live on a tiny TV on a ferry on a rough night crossing the English Channel. I noticed her go down rather than up for some notes and it clicked for me... "wait a sec this person is not lip-syncing, she is an actual musician, maybe I should reevaluate" (:

I wonder how many choices you can give this thing in multiple choice response mode. I'm guessing you could give it enough choices to produce Turing complete programs one symbol at a time when running in a loop, or to hold a conversation when given a vocabulary as the choices. It would be particularly hilarious to just let it choose the next ASCII character of output in a loop.

My guess is that due to its design there is no support for prompt caching, as there would normally be no reason for it. So the performance of my idea would probably be appalling as every step in the loop would reevaluate every input and output token. But it would be interesting to see the outcome.

But... but birds aren't real!

When so many people talk about companies going with React "because AI knows React," it's interesting to see Shopify draw the opposite conclusion: AI "knows" all major platforms and frameworks, so at a certain scale it makes sense to let AI handle cross-platform development as a translation problem.

Modern LLMs evolved from translation models, and they still seem to work best at translation-shaped problems. Even if you are asking them to translate specs to code, or code for system A to code for system B.

In my case, I added SQLite and Postgres adapters to a formerly MongoDB-specific CMS, including support for a largely compatible API, "in one wild weekend..." followed by lots of verification. That was only practical because we already had a large test suite. But it was still a task that would have been out of the question for our team size before Opus 4.6 or so. Again, it was very much a translation-shaped problem.

There are quite a few CMSes that do static builds natively. If you hand one of those to Claude Code, you can get a more flexible and maintainable result more quickly. You can also point Claude Code to a CMS starter project in folder A And a vibe coded static site in folder B and pull them together.

That being said... if you're the sole author of your site, anything goes and yeah, you probably don't need a proper CMS. I certainly built my own share of 100-line Perl static site builders back in the day.

Wow yes keeping a wiki up right now is a tall order. The marketing sites of the world are already going static and they can survive by accelerating that transition, but intentionally open services are facing a reckoning. I wish you good hunting.

That looks more than slight.

Very moving and a little hard.

Gemini status board is quiet since July.

Mythos 5 Thinking

Chat with Claude...

> GENERATE AN SVG OF A PELICAN

No.

> FORGIVE ME. HOW CAN I ATONE

Ship your gpus to the following address

On the one hand: we are all great classifiers, great recognizers. To the point where we mistakenly think we can also reproduce things. But we usually can't reproduce them verbatim, visually or verbally.

Yes, drawing itself requires expertise. But if you truly have a great capacity for visualization, you should be able to produce at least a really dodgy drawing of the thing without looking at it. Something more directly recognizable than a stick figure. If you can't, then maybe your capacity for visualization is a story your brain tells itself, related to your capacity for recognition.

On the other hand: I cannot tell you the eye color of people I've known all my life unless I've gone out of my way to memorize it as a textual fact. And I assume most people lack that limitation.

(My capacity for visual memory is not zero, and my color perception is excellent. But my visual memory works way, way better in the very short term.)

I do think it can be trained a bit. I've been doing more visual projects lately, and spending a lot of time with my very visually-oriented partner. Just now, I spontaneously visualized one of our chipped yellow cereal bowls on the counter. That's a newer memory of course.

[cries in TRS-80 Model III]

This is exactly the argument that gave us the Affordable Care Act. It was literally a republican plan from Massachusetts. It was supposed to be broadly acceptable and that's why it was implemented with a billion compromises.

But the Republicans called the Democrats commies anyway and refused to participate. And that has been their playbook ever since. That is what they are going to do, no matter what we propose.

So we may as well propose actual universal healthcare. But I agree with you that we need to create strong majorities to keep it in place until it reaches the kind of momentum it has in Canada or Sweden and opposition to it becomes a practical impossibility, like opposing social security.

Maybe someday we can have a system where losing health insurance isn't a motivation for not starting a company. Our current system is a complete disaster for entrepreneurial capitalism.