4.1 consistently surprises me in capability for the price. And I don't think I'm the only one. It's been dominating the leaderboard at OpenRouter, and I just got an email today from Fireworks saying they were _raising_ the price by about 30%. I'll probably switch, because their infra doesn't support being the highest-cost, but it's still telling.

I tried it a few times and liked the speed, but often found it ended up looping, i.e. repeating the same token sequence (e.g. the same sequence of 5 paragraphs) over and over again until it hit the max output limit. This doesn't end up happening every session, but does every now and then.

My impression of DSv4.1-flash was very positive aside from this. But that was enough for me to stick with GLM-5.3(-flash), which both gave me consistently great results

I was using a vibe coded bare bones harness. I was wondering if this was normal from DSv4.1-flash, or if its my harnesses fault.

I've had that looping issue with open models too. But never 4.1. I wonder if it's a model + harness combo? But yeah, one loop issue and I'm done with a model forever.

Harness. Especially if a tool call error doesn't say what to do next and the model is not RL'd with that tool, a retry storm is common.

So if you use MCP a lot, simplify the params, be more lenient on validation and rework the errors.

It is quite good with shell.

Wow, I'm surprised you are saying GLM 5.3 Flash is more capable. Isn't is like half the price of 4.1 Flash?

Only because we understand everything _except_ how it works.

how can we understand absolute everything and still don't know how it work ?

Trial and error :)

I wouldn't say that is "understanding". We must fully be able to explain to say that we understand it.

There are two types of understanding: knowing how a candle works and knowing how a candle works.

Is Google still chasing frontier? Seems like they haven't had a "Pro" model in forever. I think a good niche for them would be right where they are now.

They are. They were supposed to release 3.5 Pro over the summer but haven't because of persistent architectural/technical issues allegedly. Which is better than releasing it in that state imo.

Their AI leadership team has taken some hits recently too, in the form of departures. I believe when they get their bearings they will be competitive again. 3.8 Flash has been a great model for me.

These are native speech-to-speech models, so I think they've decided not to do that anymore.

Same. I use Luna for research/scout/test subagents, Sol for coordinator, Terra for delegate, and Astra high for review and simplify. Even with Astra in there, it’s fresh context, and I’m consistently amazed by how far I can stretch my $20 subscription with really good results.

Why do you use Sol for the coordinator? Does it also do planning and design? Because otherwise you are cranking Sol turns on very mundane orchestration jobs.

what harness?

Raw isn’t an image though, it’s pre-image. And we have an open raw format already. DNF I think. With pretty decent adoption.

You are thinking of DNG which is more of a container format. It can use multiple image formats internally to record the actual image, one of them being JPEG-XL.

iPhone raw images are actually a jpeg xl inside the DNG.

True. I would love to have lossless images (and video) out of my devices still. Not as a replacement for RAW but as a middle ground between RAW and lossy JPEG.

Not for quality reasons primaily but because keeping the first stages of the processing pipeline and the archival branch lossless solves some headaches.

We have to apply some weight to the fact that one is open, one isn't, right?

EDIT: Oh! Of course. AVIF is based on AV1. For some reason I was thinking HEIF the entire time I was reading the article. Man... withdrawn!

Really good read, btw. Very good points.

AVIF is royalty-free, & SVT-AV1 and libaom are open source.

Edit: saw you corrected. Much appreciated!

SVT-AV1 is also more than just open source in name only; development and discussions happen in the open; with many community forks that’s now getting upstreamed/merged.

It is a true open source project; not just an open source license.

SVT-AV1-PSY was the first fork, and I created it. When the project matured, the maintainers reached out directly to get things merged – I can confirm it is a great project with great people who are very easy to work with.

AVIF _is_ AV1 in the HEIF container.

(.heif is sometimes used as a file extension generically, but HEIF is itself a container that can support various payloads.)

Both are open I think.

Same. Have a Kickstarter one that’s been going strong ever since.

Too wide to fit in anyone's pocket, but the ratio is the sine of 45 degrees! Seems like a very non-Apple thing to do, honestly. My jeans don't care about math.

It’s the size of a passport, which does fit in most pockets.

I just put a passport in my pocket, and it did fit. It's a lot thinner, and bends a bit... but yeah, maybe it will. Interesting.

For me it's fitting the damn thing in my pocket. And unfortunately, width is the main limiter there, making this far WORSE than any other, despite the smaller screen. I can't imagine what this thing would do to your jeans. Apple needs to sell tiny man satchels with this thing.

Agree. The comparison to a passport is ludicrous since a passport can be curled around your leg when it's in a pocket where this obviously cannot.

Passports aren't actually very comfortable in the pocket even then. They are clunky things.