"Silicon Valley startups are setting up open models, with some operating on shoestring budgets because of limited interest from VCs"

https://www.wsj.com/tech/ai/the-race-to-build-an-american-alternative-to-cheap-ai-from-china-2e99a28a

See the graph in this article

https://thenewstack.io/open-weight-models-frontier-costs/

I think that's more where we are LM.

Kimi is close to Claude and 1/3 the price and other OW models (DeepSeek) are both less good and less expensive. Depending on the task, a model that matches Claude from a year ago might be enough.

https://www.together.ai/blog/kimi-k3-vs-claude-fable-5-on-deepswe-cost-and-coding
https://www.unsupervised.com/resources/what-agent-work-costs

They aren't "free" in either of the senses that Linux is free, but if you are talking about Linux from a TCO perspective, then it isn't free in the beer sense either.

> artificial sweeteners

QuoteStratified analyses indicated that the heightened risk of cognitive decline associated with high artificial sweetener consumption was particularly pronounced in adults younger than 60 years. In contrast, older adults did not demonstrate significant associations between sweetener tertiles and cognitive trajectories. Furthermore, diabetes status served as an important effect modifier. Among individuals with diabetes, high intake of total sweeteners, as well as individual agents such as aspartame, acesulfame k, and saccharin, was strongly associated with accelerated memory and global cognitive impairment.

Let's break that down.

The group that saw significant cognitive decline was under 60.

Question: what percentage of people experience significant cognitive decline before the age of 60?

Answer: not many.

So what that means is that if it just so happens that randomly a few of those people also ingested a lot of artificial sweetener, you might have a p-value low enough to say this is significant.

There was a study a few years back that got people freaked out about too much exercise. There were thousands of people and they looked at all-cause mortality. It turned out that above some very high number that almost nobody needs to worry about (something like 1000 minutes per week), death rates were higher than in the moderate category. The problem is that the denominator is tiny for that group and this was all cause mortality and it was based on three people. So if literally one person in the next category down (500 minutes say) had gotten hit by a bus, that would have flipped the curve.

So without seeing absolute numbers, this is meaningless to me (as a person who ingests almost no artificial sweetener, so I have no skin in this game).

> individuals with diabetes

But guess what? They did not get Type II diabetes from eating a lot of artificial sweetener. In most cases they got it from eating a lot of high-glycemic-index foods, lots of snacking and little activity. Now AFTER they got the diabetes diagnosis and had the fear of god put into them, they cut back on the sugared drinks and started drinking lots of artificial sweetener.

I know someone who fits that exact profile. Perhaps the most sedentary person I know, she was on a constant Mountain Dew drip. Then got diabetes and several other health issues. Now drinks large quantities of artificial sweetener.

My guess is that the causality runs in reverse: diabetes causes high consumption of artificial sweeteners.

Bah!

ICE is still 2/3 3/4* of the market. I wonder what that looks like if you take out the US

*p. 19: 21.9 out of 82.7 million

I am hoping for more of a MS/Apple vs. Linux situation.

I recently heard this described as another Apple vs Android... OpenAI and Anthropic are the premium Apple equivalent. Open weight/source models are the Android equivalent.

>Ford Maverick base is $28,100.

If it's not a dead loss, there can't be much margin in it --so no deals. I'd bet there's a good chance of it selling above msrp when it first hits the dealer lots.

Ford Maverick base is $28,100. It comes with 4 doors, stereo, and Apple Carplay. Slate is $26,400 with 2 doors, no infotainment, and unknown reliability from a startup company. I reserved an allocation on Slate's debut, but the $7500 tax credits are gone. Slate is going to be a tough sell.