jofzar

> You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache.

Direct text from tweet

> . I do have a 4k TV, but it’s not extremely large, and while I can tell the difference in resolution between SD and HD, for almost everything, 720p seems to be good enough for me.

I am actually jealous of you, I can tell the difference between 4k low bitrate and 4k high bitrate, 1080p and 720p are like mud to my eyes.

Interesting way to release this, I'm not sure how to say this, but this has a real "enterprise"/old school taste to it verses how all the other models/AI companies are releasing things.

Very refreshing.

I will also say, not having to get a team to build this for you, having to get business justification from your team for that teams hours, then spending time revising and testing that out and then you have to "prove" that it's worth having in your feature as a AI cost

Versus

"Let's put jev here and see how it works, if it works, then fantastic let's build a business case"

> But then again plain LLMs have been perfectly cheap and serviceable as zero-shot classifiers for quite some time now, so again I'm back to my internal screaming.

It's the scale of "perfectly cheap", jev (specifically) is so dirt cheap and fast that you can throw it at things that should not be justifiable in the past and you barely have to do any work other then quick testing.

That's why it's useful, it's for when you don't know what needs to be classified or if you want to offer this classification models to your users.

Testing I have seen from internal teams have shown it's still the best and cheapest model so far for pure classification work.

If you are a parent, please get your children the hpv vaccine, it's saved so many lives here in Australia.

They will follow through with it, it's because they want to get the money of the people who embedded it into their system and lazy to change it.

It's incase you want to recreate this scene from who framed Roger rabbit in svg form, duh.

https://simonwillison.net/2026/Sep/18/the-creative-spirit-of...

"Traffic routes via Cloudflare's network"?

5 years ago imo it was less relevant, if you showed me this then I would be like "cool tech, but other then a specific scenario, why would I care"

It's email reputation, it the always the answer with this kind of stuff.

My guess is that the bounce rate got too high and bot farms were using iCloud addresses like this.

I had a TP-Link deco, which while great as mesh, you couldn't limit the speed of certain devices. Which was again fine, except when the PlayStation 5 in a room could literally saturate the whole 250mbs bandwidth of my internet and when it would do an update would make our whole houses internet unusable.

It's seriously impressive how it wasnt possible to set per device speed limits or even QoS which has been a thing since like 2007

Not related to exactly OP post, but it's pretty amazing you can see the updates to LLM models "design" beliefs by the blogs that get posted here.

I'm already sick of this current look of the hard squares and solid colours.

Yeah I'm using this and have been happy with it.

It's silicon valley, literally everything is for pay or be homeless

I use luna to schedule my work day for me and then put it in todoist.

I hate to say this and this is because I fucking despise meta. But between DeepSeek and Meta, and trust they handle the training data correctly, I trust meta.

No I haven't had time to spend my afternoon rock climbing while ai generates documentation.

It's infinite work, I just did more work while codex was doing it's thing in the background.

Lmao I love their video with the idea that people will be able to do their hobbies while ai does their job.

Surely Alibaba is leading by example here by reducing work hours per week while keeping pay the same right? Right?

Honestly just a bit burnt out on posts like this

Wasn't it down last weekend?

Mate needs to move down to Australia, he would have been promoted after showing that off.

We had googlebot blast a random customer system and almost cause an outage, this is when I first learnt that google will use it for AI training also. It's honestly kind of frustrating also because you then search on it and theres (was) nothing on how you are meant to "correctly" tell google to fuck off, and not use it like that.

They actually learn it, what's more crazy in Australia is that we have videos of some birds (bush turkeys/swans) using zebra crossings to cross because they know it's safe.

Huh I'm surprised these aren't more common around the world? We have these in Australia and they work great

Microsoft/azure has a big advantage because companies will literally pay extra then "free market" cost of open models purely for the knowledge that the won't fuck around with their data retention policy.

It's $0/close to 0, they aren't at 100% demand so any leftover compute is "not spent".

The cost they are quoting is API cost, so it's already inflated on that.

Yes but grok has the literal track record of our of the box uploading your whole codebase, secrets included to a remote box.