Posted by nateb2022 20 hours ago
It’s a new fundamental innovation in math and CS that allows large scale lossy compression of natural language another data formats in a way that is semantically queryable and cross-referenceable. It also manifests some form of emergent intelligence, likely evidence of the long posited link between intelligence and data compression.
The tech is awesome. It’s one of the coolest things I’ve seen in over a decade. The industry is kind of shitty, which is not unusual. The discourse around it is almost universally insane, crazy people arguing with crazy people.
Oh and get off the AI eco bullshit train. Look up the energy cost of AI queries vs driving or running a home air conditioning system. Feel bad about using AI? Skip that DoorDash order. You probably just saved the energy of 1-2 days of heavy Claude Code use.
Still kind of shitty. But if you really hate it use open models hosted commodity.
> If the Licensee or any of its affiliates operates a Model as a Service business, and the aggregate revenue of the Licensee and its affiliates exceeds 20 million US dollars (or the equivalent in other currencies) in total over any consecutive 12 months, the Licensee must enter into a separate agreement with Moonshot AI before using the Software or its derivative works for any commercial purpose.
"Our only modification part is that, if the Software (or any derivative works thereof) is used for any of your commercial products or services that have more than 100 million monthly active users, or more than 20 million US dollars (or equivalent in other currencies) in monthly revenue, you shall prominently display "Kimi K2.7 Code" on the user interface of such product or service."
Although as you mentioned, now there is a revenue cap before custom agreements must take place. Currently there is 7 providers on OpenRouter for Kimi K3 and they have all the exact same price unfortunately.
I wouldn't be surprised if they started offering potentiall bad quantizations with much reduced capability at lower prices (without telling the users, of course)
Other than that they track Tool call error rate and Structured output error rate.
I only discovered this today, and it seems like a good idea. What are the problems in practice?