kzrdude

Yes, "this one is obsessed with the golden gate bridge" will be for real this time.

Just trying to understand, https://opencode.ai/docs/go/#privacy currently says DeepSeek V4 Flash has 0 days data retention.

> DeepSeek V4 Flash: ZDR agreement is renewed monthly. The current agreement is valid through August 31, 2026.

Is there other info I should be aware of w.r.t data retention with opencode go? It's hosted in China, so other middlemen may be active (I doubt it, but possible)?

The investment in bigger machines at the fab might set you back billions. I don't know about the lithography technology either, how easy you can scale it to larger wafers?

We don't use these models by way of "one shot" so I don't see why it's relevant. It's clearly useful to let them iterate.

People just have different tolerance. Coffee strictly before noon works well for me.

It has produced some locally useful models.

Do mathematicians have the right to say "no AI PRs please, the volume is too much" just like how some open source maintainers do it? I guess they feel a loss of control, there is no way to turn the hose off.

Thinking of this a little bit with the perspective of every new proof as a burden, dumped for review by actual mathematicians.

It's summer vacation time.

The thinking trace was (preview) frustrating to read, I think I'd prefer a summary view of it at this point.

Then DS V4 flash is pretty relevant, because it's pushing prices down.. as well as being self hostable with a large enough rig.

Thanks, any update here is important. I use them because of good data policies..

I still find this today:

> The plan is designed primarily for international users and provides stable global access. Your data will not be used for model training.

Thanks. These things change day by day I guess, AI is just moving fast (and I'm on vacation).

GPT 5.6 Luna is a new model in Go since I last checked, for example.

Not important, just a curiosity. I would expect that kind of knowledge would be pretty well burned into the weights, but what do I know about LLMs. Gemma 4 can answer this question well.

Open weights means that there is a free market on providing inference using this model. Some of them will offer good data protection. (Well, hopefully)

Does it still say that it's an anthropic model, when asked? I would guess new post-training has fixed that.

Yes, and even if they seem to have updated endpoints in place they should assign the service a version number.

Opencode-go gives you $60 worth of DS V4 api usage for $10 per month. Right now I think it's hard to exhaust that when using flash exclusively, and plain API use might even be cheaper! Anyway, for DS usage it's a good deal.

Do you notice any improvements with this update?

Nitter link: https://nitter.net/matthen2/status/2082566186785480708

DS V4 flash has been updated today!

https://api-docs.deepseek.com/updates/

The update seems to say that it's a re-post train of exactly the same model.

And Infantino has been in high leadership in UEFA (not president), then FIFA

The point of Fifa is not to make money. The point of it is to further the interests of the sport.

Besides, your tally of sports incomes is wrong since you didn't include UEFA revenue, which is also football money, in the count.

A common factor seems rather that international sports orgs are often very corrupt. Would be interesting to understand why it happens systematically. Lack of good governance structures must be part of it.

How it works in practice is that some release flows add wheels for different platforms, as they get ready, separately.

Part of the problem is that features are not added fully

This here should not need to happen:

> document that moved-from objects cannot be used

It's moved from, so of course it cannot be used. Shouldn't have to add an assertion in every method to guard a fundamental invariant.

Go is great if you prefer deepseek V4 flash. Then it's very hard to use up the allowance.

Maybe this is relevant? Just in case

> For autonomous subagents with tool calls, code execution, or multi-step reasoning: set thinking_level to "medium" or "high" to prevent premature tool termination.

I just happened to see that in the docs: https://ai.google.dev/gemini-api/docs/latest-model

He's already proven repeatedly that he'll break out of containment

It's smarter than DeepSeek v4 Pro (preview) on several benchmarks, like HLE.

Well, you can use the google models from Pi. Go to aistudio.google.com and set up an API key. There is a free quota, it's relatively large for the Flash Lite models.

...and last time I looked the limits were more generous for Gemma 4 there, but they have been tightened a bit. That's how it goes, always changing.